PRELIMINARY

THE EFFECTS OF MARGINAL TAX RATES:

EVIDENCE FROM THE INTERWAR ERA

Christina D. Romer

David H. Romer

University of California, Berkeley

February 2011

We are grateful to Alan Auerbach and Emmanuel Saez for helpful comments and suggestions, to Maria Coelho, Jeanette Ling, and Priyanka Rajagopalan for research assistance, and to the National Science Foundation for financial support.

THE EFFECT OF MARGINAL TAX RATES: EVIDENCE FROM THE INTERWAR ERA

Christina D. Romer David H. Romer

ABSTRACT

This paper uses the interwar period in the United States as a laboratory for investigating

the incentive effects of changes in marginal income tax rates. Marginal rates changed

frequently and drastically in the 1920s and 1930s, and the changes varied greatly across

income groups at the top of the income distribution. We examine the effect of these

changes on taxable income using time-series/cross-section analysis of detailed data on

income and taxes by small slices of the income distribution. We find that the elasticity of

taxable income to changes in the log after-tax share (one minus the marginal rate) is

positive but very small (approximately 0.2) and precisely estimated (a t-statistic over 5).

The estimate is highly robust. We also examine the time-series response of available

indicators of investment and entrepreneurial activity to changes in marginal rates.

Although this evidence is less clear-cut, we find little evidence of an important

relationship, suggesting that the long-run productivity effects of changes in marginal rates

may also be small.

Christina D. Romer David H. Romer Department of Economics Department of Economics University of California, Berkeley University of California, Berkeley Berkeley, CA 94720-3880 Berkeley, CA 94720-3880 cromer@econ.berkeley.edu dromer@econ.berkeley.edu

I. INTRODUCTION

A central issue in tax policy concerns the incentive effects of marginal tax rates. Do high

marginal rates reduce labor supply? Do they give rise to income shielding? Do marginal rates affect

productive investment and entrepreneurial activity? The answers to these questions are crucial for

understanding how tax changes are likely to affect tax revenues and economic growth.

Many studies have looked at these possible effects of marginal rates using data from the postwar

United States. For example, there are a variety of studies using panel data to identify the elasticity of

reported income to tax rate changes. Often, these studies use particular pieces of legislation, such as the

Reagan tax cut or the 1986 tax reform, as natural experiments. One problem with using the postwar

period as the testing ground is that the degree of variation in tax rates is relatively small. For example, the

Reagan tax cuts, which are commonly acknowledged to be the most significant in the postwar era, only

reduced top marginal rates on personal income by twenty percentage points. Because the identifying

variation is relatively small, the effects of tax changes are often measured imprecisely and the studies are

inconclusive.

The interwar United States has been greatly underused as a laboratory for analyzing the effects of

marginal tax rates. Marginal rates moved frequently and dramatically in the period between the two

World Wars. The top marginal rate at the end of World War I was 77 percent; by 1929 it had been

reduced to 24 percent; by 1936 it had been raised to 79 percent, and in 1940 it was 86.9 percent.

Furthermore, tax changes in this period did not just move the tax schedule up and down uniformly. For

example, some acts mainly changed rates at very high income levels, while others were across-the-board

changes. As a result, there was both tremendous time-series and tremendous cross-section variation in

rates. This paper seeks to use this extreme variation to provide new estimates of the incentive effects of

marginal rates.

While individual panel data, such as are used in most postwar studies, do not exist for the

interwar era, the Bureau of Internal Revenue (the precursor to the IRS) provided detailed data on reported

2

income, deductions, taxes paid, and other variables for different income ranges for this period. These data

show that the very top of the income distribution paid almost all of the personal federal income tax. The

top 2/10ths of 1 percent paid roughly 95 percent of all taxes. The vast majority of Americans paid no

income tax at all. For this reason, we focus our analysis on the behavior of the very top of the income

distribution. Specifically, we use the data in the Statistics of Income to construct estimates of marginal

rates and taxable incomes of different percentile groups of the income distribution: the top 1/200th of 1

percent of the income distribution, the next 1/200th of 1 percent, and so on.

Because the interwar tax code was very progressive, changes in business cycle conditions had a

large impact on the marginal rates paid by various slices of the income distribution. For this reason, we

are careful to separate policy-induced changes in marginal rates from other changes. We do this by

examining what marginal rates in the year of a tax change would have been had the level and distribution

of income been the same as in the previous year. The resulting series on policy-induced changes in

marginal rates for various slices of the income distribution is the crucial input for this study, and may be

useful in other studies of tax policy over time. We supplement and check this statistical identification of

policy-induced changes in tax rates with a detailed analysis of the motivation and nature of each legislated

tax change using contemporaneous presidential and Congressional documents.

To analyze the responsiveness of income to changes in marginal rates, we estimate time-

series/cross-section regressions of taxable income on policy-induced changes in the after-tax share.

Macroeconomic shocks are so large in the interwar era that purely time-series analysis is potentially

problematic. The aggregate effects of changes in tax rates may have been swamped by monetary shocks,

the changes in government spending related to the end of one world war and the start of another, and

other developments. But, we can use the fact that a given tax law often changed the marginal rate of one

slice of the income distribution much more than that of another slice. This is the cross-section variation.

And, we can pool this cross-section information over time to see if there are common responses. This is

the time-series component.

The time-series/cross-section analysis shows that changes in marginal rates have a significant

3

effect on reported taxable income. However, the effects are modest. The estimated elasticity of income

with respect to the change in the log after-tax share (that is, 1 minus the marginal rate) is 0.2, lower than

what comparable postwar studies have found for high-income taxpayers. However, because of the

extreme variation in marginal rates in the interwar era, both over time and across slices of the income

distribution, the interwar estimates are more precise than most postwar estimates. The results suggest that

supply-side or income-shielding effects of marginal rate changes, while clearly present, were of limited

economic significance.

This finding is robust to a wide variety of specifications and checks. Eliminating outliers,

considering the possibility of lagged responses, and allowing for different trends in income inequality at

the top of the distribution by decade all have little effect on the estimates. Restricting the analysis to the

shorter sample 1923–1932, a period well away from both world wars and when there were large changes

in rates but no significant changes to the structure of the tax code, increases the estimated elasticity

moderately (to 0.38) but provides no evidence of a large effect.

To look for longer-run effects of marginal rate changes, we consider time-series evidence on the

response of investment and entrepreneurship. Policymakers in the 1920s, especially President Coolidge

and his Secretary of the Treasury Andrew Mellon, felt that a key effect of high marginal tax rates was to

skew investment funds away from productive activities and toward tax-free state and municipal bonds. In

their view, high tax rates distorted behavior in a way that could have reduced economic growth over a

very long horizon. Such long-run effects would be difficult to find in our time-series/cross-section

analysis, which looks at the response of taxable income in the few years immediately following the tax

change.

We test for the presence of these more long-run effects by examining the response of a number of

high-frequency indicators of productive investment activity and business formation to the aggregate

policy-induced change in the log after-tax share. To abstract from the obvious influence of other

macroeconomic developments, we look not at the simple change in these variables, but at the change

controlling for the past behavior of the series and, in some specifications, the path of overall output. This

4

allows us to at least partially separate the effect of marginal rates on investment from the usual response

of investment to business cycle conditions. Nevertheless, because these tests focus on time-series

evidence, the results are inherently more speculative than the time-series/cross-section results about the

effects of marginal rates on taxable income.

We find no evidence that the large swings in marginal rates in the interwar era had a significant

impact on investment in new machinery or commercial and industrial construction, but some suggestive

evidence that they impacted the number of business incorporations. This suggests that the modest, fairly

immediate effects of marginal rate changes on income we identify from the time-series/cross-section

analysis may be the bulk of the supply-side effects. And, of course, those modest income effects could be

the result of income shielding or tax evasion, rather than genuine effects on labor supply and effort.

As described above, a large literature examines the response of taxable income to tax rates using

postwar data. Key contributions include Lindsey (1987), Feldstein (1995), Auten and Carroll (1999),

Goolsbee (2000), Moffitt and Wilhelm (2000), Kopczuk (2005), and Giertz (2007). One particularly

thorough study along these lines is that by Gruber and Saez (2002). Whereas earlier work generally

focuses on individual tax reforms, Gruber and Saez consider all tax changes in the 1980s. They use

individual-level data and carefully control for both trends in income distribution and mean-reversion in

individual income. They find an overall elasticity of taxable income with respect to the after-tax share of

approximately 0.4, and an elasticity among high-income taxpayers of about 0.6. Although the original

studies of Lindsey and Feldstein obtained considerably larger estimates, Gruber and Saez’s estimates are

broadly in line with most of the subsequent literature.

Only a few papers consider the incentive effects of interwar tax changes. The one that is closest

to ours methodologically is that by Goolsbee (1999). He examines the behavior of taxable income in

selected years spanning three of the interwar tax changes (as well as several postwar changes). He

concludes that the episodes suggest very different responses to changes in rates. Our analysis differs from

Goolsbee’s in considering all years and all personal income tax changes in the interwar period, and in

pooling the observations to see if the overall elasticity can be estimated with more precision. We also go

5

beyond his analysis by computing marginal tax rates more accurately and by considering the responses of

various indicators of investment.

Brownlee (2000) and Smiley and Keehn (1995) also examine interwar tax changes. Brownlee

provides a careful historical analysis of some of the political economy of the tax changes and of

policymakers’ beliefs about their impact on incentives, but does not analyze their effects. Smiley and

Keehn also provide some historical background, and examine the relationship over the period from World

War I to 1929 between marginal rates and the number of taxpayers falling in various categories of taxable

income. They find a significant negative relationship between marginal rates and the number of returns.

Their regressions, however, do not control for time fixed effects and are estimated in levels. Thus, they

may be confounded by the large swings in output and the price level over their sample. And, because of

their focus on numbers of returns, it is difficult to translate their results into estimates of the elasticity of

taxable income.

Our analysis of the effects of interwar tax changes is organized as follows. Section II discusses

the history of interwar tax changes, the available tax data, and our estimates of policy-induced changes in

marginal rates by slice of the income distribution. It also discusses the possible incentive effects of

changes in marginal rates in this period. Section III presents our time-series/cross-section estimates of the

response of reported taxable income to policy-induced changes in marginal rates. Section IV reports our

findings on the time-series relationship between changes in marginal rates and the relative strength of

entrepreneurial activity and productive investment. Section V presents our conclusions.

II. INTERWAR TAX CHANGES

The federal personal income tax was established by the Revenue Act of 1913, following

ratification of the Sixteenth Amendment. Legislation changing the income tax was passed, on average,

about every other year in the interwar period. Table 1 lists all acts affecting personal income taxes in the

period 1919-1941. It shows the estimated revenue effects of each act, the impact on the top marginal rate,

and gives a brief description of the key changes contained in the act. Appendix 1 gives a narrative

6

account of each piece of tax legislation, including a discussion of the motivation for the act, the sources of

the revenue estimates, and a more detailed description of the nature of the tax changes.

A. Impact on Aggregate Demand

The revenue estimates shown in Table 1 suggest that most interwar tax changes had small effects

on revenue—often just a few tenths of a percent of GDP. One reason for this is that tax rates were low or

zero for most households. As a result, even fairly large changes in rates translated into modest effects on

revenue because the tax base was so small. This point is well illustrated by the famous tax cuts in 1924

and 1926 (the Revenue Acts of 1924 and 1926), sometimes referred to as the “Mellon tax cuts” after the

Treasury Secretary who championed them. Though these changes cut rates dramatically, their revenue

effects, even without any correction for possible incentive effects, were expected to be small.

Obviously, the revenue impact of some tax changes in the interwar period was more substantial.

The tax cut immediately after World War I (the Revenue Act of 1921) was of moderate size, and the tax

increases leading up to World War II (particularly the Revenue Act of 1941) were large. The 1941

increase was particularly large precisely because it greatly increased the fraction of households required

to pay taxes. The tax increase passed near the nadir of the Great Depression (the Revenue Act of 1932)

was surprisingly large, measuring close to 2 percent of GDP. However, over half of the revenue increase

came from a new excise tax on all manufactured goods rather than from changes to the personal and

corporate income tax.

Another crucial fact about interwar tax changes is that they were usually balanced-budget

changes. That is, tax changes and spending changes typically went in the same direction and were similar

in magnitude. As a result, the impact on the overall budget deficit or surplus was even smaller than the

estimated effects on tax revenue.

The best evidence for this link between taxes and spending comes from the narrative record

discussed in Appendix 1. If one reads what policymakers said about why they were taking various tax

actions, it is clear that the behavior of spending and the state of the budget were crucial determinants,

7

particularly in the pre-Roosevelt era. President Coolidge, for example, often spoke of reducing spending

so taxes could be reduced. He said in his Address at the Meeting of the Business Organization of the

Government in June 1924: “this fight for economy had but one purpose, that its benefits would accrue to

the whole people through reduction in taxes” (6/30/24, p. 2). Both spending and taxes did indeed drop

during the Coolidge years.

The implication of these two key facts—that the revenue effects of interwar tax changes were

typically small, and that tax changes were usually accompanied by spending changes in the same

direction—is that interwar tax changes were unlikely to have had much effect on aggregate demand. The

budget surplus simply did not move much in response to tax changes. Thus, to the extent that tax changes

mattered, it is unlikely to have been through effects on disposable income and spending. Hence, we focus

on the incentive effects of changes in marginal rates.

B. Changes in Marginal Rates

While interwar tax changes had fairly small effects on revenues and the overall budget deficit,

they had large effects on marginal tax rates. Legislation led to dramatic variation in marginal rates both

over time and across different parts of the top of the income distribution. Before discussing those

changes, it is necessary to briefly describe how we estimate the marginal rates faced by the various groups

at the top of the income distribution in this period. The details of our calculations are presented in

Appendix 2.

Estimating Marginal Rates. Our figures for the marginal rates faced by different percentile

groups are derived from the data reported in the Statistics of Income. The key income concept in the

Statistics of Income is what the Treasury called “net income.” With a few minor differences, net income

corresponds to taxable income. The Statistics of Income divides taxpayers into various ranges of net

income, such as $90,000–$100,000, $100,000–$150, 000, and so on. For each income category, there are

data on the number of returns, income of various types, deductions, taxes due, and other variables. The

publications also provide detailed descriptions of the tax code and the rates that applied at different levels

8

of income.

If households’ tax liabilities had been completely determined by their net incomes, it would be

straightforward to find the marginal rate of each percentile group. The only complication would be that

because the Statistics of Income report data for income ranges rather than each income level, we would

need to make some assumption about the distribution of income within each range. Armed with such an

assumption, we could find each percentile group’s overall net income and its average marginal rate.

The interwar tax system was not quite this simple. The most important complication involves the

treatment of capital gains, which varied greatly over the period. We therefore exclude capital gains from

our definition of income, and focus on the relationship between taxable income exclusive of capital gains

and losses and marginal rates on non-capital-gains income. Excluding capital gains is common in studies

of tax responsiveness, both because they often do not reflect current income but the timing of realizations

and because they are often taxed differently than other types of income (for example, Gruber and Saez,

2002). As a further check that our results are not driven by the behavior of capital gains, at times we

focus on the shorter sample period 1923–1932, over which the treatment of capital gains was unchanged.

Because the interwar tax system was highly progressive, marginal rates changed not only because

of legislated changes, but also because of economic growth, inflation and deflation, behavioral responses

to legislated tax changes, and other forces. To determine the effects of changes in marginal rates, it is

therefore important to separate the changes resulting from legislation from those arising endogenously

from economic developments. Specifically, for each legislated change in taxes, we calculate the policy-

induced change in the marginal rate of a given percentile group as the change that would have occurred if

its income had not changed. When the tax code was changed retroactively (as sometimes occurred in this

period), we focus on the rates that were in effect at the time individuals were earning income, not on the

rates that were applied ex post. Appendix 2 explains the details of our calculations of taxable incomes

excluding capital gains, marginal rates, and policy-induced changes in marginal rates.

Interwar Changes in Marginal Rates. To give a sense of the time-series variation in marginal

rates over the interwar period, Figure 1 shows the top marginal rate in each year. The figure reveals large,

9

serially correlated changes. The top rate was extremely high (close to 80 percent) coming out of World

War I. It was reduced by more than two-thirds in a series of tax actions in the 1920s, most notably the

Revenue Acts of 1921, 1924, and 1926. It was then raised dramatically by the Hoover-era Revenue Act

of 1932. The Roosevelt administration increased it further in the Revenue Act of 1935, which was aimed

primarily at the very rich, and again through a series of broad-based tax increases on the eve of World

War II.

While the top marginal rate is a potentially interesting number, the marginal rate faced at different

points in the income distribution is clearly more important. Also, for calculating elasticities, it is helpful

to look not at the simple change in the marginal rate, but at the change in the log after-tax share (one

minus the marginal rate). Figure 2 shows our estimates of the policy-induced changes in the log after-tax

share for the top 10 two-hundredths of 1 percent of the income distribution. The results are in the change

in logs (approximately 0.01 times the percent change in the after-tax share). A positive value corresponds

to a tax cut; a negative value to a tax increase.

These numbers show that changes in marginal rates and thus after-tax shares, while correlated

across income groups, were highly variable. Some laws, such as the Revenue Act of 1924, lowered rates

and raised after-tax shares on all slices of the income distribution fairly uniformly. Other laws, such as

the Revenue Act of 1935, raised rates and lowered after-tax shares much more for top income groups than

for lower income groups. And laws such as the Revenue Act of 1941 raised rates and lowered after-tax

shares dramatically on slices of the income distribution below the very top, but made almost no changes

to the top marginal rate. This variation in changes in marginal rates across income groups will be central

to our identification strategy for estimating the incentive effects of tax rate changes.

The fact that there is so much variation in marginal rates in the interwar era has two important

implications. First, it means that this period has the potential to provide important evidence about the

effects of changes in marginal rates on the behavior of the wealthy: the changes in incentives due to the

tax code were large, frequent, and variable. Second, it means that the effects of the inherent imprecision

of trying to construct data on average marginal rates of different percentile groups from the data in the

10

Statistics of Income are likely to be modest: the signal provided by changes in statutory marginal rates

that frequently exceeded 10 percentage points in a year are likely to swamp the noise introduced by the

imperfections in our data construction.

Changes in Other Tax Features. The description of each tax law given in Appendix 1 gives a

sense of some of the other changes in the tax system in the interwar era. For example, Social Security

taxes were imposed by the Social Security Act passed in 1935. The initial tax rate was 1 percent on the

first $3000 of earnings received by workers and on the first $3000 of wages paid by employers. The taxes

took effect in 1937. The $3000 cut-off level was below the income of the part of the income distribution

we analyze in this paper. As a result, Social Security taxes did not affect the marginal rates of the

taxpayers in our sample.

The descriptions in Appendix 1 also show that the interwar period was a time of significant

changes in the corporate income tax. The most extreme changes involved various excess profits taxes,

which were removed after World War I, reintroduced in the National Industrial Recovery Act of 1933,

and greatly increased in the run-up to World War II. There was also a substantial undistributed profits tax

introduced in 1936 and gradually eliminated shortly thereafter. The regular corporate income tax was

changed frequently in the interwar era, but usually within a very narrow range. Between 1922 and 1933,

the corporate rate varied between 11 and 13¾ percent.

One type of corporation came in for quite extreme tax changes in the mid-1930s. A personal

holding company was a corporation set up to hold the assets of an individual or a family; the individual or

family then held stock in the corporation. Income was largely retained by the corporation, which paid the

much lower corporate tax rate, rather than distributed to the shareholders, who would have paid the much

higher personal income tax rate. The Revenue Acts of 1934, 1936, and 1937 greatly increased the tax

rates on such corporations.

Changes in the corporate income tax, for the most part, should not affect the sensitivity of

personal income to personal income tax rates. The one exception might be the changing treatment of

personal holding companies, which reduced one avenue of income shielding. The fact that changes in the

11

corporate tax were minimal from 1923 to 1932 provides another reason for examining the robustness of

the results to focusing on this shorter sample period.

C. Possible Incentive Effects

Before we turn to the statistical analysis, it is useful to discuss the possible incentive effects of

interwar tax changes. What do the features of the interwar economy and the interwar tax system suggest

are the likely effects of tax changes on behavior? In thinking about these likely effects, we consider what

policymakers at the time though were the most important linkages.

Labor Supply. A key feature of the interwar tax system was that income taxes were paid almost

entirely by the rich. Figure 3 shows fraction of total federal personal income taxes paid by households at

the top of the income distribution. Specifically, it shows the percent paid by the top 10 two-hundredths of

the top 1 percent, where the amounts are cumulated as we move down the income distribution. It shows

that between 25 and 40 percent of federal personal income taxes were paid by households in the top

1/200th of 1 percent of the income distribution. Roughly 60 percent were paid by households in the top

1/20th of 1 percent of the income distribution. And, though not shown in Figure 3, approximately 95

percent were paid by the top 2/10ths of 1 percent of the income distribution.

Most of the variation in marginal rates occurred at annual incomes above $20,000. In addition,

the exclusion of some items from taxation at low levels of income makes it harder to estimate marginal

rates at lower income levels. These considerations lead us to focus on the top 0.05 percent of the income

distribution. Specifically, the number of returns we consider in a year is 0.05 percent of the number of tax

units in the United States in that year.1

The fact that income taxes were paid almost entirely by the very rich is important for thinking

about the likely incentive effects. The income of the rich in this period did not come primarily from

Since there were about 50 million tax units in the United States in

this period, this rule implies that we consider about 25,000 returns per year. The net income cutoff for

being in this group ranged from $25,400 (in 1933) to $75,100 (in 1928).

1 Data on the number of tax units are from Piketty and Saez (2001, Table A0).

12

wages and salaries, but rather from dividends, interest, and business or partnership profits. The Statistics

of Income (Table 7) for each year breaks down total income for nominal income groups into a number of

categories. From this data, along with our calculations of the income cutoffs for various slices of the top

of the income distribution, it is possible to calculate the fraction of total income accounted for by income

from different sources in the interwar period.

For the top 1/200th of 1 percent of the income distribution, wages and salaries accounted for only

about 10 percent of total income less capital gains, while dividends and interest income accounted for

about 70 percent. Business and partnership profits accounted for another 10 percent of income. For the

top 10 two-hundredths of 1 percent of the income distribution combined (the top 1/20th of 1 percent),

wages and salaries accounted for 15 to 20 percent of total income less capital gains, while dividends and

interest accounted for roughly 60 percent. Business and partnership income accounted for another 10 to

15 percent. For both these slices of the income distribution, the importance of wage and salary income

rose slightly over time.

That wage and salary income was a fairly small fraction of total income for the groups that paid

most of the federal income tax in the interwar era suggests that the labor supply effects of marginal rates

were of limited importance. If marginal tax rates mattered in this period, it was probably not by

encouraging people to enter or exit the labor force or to work more or fewer hours. Simply too few of the

relevant taxpayers appear to have been conventional wage and salary workers.

This argument that the labor supply effects of marginal rates were probably of limited importance

is consistent with their noticeable absence from policymakers’ discussions of incentive effects. Appendix

1 documents that policymakers discussed the incentive effects of marginal rates extensively in the

interwar era. But, we have found little discussion of rates affecting decisions to work and labor effort in

the narrative record. The closest we have found to this argument is a discussion in the 1921 Treasury

Annual Report that talked of high marginal rates “drying up … the activities of individuals in trade

operations” (p. 16). This near absence of discussion is in stark contrast to the postwar narrative record,

where such labor supply effects were discussed frequently and thought to be central (see Romer and

13

Romer, 2009).

Income Shielding and Investment. The effects of marginal rates that were more likely to be

important in the interwar era are those related to income shielding, both legal and illegal. When most

income is derived from investments, changes in marginal rates create incentives to change the amount of

income received and the form in which wealth is held. In this way, changes in marginal rates can have a

large impact on reported income and government revenues. Whether they have effects on efficiency and

long-run growth depends on whether the shielding activities require substantial wasted effort or lead to

important distortions.

Policymakers in the interwar era believed that income shielding was prevalent and quite

responsive to marginal tax rates. One of the clearest statements of this belief comes from the 1921

Treasury Annual Report. It stated:

Experience teaches us that means of avoiding taxes which are regarded as excessive or unreasonable will always be found, and it would be useless to attempt to catalogue them, for new methods will constantly be developed so long as the tax rates continue so high that persons having money for investment find it unprofitable to continue their investments in productive industry (p. 15).

The method of income shielding emphasized most by President Coolidge and his Treasury

Secretary Andrew Mellon was placing wealth in tax-exempt securities such as municipal bonds. In a

letter to the chairman of the House Ways and Means Committee in 1923, Mellon stated:

Taxpayers subject to the higher rates can not afford, for example, to invest in American railroads or industries or embark upon new enterprises in the face of taxes that will tax 50 per cent or more of any return that may be realized. These taxpayers are withdrawing their capital from productive business and investing it instead in tax-exempt securities and adopting other lawful methods of avoiding the realization of taxable income (1923 Treasury Annual Report, p. 8).

The Coolidge administration believed this diversion of investment funds was significant enough that it

was impeding employment and long-run growth. It called not only for a drastic decline in marginal rates

to decrease this incentive, but also repeatedly asked for a constitutional amendment taking away the right

of states and municipalities to issue such securities.

Like their counterparts in the 1920s, policymakers in the 1930s were also very concerned about

14

income shielding. A special subcommittee of the Ways and Means Committee established in June 1933

was charged with “studying tax avoidance and the means of preventing such avoidance” (“Revenue Bill

of 1934,” 73rd Congress, 2d Session, House of Representatives Report No. 704, 2/12/34, p. 1). Among the

methods the committee found most worrisome was the creation of personal holding companies described

above. Roosevelt in 1936 suggested that the problem was broader. He argued that many corporations

were retaining earnings rather than paying dividends as a way to help shareholders avoid paying the

personal income surtax. Unlike Coolidge, who pushed for lower marginal rates as a way to lessen efforts

at income shielding, Roosevelt and the Democratic Congress responded by raising rates on personal

holding companies and placing a special tax on undistributed corporate profits. Nevertheless, their

concern about tax avoidance suggests they believed that such incentive effects were substantial in the

interwar era.

While most of policymakers’ discussions focused on legal income shielding and tax avoidance, it

is important to note that tax enforcement was relatively lax in the interwar era. There was no tax

withholding, so self-reporting was key. The only enforcement tool was auditing, which was done

relatively little (though occasionally with great fanfare, as in the cause of the audit of Andrew Mellon in

1934). As a result, there is reason to expect more illegal hiding of income in the interwar era than in the

postwar era. This would be a factor that could tend to make reported income particularly sensitive to

changes in marginal rates.

Finally, because business formation and entrepreneurial activity have the potential to generate

large increases in income over time, changes in marginal rates could affect the behavior of people not

currently paying the income tax. A low-earning entrepreneur may consider what will happen if an

investment is successful enough to lift him or her into the tax-paying brackets. The entrepreneur may not

see much potential reward from starting a new business when marginal tax rates are high, but more when

marginal rates are lower. Thus, changes in rates could affect behavior not reflected in the income

statistics of current taxpayers.

15

III. TIME-SERIES/CROSS-SECTION ESTIMATES OF THE IMPACT OF CHANGES IN MARGINAL RATES ON REPORTED TAXABLE INCOME

The first way that we investigate the impact of changes in marginal rates is to use the detailed

time-series/cross-section data from the Statistics of Income. As described in the previous section, the

Statistics of Income provide data that can be used to measure taxable income by percentile group and the

corresponding policy-induced changes in marginal rates. We have compiled these data for the 10 top

two-hundredths of 1 percent of the income distribution for the period 1919 to 1941.2

Because interwar

tax changes typically affected the percentile groups differently, we can test to see if the responses of

taxable income differ systematically. Because we have many years of data, we can use the numerous tax

actions to derive more precise estimates of the impact of changes in marginal rates.

A. Specification

Our baseline specification is:

(1) ∆ ln𝑌𝑖𝑡 = 𝛼𝑖 + 𝛽𝑡 + ∑ 𝛾𝑗∆ ln(1 − 𝜏)𝑖,𝑡−𝑗𝑃𝐼 + 𝜀𝑖𝑡 ,𝐽𝑗=0

where ∆ ln𝑌𝑖𝑡 is the change in the log of real “net income” of group i in year t (exclusive of capital gains),

and ∆ ln(1 − 𝜏)𝑖,𝑡−𝑗𝑃𝐼 is the policy-induced change in the log of the marginal after-tax share of group i in

year t–j. We regress the percent change in reported taxable income of a group on the change in the log of

the after-tax share of income of that group. The estimated coefficient is therefore the elasticity of taxable

income with respect to the share of income kept by the taxpayer. Economic theory implies a positive

coefficient: a decline in the marginal tax rate which raises the after-tax share raises taxable income.

We include a full set of group and time dummy variables. The group dummies (the 𝛼𝑖′s) capture

differences in trend growth of the income of the various percentile groups. The time dummies (the 𝛽𝑡′s)

capture year effects. The interwar period was a time of extreme movements in output, stock price swings,

and wartime mobilization and demobilization. Such time-specific macroeconomic shocks could obscure

the incentive effects of changes in marginal rates if time dummies were not included. We also consider

2 We calculate marginal rates by percentile group beginning in 1918. The policy-induced changes are therefore available starting in 1919.

16

specifications that exclude one or both sets of dummies and that replace the time dummies with GDP

growth.

In the simplest regressions, we only consider the contemporaneous relationship between income

and the after-tax share. However, it is surely possible that it takes additional time for taxpayers to adjust

their behavior to tax changes. We therefore consider specifications including 1 and 2 lags of the tax

policy variable.

Our baseline specification estimates the relationship between income and policy-induced tax

changes using OLS. An obvious alterative to this “reduced-form” specification is to use an instrumental

variables approach. In this specification, we regress the change in income (by year and percentile group)

on the change in the actual after-tax share, instrumenting with the policy-induced portion.

In the baseline specification, we estimate the relationship over the full sample period 1919 to

1941. In the robustness section, we consider a range of alternative samples.

One important alternative sample period is 1923–1932, when there were no major changes in

capital gains or corporate income taxes. Our underlying data are grouped by taxpayers’ net incomes

including capital gains. As a result, our estimate of, for example, the non-capital-gains income of the top

1/20th of 1 percent of the population ranked by net income will understate the non-capital-gains income of

the top 1/20th of 1 percent of the population ranked by non-capital-gains income. Piketty and Saez (2001)

find that this effect is non-trivial for the top 1/100th of the income distribution. However, one would

expect—and Piketty and Saez’s results suggest—that the effect would not change substantially in years

when the tax treatment of capital gains is stable. Thus, although this effect is unlikely to have a large

impact on our estimates even over the full sample, it is especially unlikely to cause bias over the 1923–

1932 sample.

Likewise, some of the response of taxable income to marginal rates likely reflects shifting

between personal and corporate forms of income, which acts to mitigate the welfare costs of the response

(Gordon and Slemrod, 2000). Focusing on a period when corporate rates barely changed, and where there

were no cases of major changes in corporate taxes occurring simultaneously with major changes in

17

personal income taxes, helps to isolate the effects of the personal tax.

B. Baseline Results

Table 2 shows the results of the baseline specification and some key econometric permutations.

The coefficient(s) of interest are those on the contemporaneous and lagged values of the after-tax share.

When we include lags of the tax variable, we focus on the sum of the coefficients on the

contemporaneous and lagged values, along with the associated standard error.

A striking feature of the results is their consistency. The estimated impact of a rise in the after-

tax share is consistently positive, small, and precisely estimated. In the simplest specification which

includes only the contemporaneous value (line 1), the coefficient estimate is 0.19 with a t-statistic over 6.

This estimated elasticity implies that a rise the after-tax share of 1 percent (say from 50.0 to 50.5 percent)

raises reported taxable income by just under 2/10ths of a percent.

The results are virtually identical when one uses IV in place of OLS. In all the specifications we

consider, the actual after-tax share moves almost one-for-one with the policy-induced change and the

effect is overwhelmingly significant. In the first stage of the IV variant of the baseline specification, for

example, the coefficient on the policy-induced change in the log after-tax share is 0.998 with a t-statistic

of 28. Using IV (line 2) barely changes the estimated elasticity and increases the standard error only

slightly. Because the interpretation of the IV specification is somewhat complicated, in subsequent

robustness checks we focus on the reduced-form OLS specification.

Including lags of the tax variable (lines 3 and 4) increases the sum of the coefficients slightly—

from 0.19 with no lags to 0.27 with 2 lags. The standard error on the sum of the coefficients rises

slightly, but the sum is still highly statistically significant (t = 4.74). The individual coefficient estimates

on the first and second lags, however, are not significantly different from zero.

As shown in line 5, leaving out the group dummy variables has almost no impact on the

estimates. On the other hand, leaving out the time dummies (line 6) matters substantially. As discussed

above, we would expect the results might be affected by other aggregate shocks when the year dummies

18

are excluded, and they are. Replacing the year dummies with the growth rate of GDP (line 7) yields

results similar to those when the year dummies are omitted. This suggests that the time dummies are

capturing more than just the systematic response of taxable income to GDP growth. For example, they

may capture changes in tax compliance associated with increased patriotism around the run-up to World

War II or different responsiveness of taxable income to output in the 1920s than in the 1930s.

C. Robustness

The previous section showed the key results and investigated their robustness to differences in the

econometric specification. In this section, we investigate their robustness to different specifications of the

tax variable and the group dummy variables. We also consider a wide range of alternative sample

periods.

Specification of the Tax Variable. As described in Appendix 1, some interwar tax changes were

retroactive to the previous year. Because people cannot change their behavior retroactively, our baseline

tax policy variable reflects the change in the year it was passed, not the year it took effect. However,

because tax returns were not filed until the subsequent year (and after the retroactive tax changes had

been passed), fraudulent income reporting could respond to the retroactive changes. Individuals could

also have responded to the retroactive changes if the changes were widely anticipated. For these reasons,

we consider an alternative specification of the tax variable that dates changes when they took effect, even

if they were passed the next year. We try specifications that replace our baseline tax variable with this

alternative and that include both the baseline variable and the difference between the alternative series

and the baseline.

The results are shown in the lines 2 and 3 of Table 3. For simplicity, in evaluating robustness we

only report the baseline specification, which includes no lags and is estimated using OLS. The results of

this specification for our baseline tax variable are repeated in line 1 of Table 3.

Replacing the baseline tax variable with the perfect-foresight alternative (line 2) has only a

moderate impact on the results. The estimated response of taxable income to tax changes falls from 0.19

19

to 0.12 and is less overwhelmingly significant. The most likely explanation for this pattern is that the

perfect-foresight variable is a less accurate measure of the incentives that taxpayers were responding to.

As noted above, taxpayers could respond to retroactive changes by altering the amount of their

income that they reported. To test for this possibility, we add the difference between the two tax variables

to the baseline regression. In this specification (line 3), the coefficient on the baseline tax variable is

virtually identical to before and the coefficient on the difference is negative, near zero, and far from

statistically significant. This suggests that fraudulent income reporting and anticipation effects in

response to retroactive tax changes may have been relatively unimportant in the interwar period.

Decade-Varying Group Dummies. Another robustness concern involves secular changes in

income inequality and taxes. The 1920s were a period of rising income inequality and falling marginal

rates, while the 1930s were a period of declining inequality and rising marginal rates. It is possible that

with only group dummies for the whole period, the regression mistakes this correlation for a behavioral

response to marginal rates. To check for this possibility, we specify two sets of group dummies—one for

the period through 1929 and one for the period beginning in 1930.

The results of this exercise are shown in line 4. Letting the average growth rate of taxable

income by percentile groups be different in the 1920s and 1930s has no impact on the results. The

estimated elasticity of income to the after-tax share is still 0.19 with a t-statistic of 6.

Another way that we test for the same effect is to interact the group dummies with real GDP

growth. It is possible that the top income groups do particularly well in good times and particularly badly

in bad times. As shown in line 5, letting the group dummies vary with real GDP has little impact on the

results.

A related robustness check just excludes the very top percentile group (the top 1/200th of 1

percent of the income distribution) from the estimation. This exercise tests whether the results are being

driven only by this one group whose relative income moved a lot in the interwar period. Line 6 of Table

3 shows that the results are very similar to the baseline when this group is eliminated.

Sample Period. We consider the robustness of the results to a wide range of sample periods. The

20

baseline sample is the full period from 1919 to 1941. This period already excludes the most extreme

wartime changes. But to avoid the impacts of demobilization in 1919 and mobilization in 1940 and 1941,

we consider the non-war sample 1920 to 1939. In addition, as discussed above, we consider the period

from 1923 to 1932, which was relatively free of changes in capital gains and corporate taxes. Finally, the

Great Depression was such an enormous shock that we consider both a pre-Depression (1919 to 1929)

sample and a Depression (1930 to 1941) sample.

Lines 7 to 10 of Table 3 report the results for the various sample periods. In general, the

alternative sample periods lead to estimates that are somewhat higher than for the full sample, but still

modest. In each of the non-war, stable corporate and capital gains tax, pre-Depression, and Depression

samples, the coefficient on the after-tax share is small, positive, and precisely estimated. The largest

estimate is for the period where other aspects of the tax system were relatively stable, where the estimated

elasticity is 0.38.3

Figure 4 shows the scatter plot of the contemporaneous percentage change in income and the

change in the log after-tax share, after partialing out the impact of the group and time dummies. That is,

both the x and y variables have been regressed on the two sets of dummies, and what is plotted are the

residuals against one another. This graph shows that there are some extreme observations—particularly

for the very top 1/200th of 1 percent of the income distribution. But, the scatter also illustrates why the

results are quite robust to excluding the very top group and to different sample periods. The extreme

observations are all cases of large tax changes and relatively small income changes, so excluding one or

two with a different sample is unlikely to change the results. Also, the extreme observations are very

much in line with the mass of the observations. As a result, excluding the extremes (for example, by

excluding the very top 1/200th of 1 percent) has little impact on the estimates.

3 For this sample, in contrast to the full sample, introducing lags reduces the overall effect somewhat. With either 1 or 2 lags, the sum of the coefficients on the tax variables is 0.28, and still highly statistically significant. As before, the coefficients on the lags are not significant.

21

D. Implications

The analysis in the previous two subsections shows that the results of the time-series/cross-

section analysis are exceedingly robust to econometric specification, tax and dummy variable

specification, and sample period. Changes in marginal tax rates and hence in the after-tax share have a

precisely estimated but small impact on reported taxable income.

One way to quantify just how small the estimated elasticity of income to after-tax share is is to

calculate what it implies about the revenue-maximizing tax rate. If the tax rate is constant and the

elasticity of taxable income with respect to the after-tax share is constant across income levels, tax

revenue is maximized when the tax rate equals 1 divided by 1 plus the elasticity of taxable income. Thus,

an elasticity of 0.19 implies that the revenue-maximizing tax rate is 1/(1+0.19), or 84 percent. This is

substantially above marginal rates in the postwar period and the interwar period after the Revenue Act of

1922. Even an elasticity of 0.38, which is the highest we find in our robustness checks, implies that the

revenue-maximizing rate is 73 percent.

The estimated elasticity of taxable income to the after-tax share obviously includes a variety of

incentive effects. It captures incentives to adjust labor effort and hours, to take actions to legitimately

shield income from taxes, and to risk prosecution by falsely reporting income. The time-series/cross-

section results suggest that the sum of all of these incentive effects was very small in the interwar period.

As discussed above, the interwar period is an excellent period for investigating these incentive

effects because there were large changes in marginal rates that affected various slices of the income

distribution very differently. However, the interwar period is obviously far removed from the present.

Reporting requirements, tax enforcement, and financial sophistication were all quite different in the 1920s

and 1930s than they are today. As a result, it is not clear how closely the interwar estimates are likely to

reflect current behavior.

In this regard, the fact that our estimated elasticity is similar, though smaller and more precisely

estimated, than Gruber and Saez’s estimate from the 1980s is interesting. It confirms that the relatively

small estimates from the detailed postwar sample are plausible. It also suggests that differences between

22

the interwar and postwar eras may have had offsetting impacts. Weaker tax enforcement in the 1920s and

1930s may have allowed the fraud response to be stronger in those decades. But less financial

sophistication and a simpler tax system may have caused the income-shielding response to be weaker.

Likewise, the fact that much more interwar income at the top of the income distribution was unearned

suggests that the labor-supply response was less important for this sample than in later periods.

IV. TIME-SERIES EVIDENCE ON THE EFFECT OF MARGINAL RATES ON INVESTMENT

The previous section found that changes in marginal tax rates in the interwar period had a small

and precisely estimated impact on the taxable incomes of taxpayers at the top of the income distribution

in the years immediately following the changes. This suggests that the very large movements in marginal

rates did not have a large short-run impact on the behavior of the taxpayers directly affected.

Many concerns about the effects of marginal tax rates, however, involve more than taxpayers’

short-run responses through such channels as income-shielding, tax evasion, and labor supply. Instead,

they focus on the possibility that high marginal rates discourage investment, innovation, and

entrepreneurship, and so slow the economy’s long-run growth.

Our findings about the response of taxable income do not rule out the possibility of such effects.

For example, suppose a small component of overall investment, such as machinery investment, is

particularly important for long-run growth (as argued by DeLong and Summers, 1991). In this case, a

change in marginal rates that changed investment behavior might have little impact on the income of the

wealthy in the short-run, but a substantial effect over time for the entire economy. Similarly, as discussed

in Section II, changes in marginal rates could affect the entrepreneurial behavior of individuals not

currently paying taxes. If a cut in marginal rates increased business formations by the less wealthy, this

would not be apparent in the short-run income response at the upper end of the income distribution, but

could have a large impact on growth.

This section therefore investigates the responses of available interwar series on investment and

entrepreneurial activity to policy-induced changes in marginal rates at the top of the income distribution.

23

We examine whether variables such as machinery investment and business incorporations responded

systematically to changes in marginal rates.

This exercise is inherently more tentative than the examination of the response of taxable income

in Section III. Because we do not have data on investment and entrepreneurship by income level, we can

only exploit the time-series variation in marginal rates. Thus we lose a considerable part of our

identifying variation. Even more important, the fact that we are forced to rely on the time-series variation

means that the effects we are interested in may be confounded by the enormous shocks affecting the

economy over this period. In addition, because the investments and entrepreneurial decisions that drive

growth are hard to isolate, we do not have high-quality measures of the activities that we are interested in.

Thus, our results need to be interpreted with caution.

A. Series Analyzed

Changes in marginal tax rates were frequent in the interwar period. It is therefore useful to

consider high-frequency indicators of productive investment and entrepreneurial activity. While the

number and quality of such indicators is obviously much more limited for the interwar period than for

today, there are some potentially useful measures.

Investment and Entrepreneurial Activity. We focus on three indicators of business investment

and entrepreneurial activity. The first is the Federal Reserve index of industrial production for

machinery.4

The machinery series for this period is derived almost entirely from man-hour data collected by

the Bureau of Labor Statistics. Since productivity is generally procyclical, the use of man-hours means

This series is available starting in 1923:1. The main types of machinery included are

electrical machinery, foundry and machine-shop products, engines, and agricultural implements. This

series measures the production of these types of investment goods rather than actual investment, but the

two are likely to be highly correlated.

4 This series was first described and presented in an article in the Federal Reserve Bulletin (“A New Federal Reserve Index of Industrial Production,” Vol. 26, August 1940, pp. 753-769). Data through 1941:12 were collected from later issues of the Federal Reserve Bulletin. We use the seasonally adjusted version of the series.

24

that cyclical moments are likely to be understated. But there is no reason to think that the series is

inconsistent over the interwar period.

The second series we consider is the value of construction contracts for commercial and industrial

buildings. These data are from Lipsey and Preston (1966, pp. 88-90), and are based on information

compiled on a relatively consistent basis by the F. W. Dodge Company. They are available throughout

the interwar period. The series covers mainly the eastern half of the country and shows commitments to

start work within about 60 days. Lipsey and Preston report that various tests suggest that the data cover

most non-residential construction and are reasonably accurate. We deflate the series by the Consumer

Price Index to convert it from nominal to real.5

The third series we consider is an index of business incorporations from Evans (1948, Table 38,

pp. 80-81). This series is based on detailed data culled from various states. The particular group of states

included in the index varies over our sample, but data for New York and Delaware are included for most

of the period. Evans is careful to construct substantial periods of overlap between series using different

sets of states and to splice the series together appropriately. Nevertheless, the changing sample could

affect trend behavior if states had substantially different trends, for example because of state-specific

legislation. The resulting index is a measure of one type of business formation.

The construction contracts series is a leading indicator of

investment in commercial and industrial buildings. As such, it is a high-frequency indicator of this type

of fixed investment.

6

Interest-Rate Ratio. As discussed in Section II, interwar policymakers were convinced that high

marginal tax rates skewed investment greatly toward local public investment and away from private

investment. To test this idea, we examine the ratio of the interest rate on municipal bonds to the rate on

5 The number of states covered by the Dodge series rises from 27 before 1923 to 36 in 1923 and 1924 (9 southern states were added) to 37 starting in 1925 (Texas was added). We take the measure with the widest coverage and join the series using a ratio splice in the latest year of overlap. We use the seasonally adjusted version of the series. Because the movements in the nominal series are so large relative to movements in prices, the specifics of how we deflate it are unlikely to be important. The specific series we use is the Consumer Price Index for all urban consumers (series CUUR000SA0,CUUS0000SA0, downloaded 1/19/2011). Only the not seasonally adjusted series is available for this period. However, seasonal movements in the series are small. 6 Evans presents two indexes, one covering the period through 1925 and the other beginning in 1924. Following the procedure he uses in other cases, we splice the two series together using their annual averages in 1924. The series is seasonally adjusted.

25

AAA corporate bonds. This series is not a measure of investment, but an indicator of the extent to which

the tax system distorted investment incentives. These data are available for our full sample period.7

Aggregate Tax Variables. Since we are focusing only on the time-series dimension of the data,

we need a measure of the overall policy-induced change in the log after-tax share for the upper end of the

income distribution. We calculate this series by aggregating the changes for the different percentile

groups in our earlier analysis, weighting them by income. That is, consistent with the time-series/cross-

section analysis, we calculate the marginal after-tax share of the top one-twentieth of 1 percent of the

income distribution for their year t – 1 income under both the year t – 1 tax code and the year t tax code.

The difference between the two measures is the policy-induced change in year t.

We use a slightly different tax measure when we are considering the ratio of the interest rates on

tax-free and taxable bonds. If the bonds are otherwise comparable, one would expect the ratio of the two

interest rates to equal 1 minus the marginal tax rate of the marginal investor. In analyzing this series, we

therefore focus on the policy-induced change in 1 minus the average marginal rate of wealthy taxpayers

(rather than in the log of 1 minus their marginal rate).

Because our investment indicators are monthly, we need the tax series at a monthly frequency as

well. For tax changes that were passed before they took effect, we date the changes as occurring when

they went into effect (which is always in January of the year). When a change was retroactive, we date

the changes as occurring when the legislation was passed. That is, we ignore the retroactive component

and compare the new rate with the rate under the previous legislation. However, for completeness, we

test the robustness of our results to using the perfect-foresight variant of our series. This series dates

retroactive changes as occurring when they took effect rather than at the later dates when they were

passed, even though individuals would not have had full information about them at the time.

7 The data on both interest rates are from the U.S. Board of Governors of the Federal Reserve System (1943, Table 128, pp. 468-471). Interest rate data are generally thought to be accurate for the interwar period. The markets were thick and information on rates was widely published.

26

B. The Behavior of the Investment Series

Given the exceptional nature of our sample period, it is helpful to start by simply looking at the

data. Figure 5 presents graphs of each investment series and the aggregate tax variable. Our tax series

begins in 1919; all of the investment series are available by then except for machinery production, which

begins in 1923.

One fact apparent from the figure is just how extreme the tax changes were in the interwar period.

The largest changes moved the after-tax share by 20 percent or more. The positive values in the 1920s

correspond to the large cuts in marginal rates in this decade. The negative values in the 1930s and early

1940s correspond to the large increases in marginal rates, particularly in 1932 and in the mobilization

leading up to World War II.

A second fact apparent from the figure is that there were huge swings in the investment series that

almost certainly reflect the effects of the large macroeconomic shocks of this period and of additional

forces other than marginal tax rates. Machinery production, for example, skyrocketed in the late-1920s

boom, fell by over a factor of three in the Depression, and returned almost to its pre-Depression peak in

1937 before going through an additional gyration in the years leading up to World War II. The

construction and incorporations series also exhibit tremendous movements over extended periods, and

those movements differ substantially from those of machinery production.

Investment and Entrepreneurial Activity. Now consider the behavior of each series and its

relation to changes in taxes in more detail. Panel (a) shows the behavior of the production of machinery

beginning in 1923. It suggests no clear link between tax rates and this component of investment.

Machinery production grew strongly after the 1924 tax cut, but was largely flat after the larger 1926 cut.

It was likewise flat immediately after the very large 1932 tax increase, before surging in the early stages

of the recovery from the Depression. Not surprisingly, it also surged during the mobilization for World

War II despite two substantial tax increases.

Panel (b) shows the behavior of commercial and industrial construction contracts beginning in

1919. This graph also suggests that fluctuations in construction were driven largely by factors other than

27

taxes. Construction more than quadrupled after the large tax increase in 1919 to pay for World War I.

After plummeting in the recession of 1920–1921, it rose over the 1920s, but choppily and in a way that

was not clearly related to the three tax cuts in this period. Industrial construction changed little following

the 1932 tax increase, and rose temporarily after the 1935 increase. And like machinery production, it

rose strongly after the tax increases leading up to World War II.

Panel (c) shows the behavior of incorporations from 1919. This graph shows potentially more of

a relationship between investment and legislated changes in taxes. Incorporations largely rose over the

1920s, when the after-tax share was rising, and fell over the 1930s, when it was falling. Moreover, there

are noticeable surges in incorporations after each of the tax cuts in the 1920s, and noticeable declines after

the tax increases in 1940 and 1941. However, there is a dramatic surge after the 1919 tax increase that

goes strongly against this correlation, and little change following the tax increases of 1932 and 1935.

Nevertheless, incorporations appear to be the one investment series that may suggest an impact of

marginal rates.

Interest-Rate Ratio. Finally, Figure 6 shows the ratio of the interest rate on municipal bonds to

the rate on AAA corporate bonds. In keeping with what one would expect about the determinants of this

ratio, we look at the after-tax share rather than its log, and at the level of the series rather than only at

policy-induced changes. Thus the tax series shown is simply 1 minus the average marginal rate of the top

one-twentieth of 1 percent of the income distribution. In addition, for simplicity the figure uses the actual

tax rates paid, without adjustment for the fact that in some cases the rates were set retroactively. (Using

the rates that people would have thought were in effect at the time would yield a very similar picture.)

Figure 6 shows that to the degree the two series are related, the relationship is extremely muted.

In the 1920s, the after-tax share was rising sharply while the interest-rate ratio was at most creeping

upward. In the 1930s, as the after-tax share was falling, the interest-rate ratio was falling as well, but less

than one-for-one. And following the 1932 tax increase, the ratio surged rather than fell. The simple

picture therefore suggests that it will be difficult to find Mellon’s hypothesized relationship.

28

C. Statistical Tests

We now turn to more formal examination of the relationship between legislated changes in

marginal rates and our investment indicators. We begin by describing our specifications and sample

periods. We then present the baseline findings, and finally discuss robustness.

Specifications and Samples. Our baseline specification is a two-variable vector autoregression

(VAR) with the investment measure of interest and the policy-induced change in the log after-tax share of

the top 1/20th of 1 percent of the income distribution. The investment measures are entered in log levels,

and the VAR includes 24 lags. The tax variable is ordered first, so that investment can potentially

respond to taxes within the month but taxes cannot respond to investment.

This specification asks how investment behaves relative to its usual behavior following a

legislated change in taxes that is not predictable based on the past behavior of investment and of changes

in taxes. Throughout, we find that the tax changes are essentially unpredictable; for example, the adjusted

R2 of the tax equation in the VAR is almost always negative. Thus the VAR in effect summarizes how

investment behaves relative to its usual behavior following a legislated change in taxes.

To help address the fact that there were enormous macroeconomic fluctuations in this era, we

consider two variations on this specification. The first includes the overall index of industrial production

(in logs) as another variable in the VAR. This controls for the effects of movements in overall economic

activity prior to the tax changes. The second and larger variation includes the contemporaneous value and

twelve lags of industrial production but treats them as exogenous. Thus, this specification asks how

investment behaves in the wake of tax changes given the path of overall economic activity following the

changes. The specification is reasonable if the effects of the tax changes on the overall economy are

small, which is plausible in light of their small impact on aggregate demand discussed in Section II.

Under this assumption, this specification can help address the possibility that the effects of the tax

changes might be swamped by the large cyclical fluctuations of this period.

When we consider interest rates, the variables are slightly different. The two variables in our

basic VAR are the ratio of the municipal bond interest rate to the AAA corporate bond interest rate and

29

the policy-induced change in 1 minus the average marginal rate at the top of the income distribution. As

described above, if the taxpayers we consider are the ones relevant to the determination of the relative

interest rates, one would expect the interest-rate ratio to move one-for-one with the tax variable.

Our basic specification uses data back to 1919:1 when it is available; the end date is 1941:12.

Since the VAR includes 24 lags, this means that for business construction, incorporations, and the

interest-rate ratio, the sample period is 1921:1–1941:12. Because the machinery data only begin in 1923,

for that VAR the basic sample period is 1925:1–1941:12.

As with our analysis of the responses of taxable income, we consider the effects of limiting the

sample to the period when there were few changes in capital-gains or corporate taxes (so the 1921:1

starting dates are changed to 1923:1, and the end date is 1932:12). We also examine the effects of

limiting it to the period before the Depression (so the ending date is 1929:8, the month of the NBER

business cycle peak), and of excluding the tax increases leading up to World War II (so the end date is

1939:12). We do not consider the pre-Depression sample for the machinery VAR, since that sample

would consist of only 56 observations.

As we have described, when taxes were changed retroactively, we date the changes as occurring

when they were passed. But we also consider a variation where we date them retroactively to the times

they went into effect. This would be appropriate if investors had considerable information about tax

changes well before they were enacted.

Basic Results. Figure 7 shows the impulse response functions of machinery investment, business

construction, and incorporations to a 1-percentage-point innovation to the measure of policy-induced

changes in the after-tax share. The figure also shows the 1-standard-error confidence bands.

The results largely echo the patterns suggested by the simple plots in Figure 6. For machinery

investment and commercial and industrial construction, there is no evidence that increases in marginal

rates lower investment. For machinery, the estimated impact of an increase in the after-tax share is

positive, small, and insignificant for a few months, then consistently negative. A 1 percent increase in the

after-tax share is followed by a fall in machinery investment of about 1½ percent. From 12 to 21 months

30

after the change, the t-statistic exceeds 2; at other horizons, the null hypothesis of no effect cannot be

rejected.

For the construction measure, there is no evidence of any relationship. The point estimates are

negative and insignificant at short horizons, and positive and highly insignificant at long horizons.

For business incorporations, in contrast, there is evidence of a positive effect of tax reductions.

For the first year, the estimated effect of a 1 percent rise in the after-tax share is generally positive, but

irregular and almost always insignificant. After about a year, however, the estimated impact rises rapidly

to almost 1 percent and is highly significant, with the t-statistic often over 3.

Figure 8 shows the estimated response of the interest-rate ratio to a fall in marginal rates of 1

percentage point. The response is at best modest. The hypothesis that it is one-for-one is

overwhelmingly rejected at all horizons, and at medium and long horizons the estimates are negative

rather than positive. The estimates are positive only for the first eight months, and even then they are

small and never significantly different from zero.

Robustness. The findings for business incorporations and interest rates are extremely robust to

all of the variations in specification, sample, and measurement of the tax variable described above. The

largest changes in the estimates occur when retroactive tax changes are dated when they took effect rather

than at the later date when they were passed. For incorporations, the response is smaller (roughly 0.8

rather than 1.0) and no longer overwhelmingly significant (the maximum t-statistic is 2.4 rather than 3.6),

but not dramatically changed. For the ratio of tax-free to taxable interest rates, the short-run response

changes from insignificantly positive to significantly negative (that is, it is in the opposite direction from

what one would expect), but the response at longer horizons is little changed. The same is true for the

sample ending in 1939:12.

For both machinery production and commercial and industrial construction, we never find a

significant positive response to increases in the after-tax share. For machinery production, however, the

result that the impact is significantly negative at some horizons is not robust: the estimated effects are

generally negative but insignificant when we include industrial production as another variable in the

31

VAR, when we treat industrial production as exogenous, when we use the alternative dating of the timing

of the retroactive changes, and when we end the sample in 1939:12. For commercial and industrial

construction, where the estimated effect has no consistent sign in the baseline specification, there are

some variations where it is generally negative and some where it is generally positive. But there are none

where it is significantly positive (and one—the 1923–1932 sample—where it is significantly negative).

As noted above, taxes were generally falling and incorporations generally rising in the 1920s, and

taxes were generally rising and incorporations generally falling in the 1930s. If this pattern were driving

the results for incorporations, there would be reason to be concerned that other secular changes might be

responsible for the correlation. To check for this possibility, we add a dummy variable equal to 1

beginning in 1929:9. This addition reduces the estimated impact almost in half and cuts its statistical

significance somewhat. But the effect remains substantial and significant: the peak effect is 0.6, and the

maximum t-statistic is 2.9. Thus, although the decadal swings in taxes and incorporations are

contributing to the estimates, they are not their main source.

It is also important to note that because of lags in the VAR, the regressions do not include 1919,

which is a time when incorporations and tax changes moved strongly in opposite directions. It is likely

that if 1919 entered the estimation, the positive impact of a rise in the after-tax share on incorporations

would be substantially reduced.

D. Discussion

The finding that is most straightforward to interpret concerns the relative interest rates on tax-free

and taxable bonds. The natural interpretation is twofold. The first part concerns the anomalous behavior

of the interest-rate ratio following the 1932 tax increase: the obvious possibility is that this pattern is not

a consequence of the tax change, but related to the extreme financial stress of this period, the threats of

defaults by municipal governments, and the full-blown financial crisis of early 1933. The second part

concerns the remainder of the sample. What appears to be occurring is that the taxpayers at the top of the

income distribution (who, as described in Section II, paid the vast majority of taxes) were inframarginal,

32

and that the relative interest rates were determined by taxpayers who faced marginal rates that were much

lower and moved much less over this period. As a result, the interest-rate ratio moved in the direction one

would expect on the basis of the changes in the marginal rates faced by high-income taxpayers, but the

movements were muted. This finding is similar in sprit to the findings of the time-series/cross-section

analysis of taxable income: taxes were distortionary, but the distortions were small.

The findings for the other measures are somewhat harder to interpret. Taken at face value, they

suggest no important effect of tax cuts on machinery investment and commercial and industrial

construction, with hints of a possible perverse effect. And they suggest a positive effect of tax cuts on

business formation.

As noted at the start of this section, however, the results of this analysis need to be interpreted

cautiously. The data are imperfect and not necessarily fully consistent over time (especially in the case of

incorporations), and during this period the economy was subject to a series of enormous shocks with far-

reaching consequences. But the result about business incorporations is suggestive. If it holds up under

further scrutiny—either of the interwar period or of other settings—it could mean that, despite the

apparent lack of strong effects on taxable income and investment, marginal tax rates may have substantial

effects on long-run economic performance.

V. CONCLUSION

Determining the incentive effects of marginal tax rates is important for welfare and public policy.

This paper shows that the interwar United States provides an excellent laboratory for investigating this

issue. Changes in marginal rates were frequent, often dramatic, and very heterogeneous across income

groups. In addition, income taxes were paid almost entirely by the wealthy, and changes in the overall

level of taxes were generally modest relative to the scale of the economy and accompanied by changes in

government spending similar in direction and size to the changes in taxes. As a result, the main channel

through which changes in taxes are likely to have affected economic performance is through their impact

on incentives.

33

We use this laboratory to examine the incentive effects of marginal rates in two ways. The first is

through time-series/cross-section regressions estimating the responsiveness of taxable income to marginal

rates. The use of time-series/cross-section data allows us to control for potential sources of differential

trends in income across sub-groups of taxpayers, and more importantly, for aggregate shocks affecting

taxpayers’ overall taxable income.

The estimates have four important features. First, consistent with what one would expect given

the tremendous identifying variation, they are very precise. The standard errors of the estimates of

responsiveness are about three times smaller than those in postwar studies of the population as a whole,

and almost ten times smaller than those in postwar estimates for the wealthy. Second, they show that

taxes are indeed distortionary: the null hypothesis of no effect is overwhelmingly rejected. Third, they

indicate that the distortions are small. Our baseline estimate of the elasticity of taxable income with

respect to the after-tax share (that is, one minus the marginal tax rate) is approximately 0.2. This is

considerably smaller than the findings of postwar studies (though generally within their confidence

intervals). If this elasticity characterized the entire population, the constant tax rate that maximized

revenue would be over 80 percent. Finally, the estimates are extremely robust.

Our second approach is more speculative. We ask how the time-series variation in overall

marginal rates is related to indicators of investment and entrepreneurial activity that could be important to

long-run growth. We find clear evidence that one major concern of interwar policymakers was of little

importance: the changes in marginal rates in this era had at most small effects on the relative interest

rates on municipal and corporate bonds, and so caused only small distortions in the incentives for the

composition of investment on this dimension. We find no evidence that high marginal rates reduced

machinery investment or business construction, but suggestive evidence that they reduced business

formation. This last result, which clearly deserves further study, identifies one possible channel through

which high marginal rates might have had important consequences.

The obvious disadvantage of the interwar period for studying the effects of marginal rates is that

the economic environment was very different from today’s. For example, wealthy taxpayers earned much

34

less of their income from wages and salaries than do their modern counterparts, and the tax enforcement

regime and the structure of the tax code were markedly different. Thus, one important question raised by

our findings is whether changes over the past three-quarters of a century are likely to have substantially

increased the distortionary effects of high marginal rates. Another is whether there are features of the

interwar tax system—most obviously, its comparative simplicity—that contributed to its relatively low

distortionary effects and that could help guide changes in the tax system today.

35

Table 1 Interwar Tax Legislation

Act Revenue Change in Top Nature of Tax (Date Enacted) Estimate Marginal Rate Change (Percentage Points)

Revenue Act of 1918 +$1,608 million +10 (1918) Raised normal tax rates in 1918 (2/24/19) +2.05 % of GDP –4 (1919) and then lowered partially in 1919; raised surtax rates; introduced war-profits tax Revenue Act of 1921 –$835 million –15 Reduced surtax rates; changed (11/23/21) –1.14 % of GDP treatment of capital gains Revenue Act of 1924 –$341 million –14.5 (1923) Reduced both normal and surtax (6/2/24) –0.39 % of GDP + 2.5 (1924) rates by roughly 25 percent Revenue Act of 1926 –$326 million –21 Reduced surtax rates roughly in (2/26/26) –0.34 % of GDP half; large increase in personal exemption Revenue Act of 1928 –$233 million 0 Increased earned-income credit; (5/29/28) –0.24 % of GDP primarily changed corporate income tax Joint Resolution No. 133 –$160 million –1 (1929) Temporarily reduced the normal (12/16/29) –0.15 % of GDP +1 (1930) personal income tax and the corporate income tax by 1 point Revenue Act of 1932 +$1,121 million +38 Raised normal and surtax rates; (6/6/32) +1.91 % of GDP surtax rates doubled at most income levels; raised corporate income tax and excise taxes Revenue Act of 1934 +$258 million 0 Rearranged normal and surtax (5/10/34) +0.39 % of GDP rates; changed treatment of capital gains; closed loopholes Revenue Act of 1935 +$270 million +16 Raised surtax rates on incomes (8/30/35) +0.37 % of GDP over $50,000; raised estate tax; established graduated corporate income tax Revenue Act of 1936 +$620 million 0 No change in personal tax rates; (6/22/36) +0.74 % of GDP subjected dividends to normal tax; large change in corporate tax, including graduated tax on undistributed profits

36

Table 1 (Continued) Interwar Tax Legislation

Act Revenue Change in Top Nature of Tax (Date Enacted) Estimate Marginal Rate Change (Percentage Points)

Revenue Act of 1937 Trivial 0 Raised surtax on undistributed net (8/26/37) income of personal holding companies; closed loopholes

Revenue Act of 1938 Trivial 0 Changed treatment of capital (5/28/38) gains so tax depends on how long asset was held; largely eliminated undistributed profits tax; made other fundamental changes in corporate income tax Revenue Act of 1940 +$1,004 million +7.9 Lowered personal exemption; (6/25/40) + 0.99 % of GDP raised surtax rates on incomes between $6,000 and $100,000; temporary “defense tax” equal to 10 percent of all regular taxes Revenue Act of 1941 +$3,500 million –5.9 Raised surtax rates dramatically (9/20/41) +2.76 % of GDP except at very top; subjected all income levels to surtax; reduced personal exemption

37

Table 2 Time-Series/Cross-Section Results

Estimation Lags Control Elasticity of Taxable Method Included Variables Income with Respect to After-Tax Sharea

(1) OLS None Year, group 0.189 dummies (0.031) (2) IV None Year, group 0.189 dummies (0.033) (3) OLS 1 Year, group 0.312 dummies (0.048) (4) OLS 2 Year, group 0.272 dummies (0.057) (5) OLS None Year dummies 0.191 (0.031) (6) OLS None Group dummies 0.090 (0.067) (7) OLS None Group dummies, 0.067 real GDP growth (0.047)

Notes: Standard errors are in parentheses. aWhen lags of the tax variable is included (rows 3 and 4), coefficients and standard errors are for sum of coefficients.

38

Table 3 Robustness of Time-Series/Cross-Section Results

Specification Elasticity of Taxable Income with Respect to After-Tax Share (1) Baseline 0.189 (0.031) (2) Using tax rates applied 0.120

retroactively, rather than (0.038) tax rates in effect when income was earned

(3) Using baseline tax measure, 0.174 but including difference (0.037) between rate applied retro- actively and baseline measurea (4) Include separate group 0.191 dummies by decade (0.032) (5) Include group dummies 0.168 interacted with real GDP (0.035) growth (6) Exclude top 1/200th of 1 0.160 percent of income distribution (0.027) (7) Exclude war years (1919 0.288 and 1940–1941) (0.039) (8) Restrict sample to period 0.378 of stable capital gains and (0.037) corporate taxes (1923–1932) (9) Pre-Depression sample 0.198 (1919–1929) (0.044) (10) Depression sample (1930– 0.186 1941) (0.046)

Notes: All regressions are estimated by OLS and include group and year dummies, and are estimated with no lags. Standard errors are in parentheses. aThe coefficient on difference is –0.034, with a standard error of 0.047.

39

0

10

20

30

40

50

60

70

80

90

10019

1819

1919

2019

2119

2219

2319

2419

2519

2619

2719

2819

2919

3019

3119

3219

3319

3419

3519

3619

3719

3819

3919

4019

41

Perc

ent

Figure 1Top Marginal Tax Rate

40

Note: Each line represents the change for a given tenth of the top 1/20th of 1 percent of the income distribution.

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.419

1919

2019

2119

2219

2319

2419

2519

2619

2719

2819

2919

3019

3119

3219

3319

3419

3519

3619

3719

3819

3919

4019

41

Cha

nge

in lo

garit

hms

Figure 2Policy-Induced Change in the Log After-Tax Share for Different Percentile Groups

1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th

41

0

10

20

30

40

50

60

70

8019

1819

1919

2019

2119

2219

2319

2419

2519

2619

2719

2819

2919

3019

3119

3219

3319

3419

3519

3619

3719

3819

3919

4019

41

Perc

ent

Figure 3Percent of Total Income Tax Paid by Tenths of the Top 1/20th of 1 Percent of

the Income Distribution (Cumulative)

1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th

Top 1/20 of 1 percent

Top 1/200 of 1 percent

42

Notes: Both variables are expressed as residuals from a regression on the year and group dummy variables. The labels on particular observations report the tenth of the top 1/20th of 1 percent of the income distribution (1 to 10) and which year (1919 to 1941) the observation corresponds to.

-0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5

Cha

nge

in R

eal R

epor

ted

Net

Inco

me

Policy-Induced Change in Log After-Tax Share

Figure 4Scatter Plot of Change in After-tax Share and Change in Real Reported Net Income

1, 1932

2, 19191, 19362, 1932

1, 1919

1, 1926

1, 1922

1, 1941

43

-0.4-0.3-0.2-0.100.10.20.3

0

50

100

150

200

250Ja

n-19

Jan-

20Ja

n-21

Jan-

22Ja

n-23

Jan-

24Ja

n-25

Jan-

26Ja

n-27

Jan-

28Ja

n-29

Jan-

30Ja

n-31

Jan-

32Ja

n-33

Jan-

34Ja

n-35

Jan-

36Ja

n-37

Jan-

38Ja

n-39

Jan-

40Ja

n-41

Polic

y-In

duce

d C

hang

e in

Afte

r-Tax

Sh

are

(Cha

nge

in L

ogar

ithm

s)

Mac

hine

ry In

dex

(193

5-39

+100

)Figure 5

Investment and Policy-Induced Changes in the After-Tax Share

a. Industrial Production of Machinery

0

2

4

6

8

10

12

Jan…

Jan…

Jan…

Jan…

Jan…

Jan…

Jan…

Jan…

Jan…

Jan…

Jan…

Jan…

Jan…

Jan…

Jan…

Jan…

Jan…

Jan…

Jan…

Jan…

Jan…

Jan…

Jan…

-0.4-0.3-0.2-0.100.10.20.3

Rea

l Con

tract

s

Polic

y-In

duce

d C

hang

e in

Afte

r-Ta

x Sh

are

(Cha

nge

in L

ogar

ithm

s)

b. Commercial and Industrial Building Contracts

0

20

40

60

80

100

120

140

Jan-

19Ja

n-20

Jan-

21Ja

n-22

Jan-

23Ja

n-24

Jan-

25Ja

n-26

Jan-

27Ja

n-28

Jan-

29Ja

n-30

Jan-

31Ja

n-32

Jan-

33Ja

n-34

Jan-

35Ja

n-36

Jan-

37Ja

n-38

Jan-

39Ja

n-40

Jan-

41

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

Inco

rpor

atio

ns (1

925=

100)

Polic

y-In

duce

d C

hang

e in

Afte

r-Ta

x Sh

are

(Cha

nge

in L

ogar

ithm

s)

c. Index of Incorporations

44

0

20

40

60

80

100

120Ja

n-19

Jan-

20

Jan-

21

Jan-

22

Jan-

23

Jan-

24

Jan-

25

Jan-

26

Jan-

27

Jan-

28

Jan-

29

Jan-

30

Jan-

31

Jan-

32

Jan-

33

Jan-

34

Jan-

35

Jan-

36

Jan-

37

Jan-

38

Jan-

39

Jan-

40

Jan-

41

Perc

ent

Figure 6Ratio of Municipal Bond Rate to AAA Rate and After-Tax Share

Interest Rate Ratio

After-Tax Share

45

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

Perc

ent

Figure 7Impulse Response of Investment to Policy-Induced Change in After-Tax Share

a. Industrial Production of Machinery

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

Perc

ent

b. Commercial and Industrial Building Contracts

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49-0.5

0.0

0.5

1.0

1.5

2.0

Perc

ent

c. Index of Incorporations

46

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48

Perc

ent

Figure 8Impulse Response of Interest Rate Ratio to After-Tax Share

47

APPENDIX 1 A NARRATIVE HISTORY OF INTERWAR TAX CHANGES

Revenue Act of 1918 Enacted: February 24, 1919

The motivation for the Revenue Act of 1918 was to raise revenue to cover the additional expenditures related to World War I. President Wilson addressed a joint session of Congress on May 27, 1918 to ask members “to prolong your session long enough to provide more adequate resources for the Treasury for the conduct of the war” (speech reproduced in “Hearings before the Committee on Ways and Means, House of Representatives, on the Proposed Revenue Act of 1918, Part I: Income, Excess Profits, and Estate Taxes, June 7 to July 17 and August 5, 14, and 15, 1918, p. 5). Wilson argued that “Additional revenues must manifestly be provided for. It would be a most unsound policy to raise too large a proportion of them by loan” (p. 5). Wilson also emphasized that “We can not in fairness wait until the end of the fiscal year is at hand to apprise our people of the taxes they must pay on their earnings of the present calendar year, whose accountings and expenditures will then be closed” (p. 5). However, this ended up to be exactly what happened. The act was not passed until February 1919 and raised taxes retroactively in 1918.

Secretary of the Treasury William McAdoo followed up on Wilson’s speech with a letter to the

chairman of the Ways and Means Committee on June 5, 1918 with a detailed recommendation. McAdoo estimated that expenditures for the 1919 fiscal year (July 1, 1918 to June 30, 1919) would be roughly $24 billion, approximately twice what they were in the previous year (letter also reproduced in “Hearings before the Committee on Ways and Means, House of Representatives,” p. 9). He argued that failing to raise taxes to pay for some of the increase “would be a surrender to the policy of high interest rates and inflation, with all of the evil consequences which would flow inevitably therefrom, and which would … bring ultimate disaster to the country” (p. 9). McAdoo specified $8 billion, or one-third of total expenditures, as the amount of revenue that should be raised in fiscal year 1919. He also made numerous recommendations about the form of the tax increase. In particular, he supported imposing a war-profits tax in addition to the existing excess-profits tax, an increase in the normal tax on unearned income, and “heavy taxation … upon all luxuries” (p. 12).

` The Ways and Means Committee produced a bill be September that followed fairly closely

McAdoo’s suggestions. One aspect of the bill that was common at the time was that it was designed to replace existing revenue laws and stand on its own. The Ways and Means Committee report on the bill states: “Your committee has endeavored to wipe out all inequalities in the operation of existing law and recommends the repeal of the major portions of the revenue acts of 1916 and 1917 in order that the existing internal-revenue laws so far as deemed practicable will be in one act and therefore more readily accessible to the taxpayer” (“Revenue Bill of 1918,” 65th Congress, 2d Session, House of Representatives Report No. 767, 9/3/18, p. 2).

The Senate did not take up the bill until the lame-duck session in December 1918. By this time,

two events had changed conditions appreciably. One was the introduction of Prohibition, which reduced expected beverage tax revenue by more than $1 billion per year. The second was the abrupt end of World War I in November 1918. According to the Senate Finance Committee report, “Taxes which can be easily borne amid the feverish activity and patriotic fervor of war times, are neither so welcome nor so easily sustained amid the uncertainties, the depreciating inventories, and the falling markets which are apt to mark the approach of peace” (“Revenue Bill of 1918,” 65th Congress, 3d Session, Senate Report No. 617, 12/6/18, p. 2). Because expenditures in fiscal year 1919 were now likely to be only $18 billion, the

48

Secretary of the Treasury recommended and the Senate Finance Committee endorsed a bill that raised $6 billion in revenue (p. 2).

According to a table in the Senate report, the existing law would raise $4.370 billion in fiscal year

1919, whereas the proposed law would raise $5.978 billion, or an increase of $1.608 billion (p. 3). The Senate Finance Committee recommended that the war-profits tax be largely eliminated in 1919, the excess-profits tax rates be reduced substantially, and the normal tax on personal and corporate incomes be reduced by one-third. The report estimated “that these changes would reduce the revenue for 1920 as compared with 1919 by approximately $1,400,000,000” (p. 3). Thus, the Senate version contained a large net tax increase in fiscal year 1919 (calendar year 1918), and a near-return to previous tax levels in fiscal year 1920 (calendar year 1919). The final act was very similar to the Senate version. An academic article from June 1919 gives revenue estimates for the act very similar to those in the Senate report (Roy G. Blakey and Gladys C. Blakey, “The Revenue Act of 1918,” American Economic Review, 9 (June 1919): 213-243, p. 216).

As described, the Revenue Act of 1918 was a large, retroactive tax increase affecting primarily calendar year 1918. However, while the revenue effects were largely felt in fiscal year 1919, it had more long-lasting effects on income tax rates. The normal tax rate on the first $4000 of income was raised from 2 percent in 1917 to 6 percent in 1918 and then to 4 percent in 1919 and after. The normal tax on incomes above $4000 was raised from 4 percent in 1917 to 12 percent in 1918 and 8 percent in 1919 and after. The surtax rates were also raised for 1918 and all subsequent years. The top surtax rate only increased from 63 percent in 1917 to 65 percent in 1918, but the rates rose much more quickly with income under the Revenue Act of 1918. For example, the surtax rate on net incomes between $50,000 and $52,000 rose from 12 percent in 1917 to 24 percent in 1918. In general, for incomes between $50,000 and $150,000, surtax rates were roughly double under the Revenue Act of 1918.

Corporate income tax rates were also raised. The rate rose from a top value of 4 percent of net

income in 1917 to 12 percent in 1918 and 10 percent in 1919 and subsequent years (“Revenue Bill of 1918: An Analysis of the Bill (H.R. 12863) to Provide Revenue, and for Other Purposes,” 65th Congress, 3d Session, Senate Document No. 391, February 13, 1919, p. 5).

The act included a war-profits tax in 1918 equal to 80 percent of the excess of net income of the

corporation over its prewar profits (with an adjustment for additional capital invested). There was also an increase in the excess-profits tax. In 1917, tax rates ranged from 20 to 60 percent of net income in excess of various fractions of invested capital. In the 1918 law, cutoffs income levels were lowered and rates were raised. The rate was 30 percent of net income in excess of 8 percent of invested capital (and not in excess of 20 percent of invested capital), plus 65 percent of net income in excess of 20 percent of invested capital. Corporations paid whichever of the war-profits and excess-profits tax was larger. The war-profits tax was essentially eliminated for 1919, and the excess-profits tax rates were reduced from 30 and 65 percent to 20 and 40 percent in 1919 (see 1930 Statistics of Income, pp. 322-324, for a summary of the rates in various years).

Finally, the Revenue Act of 1918 included a wide array of other tax changes. It lowered the

estate tax slightly, raised taxes on tobacco products, and raised excise taxes on automobiles, jewelry, phonographs, and a number of other luxury goods. It also included a tax of 10 percent on the net profits of any business which employed children under a certain age.

49

Revenue Act of 1921 Enacted: November 23, 1921 The tax cut contained in the Revenue Act of 1921 was unquestionably tied to the reduction in spending following World War I. According to the 1921 Treasury Annual Report, government spending had fallen not only because of the end of hostilities, but also because of the introduction of an explicit budget process and the creation of the Bureau of the Budget (p. 1). The mindset of policymakers was very clear that tax reductions could only occur if spending declined; deficit-inducing tax cuts were not contemplated. The link between spending declines and tax declines is very clear in the size of the final tax cut. The 1921 Treasury Annual Report states that because of new lower estimates of expenditure for fiscal years 1922 and 1923, the Treasury tax proposal was changed to eliminate some tax increases that had been thought necessary to keep the budget in balance (p. 6). Treasury Secretary Mellon’s statement on August 13, 1921 took as given that the fundamental requirement for a tax bill was that it “raise the needed revenue within reasonable certainty” (Treasury Annual Report, Exhibit 72, p. 372). Likewise, President Harding echoed this same view in his First Annual Message. In discussing policy in the future, he stated: “By your sustainment of the rigid economies already inaugurated, with hoped-for extension of these economies and added efficiencies in administration, I believe further reductions may be enacted” (12/6/21, p. 2). The role of the reduction in spending is also evident in the Senate Finance Committee report on the bill. It stated: “The revenue bill which your committee recommends is designed to produce enough revenue to meet without borrowing all ordinary expenditure, … but not enough to create a current surplus and thus encourage unnecessary spending” (“Internal Revenue Bill of 1921,” 67th Congress, 1st Session, Senate Report No. 275, September 1921, p. 1). The committee appears to have felt not only that it was appropriate that taxes fall when spending declines, but also that reducing taxes encouraged government economy (or discouraged profligate spending). While the reduction in spending was the proximate cause (or a necessary condition) for tax reduction, the form of the tax cut certainly reflected beliefs about incentives. The 1921 Treasury Annual Report had an extensive discussion of the dangers of high surtax rates. One argument was that “the higher surtax rates are rapidly ceasing to be productive of revenue,” because they provided strong incentives for tax avoidance (p. 14). In particular, “[t]here is no doubt that a large and steadily increasing amount of money formerly invested in productive industry is now going into tax-exempt securities” (p. 14). The 1921 Report also argued that high surtax rates were discouraging saving and capital formation, and that this was very bad for future growth and standards of living. More generally, it worried that “[a]nother serious effect of these high tax rates is the destruction of incentive—the drying up of the activities of individuals in trade operations—with consequent lessening of business transactions, the slowing down of production, and ultimately a loss of revenue to the Government” (p. 16). There can be no question that the nature of the tax cut had a very strong philosophical component. The Senate report on the bill did not give much information on the reason for the form of the tax reduction. The one thing that it did emphasize was the motivation for eliminating the excess profits tax. It said: “The repeal of this tax is recommended because of its inequalities and difficulty of administration and because of the manner in which it discriminates against corporations with small invested capital” (Senate Report No. 275, p. 21). The size of the tax reduction was substantial. The 1921 Treasury Annual Report said it reduced revenues by $835 million relative to existing law during the first full fiscal year it was in operation, which

50

was 1923 (p.10). This accords with President Harding’s description of it as a “billion dollar reduction” (First Annual Message, 12/6/21, p. 2). The Revenue Act of 1921 changed taxes along a number of dimensions. Most obviously, it reduced surtax rates (while leaving the normal tax unchanged). For incomes up to $100,000, the act reduced surtax rates by just 1 percentage point. But on higher incomes, the reduction was more substantial. The top surtax rate was reduced from 65 percent to 50 percent. The Revenue Act of 1921 also greatly affected the tax treatment of capital gains. Rather than being taxed at the surtax rate, taxpayers were allowed to pay the lower of the surtax rate or 12½ percent (the surtax rate hit 13 percent at taxable incomes of $30,000). Various excise taxes were also eliminated by the act. The excess profits tax, which had been introduced during the war, was eliminated, but a flat 2½ percentage points was added to the corporation income tax. The tax changes all took effect on January 1, 1922. Revenue Act of 1924 Enacted June 2, 1924 The fundamental motivation for the Revenue Act of 1924 was a belief in the virtue of limited government. President Coolidge and his Secretary of the Treasury Andrew Mellon spoke eloquently of the dangers of high taxation. For example, in his Address at the Meeting of the Business Organization of the Government in June 1924, Coolidge said: “any oppression laid upon the people by excessive taxation, any disregard of their right to hold and enjoy the property which they have rightfully acquired, would be fatal to freedom” (6/30/24, p. 1). Like virtually all the tax changes in the interwar period, this tax change was tied to spending changes. The notion of a tax cut without a budget surplus or a reduction in spending was simply not contemplated. Coolidge was an ardent proponent of reducing spending so taxes could be reduced. In December 1923, for example, Coolidge discussed the fact that spending had been reduced substantially. He went on to say: “It is possible, in consequence, to make a large reduction in the taxes of the people, which is the sole object of all curtailment” (First Annual Message, 12/6/23, p. 3). Likewise, in his June 1924 address, he stated: “this fight for economy had but one purpose; that its benefits would accrue to the whole people through reduction in taxes” (6/30/24, p. 2). The 1924 Treasury Annual Report stated: “As a result of this reduction in expenditures [since 1920] two revenue relief measures were made possible, the revenue act of 1921 and the revenue act of 1924” (p. 2). The Ways and Means Committee report on the bill also made it clear that the current budget surplus was a driving force behind the tax cut. It quoted a letter from the Secretary of the Treasury that said surpluses of over $300 million a year were anticipated for the next four or five years. The report then said: “These figures indicate beyond any question that present taxes are yielding more revenue than the needs of government demand. Tax reduction is therefore imperative” (“The Revenue Bill of 1924,” 68th Congress, 1st Session, House of Representatives Report No. 179, 2/11/24, p. 1). The committee urged “the reduction of taxes to the full extent justified by the Treasury surplus” (p. 1). The Treasury proposal for the bill included a much larger reduction in top marginal rates than the act ultimately included. There can be no doubt that the form of the tax cut proposed was motivated by a belief that high marginal rates distorted incentives. The administration felt that high wartime rates were gradually generating less revenue as people responded to the incentives they provided for tax evasion (1923 Treasury Annual Report, pp. 4-5). In a letter to the chairman of the Ways and Means Committee on November 10, 1923, Secretary of the Treasury Andrew Mellon spelled out these arguments. He said:

51

The high rates put pressure on taxpayers to reduce their taxable income, tend to destroy individual initiative and enterprise, and seriously impede the development of productive business. Taxpayers subject to the higher rates can not afford, for example, to invest in American railroads or industries or embark upon new enterprises in the face of taxes that will tax 50 per cent or more of any return that may be realized. These taxpayers are withdrawing their capital from productive business and investing it instead in tax-exempt securities and adopting other lawful methods of avoiding the realization of taxable income (1923 Treasury Annual Report, p. 8).

The administration was sufficiently concerned about the distortion of incentives regarding tax-exempt securities that, in addition to reducing marginal tax rates, it proposed a constitutional amendment abolishing the right of states and municipalities to issue such securities. President Coolidge endorsed Mellon’s supply-side views in his First Annual Message. He said: “a revision of the surtaxes will not only provide additional money for capital investment, thus stimulating industry and employing more but will not greatly reduce the revenue from that source, and may in the future actually increase it” (12/6/23, p. 3). The Ways and Means Committee report also emphasized the importance of tax reform. In addition to tax reduction, it sought to “simplify the law, so far as possible, and endeavor to close the gaps which now give opportunity to evade its provisions” (House of Representatives Report No. 179, p. 1). In justifying the reduction in marginal rates, the report quoted at length from Mellon’s November letter and statements from two former Secretaries of the Treasury. The report showed less commitment to the extreme reduction in top rates that the Treasury proposed. It said: “In making his recommendations, the purpose of the Secretary was obviously to fix the maximum surtax rates … at the point of maximum productivity. It is, of course, impossible accurately to determine at what rate of tax this point is reached, but it seems to be generally conceded that a 50 per cent surtax has a constantly increasing effect in creating evasions, and that it is inadvisable for other reasons” (p. 5). This less than full commitment to the 25 percent top rate may explain why the ultimate reduction was only to 40 percent. The House report also contained a lengthy discussion of the reason for the proposed reduction of the tax on earned income. The basic argument was one of fairness. It stated: “The taxpayer who receives salaries, wages, and other earned income must each year save and set aside a portion of his income in order to protect him in case of sickness and in his old age, and in order to provide for his family upon his death. On the other hand, the person whose income is derived from investments already has his capital and is relieved of the necessity of saving to establish it” (p. 5). Coolidge and Mellon were unhappy with the final form of the bill. The 1924 Treasury Annual Report stated that the President “viewed the bill as a measure of temporary relief but not a genuine tax reform” (p. 4; see also Statement by President Coolidge Concerning the Revenue Bill of 1924, Exhibit 56, 1924 Treasury Annual Report, pp. 264-267). They were concerned that the reduction in marginal rates was not sufficient to limit efforts at tax evasion (1924 Treasury Annual Report, pp. 3-6). The revenue effects of the bill were not expected to be large. However, revenue estimates for this time period are highly inexact, and it is unclear how policymakers were treating the potential growth effects of the bill. The most concrete estimates are for the Treasury’s initial proposal. The 1923 Treasury Annual Report said the plan would cut revenues by $323 million (p. 10). The House report said that in a full year of operative, the form of the bill that it proposed would result in a net revenue loss of $341,440,000 (House of Representatives Report No. 179, p. 2). The Revenue Act of 1924 changed taxes in a number of ways. It reduced normal tax rates from 4 to 8 percent to 2 to 6 percent, and lowered surtax rates by approximately 25 percent at most income levels (20 percent at the very top). The top surtax rate was reduced from 50 percent to 40 percent. The

52

new rates applied to income earned in 1924. In addition, the tax from the surtax for 1923 was reduced retroactively by 25 percent. The Revenue Act of 1924 continued the wartime estate tax and added a gift tax as a means of stemming avoidance. It also established a credit of 25 percent of the normal tax on earned income up to $10,000. Revenue Act of 1926 Enacted February 26, 1926 In proposing and passing the Revenue Act of 1926, President Coolidge emphasized the need for tax reform over the need for tax reduction. The groundwork for this action was laid in Coolidge’s signing statement on the Revenue Act of 1924. He stated: “The bill as passed provides a certain amount of tax reduction. … But it is not only lacking in tax reform, it actually adds some undesirable features to the present law” (1924 Treasury Annual Report, Exhibit 56, p. 264). Among the key reforms that Coolidge and Mellon wanted were a reduction in high surtax rates and a repeal of the estate tax (1924 Treasury Annual Report, p. 4). The Coolidge administration gave many justifications for these reforms. One argument was the practical one that high tax rates make efforts at avoidance inevitable (1924 Treasury Annual Report, p. 6; see also, p. 264). Mellon went even further in his statement before the Ways and Means Committee of the House on October 19, 1925. He stated: “it is important to bear in mind the distinction between a reduction of taxes which reforms the tax system and a reduction in taxes which simply reduces revenue. It has been the experience of the Treasury that every time there has been a material reduction in surtaxes it has stimulated business and brought about an increase in taxable income which has made up a great part, if not all, of the loss in revenue” (1925 Treasury Annual Report, Exhibit 87, p. 350). Coolidge also stressed the positive revenue effects of lower rates (see, for example, Second Annual Message, 12/3/24, p. 2). In addition to the revenue argument, Coolidge made much broader claims about the supply-side benefits of lowering marginal rates. His Third Annual Message, given just shortly before the act was passed, gave an impassioned listing of these benefits:

All these economic results are being sought not to benefit the rich, but to benefit the people. They are for the purpose of encouraging industry in order that employment may be plentiful. They seek to make business good in order that wages may be good …. They seek to lay the foundation which, through increased production, may give the people a more bountiful supply of the necessaries of life (12/8/25, p. 3).

In his Second Annual Message on December 3, 1924, he made the flip side of this argument, saying that the severe postwar recession “resulted in no small measure from the prohibitive taxes which were then levied on all productive effort” (p. 1). Coolidge also added a moral argument to his belief in lower tax rates. His Inaugural Address, for example, said: “I am opposed to extremely high rates, because they produce little or no revenue, because they are bad for the country, and, finally, because they are wrong” (3/4/25, p. 4). He added: “Under this republic the rewards of industry belong to those who earn them” (p. 4). As with the earlier tax cuts, Coolidge stressed the need to reduce government expenditure before taxes could be cut. In essence, the tax cut was to be a balanced-budget measure. Mellon made this argument very clearly in his testimony to Congress. He said: “The first matter which must be considered in any revenue bill is how much revenue the Government requires” (1925 Treasury Annual Report,

53

Exhibit 87, p. 346). He continued: “I think, however, that the surplus in 1927, … would be between $250,000,000 and $300,000,000. This, it seems to me, is a figure which it is safe to take as the amount by which taxes can now be permanently reduced” (p. 347). Coolidge made similar arguments repeatedly. For example, in his Address at a Meeting of the Business Organization of the Government, he said: “The object sought is not merely a cutting down of public expenditures. That is only the means. Tax reduction is the end” (6/22/25, p. 1). The Ways and Means Committee Report (“The Revenue Bill of 1926,” 69th Congress, 1st Session, House of Representatives Report No. 1, 12/7/25) stressed the existence of a surplus as the main justification for the cut. It said: “It was estimated that under the revenue law then in force that the fiscal year 1925 would produce a surplus of $250,000,000; the fiscal year 1926, over $200,000,000; and the fiscal year 1927, about $300,000,000. Such being the case, it was obvious that our Federal taxes ought to be reduced” (p. 1). It went on to say: “the committee, having first determined the total amount of reduction in revenues which could properly be made, proceeded to apportion the benefits of such reduction … as far as possible, to so distribute them as to bring the maximum good to all of our people” (p. 2). The expected revenue effects of the action were fairly modest. The 1927 Treasury Annual Report said that the bill was expected to reduce tax revenues by $46 million by cutting the normal tax, $98 million by cutting the surtax rates, and $42 million by increasing the personal exemption, for a total of $186 million (p. 44). This is slightly less than the $250 to $300 million that Mellon said taxes could be cut. Coolidge said in a speech on June 21, 1926 that the bill also reduced wartime excise taxes by $275 million (Address at the Eleventh Regular Meeting of the Business Organization of the Government, p. 4). Including this would increase the total to $461 million. This sum is somewhat higher than the Ways and Means Committee estimate that “[f]or the calendar year 1926 the reduction in revenues is estimated at $325,736,000” (House Report No. 1, p. 2). The tax reductions were made retroactive to January 1925. The Revenue Act of 1926 had a number of key features. The most extreme was a dramatic reduction in surtax rates on personal income. The top rate was reduced from 40 percent to 20 percent. Rates at all but very low income levels were roughly cut in half. The normal tax rate was also reduced (from a top rate of 6 percent to a top rate of 5 percent). The personal exemption was raised by 50 percent, which substantially increased the number of people paying no federal income tax (1927 Treasury Annual Report, p. 10). The bill also removed the capital stock tax on corporations, and partially replaced it by raising the corporate income tax rate by 1 percentage point (from 12½ percent to 13½ percent). Revenue Act of 1928 Enacted May 29, 1928 The Revenue Act of 1928 was a relatively minor tax action that primarily reduced the corporate income tax. It also made small adjustments to the personal income tax and certain excise taxes. The proximate motivation for the tax reduction appears to have been current and prospective budget surpluses. For example, the Ways and Means Committee report on the bill began the general discussion saying: “We are again in the happy position of having a surplus of revenue in the Treasury which is being applied on the national debt, but which enables us to reduce taxation” (“The Revenue Bill of 1928,” 70th Congress, 1st Session, House of Representatives Report No. 2, 12/7/27, p. 1). This sentiment was echoed by Treasury Secretary Mellon’s statement before the committee. He stated: “The factor which definitely determines the extent to which we may reduce taxes is the 1929 surplus” (“Revenue Revision 1927-28,” Interim, 69th-70th Congresses, Hearings before the Committee on Ways and Means, House of Representatives, October 31 to November 10, 1927, p. 6). Likewise,

54

Undersecretary of the Treasury Mills stated on November 11, 1927: “For the fourth time in seven years the state of Federal finances is such as to permit a substantial reduction of taxes” (1928 Treasury Annual Report, Exhibit 26, p. 278). President Coolidge made many statements in 1927 and 1928 to the effect that the state of the budget was the crucial determinant of tax legislation. In June 1927, he stressed the importance of considering extended budget forecasts: “In considering the possibility of tax reduction, we must keep in mind that our revenue laws can not be written from the standpoint of a single year, but must be expected to yield adequate revenue over a period of years” (Address at the Thirteenth Regular Meeting of the Business Organization of the Government, Washington D.C., 6/10/27, p. 2). In December 1927, he said: “The immediate fruit of economy and the retirement of the public debt is tax reduction.” Coolidge was explicit that budget balance was not negotiable, saying: “We must keep our budget balanced for each year” (both quotations, Fifth Annual Message, 12/6/27, p. 1). In 1928, Coolidge stated: “This Nation is committed irrevocably to balancing the Budget. Nothing short of a national emergency can trespass upon that commitment” (Address at the Fifteenth Regular Meeting of the Business Organization of the Government, 6/11/28, p. 3). This same view was expressed by the Ways and Means Committee report, which stated: “The majority of the committee are opposed to any plan that would produce a deficit” (House of Representatives Report No. 2, p. 2). Shortly after passage of the Revenue Act of 1928, Coolidge reiterated the view expressed previously that his administration was reducing spending so that it could reduce taxes. He said: “We have approached the tax question from the angle of requiring no more from the people than necessary efficiently to operate the Government. The effort has been to reduce the cost of Government so as to make room for tax reduction” (Address at the Fifteenth Regular Meeting of the Business Organization of the Government, 6/11/28, p. 2). However, in this case, it is also clear that unexpectedly large revenues played a role in causing the budget surplus driving the tax cut. In his statement to the Ways and Means Committee, Secretary Mellon stated: “The Treasury Department has always contended that lower rates would be more productive than the very high rates which prevailed [before the Revenue Act of 1926], but neither the Treasury Department nor the Congress had anticipated such an immediate increase, an increase which was, of course, greatly accelerated by the rising tide of prosperity” (“Revenue Revision 1927-28,” Interim, 69th-70th Congresses, Hearings before the Committee on Ways and Means, House of Representatives, October 31-November 10, 1927, p. 3). In his statement, Mellon suggested that budget surpluses of about $250 million were likely to continue. Taking into account possible spending increases from new legislation, he concluded: “The Treasury believes that tax reduction should not in any event be in excess of approximately $225,000,000” (Hearings, October 31-November 10, 1927, p. 7). The 1928 Treasury Annual Report suggested that the effect of the 1928 act (at a given level of income) was a reduction in tax revenues of $222 million (p. 34). This is slightly lower than the Ways and Means Committee estimate of a reduction of $232,735,000 (House of Representatives Report No. 2, p. 2). President Coolidge said the proposed tax changes “would give us a much better balanced system of taxation” (Fifth Annual Message, 12/6/27, p. 1). The key component was a reduction in the corporate income tax rate. Secretary Mellon suggested that such a reduction in corporate taxes was necessary for both fairness and efficiency. He said:

Corporations last received relief from taxation in the revenue act of 1921, which repealed the excess-profits tax, and even then the income tax rate was increased. Since that time, while other classes of taxpayers have benefited either by the repeal of war taxes or the sharp reduction of war-time rates, corporations have continued to bear a heavy burden. The time has come to revise the corporation tax rates downward. Business conducted

55

under the corporate form is to-day overtaxed as compared with individual business enterprises and partnerships, a condition which spells particular hardship to the small corporations with a limited net income and to the stockholder of limited means (Hearings, October 31-November 10, 1927, p. 8).

Undersecretary Mills, in his speech in November 1927, also stressed the unfairness of taxing corporate income more highly than other types of income. He said: “There is no logic or justice in such a discrimination” (1928 Treasury Annual Report, Exhibit 26, p. 281). Similar arguments were made by the Ways and Means Committee (see House of Representatives Report No. 2, pp. 3-4). The key feature of the Revenue Act of 1928 was a reduction in the corporate income tax rate from 13½ to 12 percent. It also raised the specific credit from $2000 to $3000. The original proposal included a reduction in individual surtax rates for incomes in the middle range of taxable incomes (see Mellon’s statement, Hearings, October 31-November 10, 1927, p. 7). However, this change did not make it into the final law. The only change in individual income taxes was an increase in the limit on which the earned income credit could be taken from $20,000 to $30,000. A proposal to abolish the estate tax also did not make it into the final bill. Joint Resolution of Congress No. 133 Enacted December 16, 1929 Joint Resolution No. 133 was a fairly small, explicitly temporary tax cut passed at the end of 1929. Like the Revenue Act of 1928, it appears to have been motivated by the existence of a modest, largely unintended budget surplus. The 1929 Treasury Annual Report said that it expected a surplus of $226 million in fiscal year 1930 and $123 million in fiscal year 1931. It continued: “The Treasury Department believes, therefore, that the taxpayers should receive the benefit of any prospective surplus in the form of tax reduction” (p. 22). The 1930 Annual Report was equally explicit about the link to the surplus, and added that the surplus was not the result of government intentions. It said: “It was then apparent that the tax yield at 1928 rates would be more than sufficient for budget requirements in the fiscal year 1930…. This was due primarily to the increase in incomes of both corporations and individuals during the years immediately preceding, especially in the calendar year 1928. Accordingly, provision was made to reduce by 1 per cent the normal rates on individual income and the rate on corporation income applicable to incomes reported for the calendar year 1929” (p. 4). President Hoover, in his Annual Budget Message to the Congress, Fiscal Year 1931, repeated the same view. He said: “With an estimated surplus … it is felt that some measure of reduction in taxes is justified” (12/4/29, p. 10). He also added the argument that there were possible efficiency gains from cutting taxes. He stated: “Such reduction gives the taxpayer correspondingly more for his own use and thus increases the capital available for general business” (p. 10). The tax cut was made explicitly temporary because the Treasury was unsure that the surpluses would continue. The 1929 Treasury Annual Report said: “the problem of estimating future revenue is attended by extraordinary difficulties at the present time due to the existence of a number of factors the effect of which it is almost impossible to foresee” (p. 23). Chief among these factors was an unusual surge in both income and capital gains in 1928, and a possible recession beginning in 1929. The 1930 Treasury Annual Report said that the 1929 tax cut “is the first instance in which income tax rates have been reduced for a single calendar year in order to relieve individuals and corporations from taxes when a surplus of receipts was anticipated without assurance that this surplus would continue for more than one year” (p. 2). Though an explicitly countercyclical motivation was not mentioned, the 1930 Annual Report

56

pointed out that “During the calendar year 1930 the income tax reduction afforded relief to both individuals and corporations during a period of unfavorable business developments” (p. 2). The Ways and Means Committee report on the bill seconded both the idea that the existing surplus was the key motive for tax reduction, and the notion that budget uncertainty was the reason for making the cut temporary. It said: “Because of the unusual conditions, your committee believes that the estimated surplus for 1930 and 1931 does not permit a permanent tax reduction. Nevertheless, it is convinced that the benefits of the probable surpluses should be passed on to the taxpayers” (“Tax Relief for 1929,” 71st Congress, 2d Session, House of Representatives Report No. 24, 12/4/29, p. 2). The key motivation given for the particular form that the tax cut took was a desire to spread the cut broadly. The 1929 Treasury Annual Report said the cut in both the normal tax rate and the corporate income tax rate “distributes the benefits as widely as possible and while giving all income taxpayers some measure of relief favors those of moderate incomes” (p. 25). It particularly stressed that since few people paid the income tax, but many held stocks and received dividends, “the way to give the greatest Federal tax relief to the greatest numbers is through a reduction of the corporation rate” (p. 25). It also pointed out that “under our system of graduated surtaxes the reduction of the normal rate is relatively of greater benefit to those with small or moderate incomes” (p. 25). The Congressional report on the bill made similar arguments for the form the tax reduction took. It stated: “the relief should be granted to the greatest extent consistent with sound Government finance and to the greatest possible number of taxpayers” (House of Representatives Report No. 24, p. 3). It also presented calculations suggesting “the present corporation rate of 12 per cent is out of line with the rates imposed upon individuals,” and said: “It can hardly be denied that the way to give the greatest Federal tax relief to the greatest number is through a reduction of the corporation rate” (both quotations, p. 3). The estimated revenue effects were relatively small. The 1929 Treasury Annual Report gave the effect in calendar year 1930 as $160 million (p. 24). This estimate is repeated in the 1930 Report (p. 4) and the Ways and Means Committee report (House of Representatives Report No. 24, p. 1). The 1930 Treasury Annual Report indicated that $90 million of this reduction came from the cut in the corporate rate and $70 million from the reduction in the individual income tax rate (p. 4). The lower rates only applied to 1929 taxes, which were paid in early 1930. As described above, the tax action took the form of reducing both the normal income tax and the corporate income tax rates by 1 percentage point. The normal tax rate, which varied from 1½, 3, and 5 percent depending on income, was reduced to ½, 2, and 4 percent; the corporate income tax rate was reduced from 12 percent to 11 percent. Revenue Act of 1932 Enacted June 6, 1932 The Revenue Act of 1932 contained a very large tax increase. The sole purpose of the tax increase was to close the budget deficit caused by the reduction in revenues (and the much smaller increase in expenditures) brought about by the Great Depression. The 1932 Treasury Annual Report contained a thorough discussion of the decline in revenues caused by the Depression. It pointed out that with a highly progressive tax system, as the United States still had despite the large tax cuts of the 1920s, tax revenues decline even more rapidly than income (p. 9). At the time the bill was being considered by Congress, the projected federal deficit for fiscal year 1933 was more than $1,700 million (1932 Treasury Annual Report, Exhibit 23, p. 259). These estimated deficits are mentioned in the first page of the Ways

57

and Means Committee report on the bill (“The Revenue Bill of 1932,” 72nd Congress, 1st Session, House of Representatives Report No. 708, 3/8/32, p. 1).

To some degree, policymakers acted as if it was obvious that eliminating a budget deficit was justification enough for the substantial tax increase. For example, Ogden Mills, who became Treasury Secretary under Hoover, said in December 1931: “over a period of years, revenues must be equal to expenditures. Deficiency for a time may be inevitable, but the principle of a balanced Budget must never be abandoned, and when emergency conditions upset the balance, every effort must be made to restore it at the earliest possible opportunity” (1932 Treasury Annual Report, Exhibit 22, p. 256). Likewise, the 1932 Treasury Annual Report described the purpose of the tax increase as “to provide additional revenue to meet the emergency situation” (p. 13).

Both Treasury Secretary Mills and President Hoover, however, went somewhat further. Hoover,

in his December 1931 Annual Budget Message said: “We can not maintain public confidence nor stability of the Federal Government without undertaking some temporary tax increases” (12/9/31, p. 2). He reiterated these motivations in his short Statement on Signing the Revenue Act of 1932, saying: “the bill will effect the great major purpose of assurance to the country and the world of the determination of the American people to maintain their finances and their currency on a sound basis” (6/6/32, p. 1).

Mills placed particular emphasis on the credit crisis facing the country. After saying that the

decline in bank credit was a key impediment to recovery, he emphasized that “our private credit structure is inextricably bound to the credit of the United States Government” (1932 Treasury Annual Report, Exhibit 23, p. 259). Like Hoover, Mills also mentioned a link between a balanced budget and currency stability. He said: “Our currency rests predominantly upon the credit of the United States. Impair that credit and every dollar you handle will be tainted with suspicion” (p. 259).

The Ways and Means Committee report made many similar arguments about why budget balance

was important. It stated: “any failure to balance the Budget for 1933 showing as it would a continuing failure in the face of known conditions to meet current expenditures out of current receipts would evidence such a lack of sound business methods in the conduct of our national finances as to cause a loss of confidence and apprehension as to the future” (House of Representatives Report No. 708, p. 4). It continued: “Our commercial credit system is inextricably tied up with the credit of the Federal Government, and anything that shakes public confidence in that credit necessarily affects the entire commercial credit system upon which business development and expansion are dependent” (p. 5).

In contrast to the 1920s, policymakers were not terribly worried about the incentive effects of the

tax increase. The administration’s proposal involved essentially reenacting the Revenue Act of 1924. Hoover stated that this plan “has the great advantage that the Government is equipped by experience with similar legislation for its systematic and economical collection” (Annual Budget Message to the Congress, Fiscal Year 1933, 12/9/31, p. 2). The one thing that Secretary Mills emphasized repeatedly was the need to make the tax increase more broadly based. He said that raising taxes only at the top of the income distribution simply would not raise adequate revenue (1932 Treasury Annual Report, Exhibit 24, p. 264). Congress expressed slightly more concern about the incentive effects of high marginal income tax rates when it said: “these increases in the rates, particularly on the higher incomes, reaching, as they do in the proposed bill, a maximum of 46 per cent, equal if they do not exceed the point of diminishing return. No more revenue can be obtained out of a tax on large incomes” (House of Representatives Report No. 708, p. 7).

The 1932 Treasury Annual Report called the Revenue Act of 1932 “one of the largest increases in

taxes ever imposed by the Federal Government in peace times” (p. 21). It also said that “In a year in which the enactment of any new revenue measure presented grave difficulties, the placing on the statute

58

books of an act so substantial in scope was an impressive achievement” (pp. 21-22). When the law was enacted, it was estimated to yield a revenue increase of $1,118.5 million for fiscal year 1933 (1932 Treasury Annual Report, p. 21). The Ways and Means Committee report gave the slightly larger revenue estimate for fiscal year 1933 of $1,121 million (House of Representatives Report No. 708, p. 5).

The tax increase had many components. It raised the normal tax from 1½ to 5 percent to 4 to 8

percent. It roughly doubled surtax rates at most income levels; the top marginal rate rose from 20 percent to 55 percent. The law also eliminated the earned income credit and reduced the personal exemption substantially. The act permanently increased the corporate income tax from 12 percent to 13¾ percent, and added a temporary extra ¾ percent tax for 1932 and 1933. While the changes in the individual and corporate income taxes were substantial, the majority of the revenue effects were due to a vast increase in excise taxes. The most notable of these was an across-the-board tax of 2¼ percent on all manufactured articles.

National Industrial Recovery Act Enacted June 16, 1933 The National Industrial Recovery Act was an extremely complex and wide-ranging piece of legislation. While modern analysis has focused largely on the anti-competitive practices and labor provisions contained in the resulting industry codes, a very large part of the act centered on public works spending. Title II of the act authorized appropriations of $3,300 million for public works and construction projects. The act provided for additional taxes to meet the service charges on the funds borrowed for this additional spending. The Roosevelt administration believed strongly that regular recurring expenditures should be covered by regular, recurring revenues. But, it viewed the Depression as a national emergency, much like a war, that warranted increased borrowing. For example, Director of the Budget Lewis Douglas testified at a hearing on the NIRA:

Recurring items of expenditure should be met out of recurring revenue. … But just as during the war there were emergency expenditures which had to be made, so now there are emergency expenditures which have to be made—expenditures of a nonrecurring and extraordinary nature. … We propose to undertake this public works program, under which the Government will have to borrow a very substantial sum of money—$3,300,000,000 (“National Industrial Recovery,” 73rd Congress, 1st Session, Hearings Before the Committee on Ways and Means, House of Representatives, No. 1, May 18, 1933, p. 29).

Likewise, in his Fireside Chat on the Recovery Program, Roosevelt talked of bringing regular expenses within our revenues. He then said: “It may seem inconsistent for a government to cut down its regular expenses and at the same time to borrow and to spend billions for an emergency. But it is not inconsistent because a large portion of the emergency money has been paid out in the form of sound loans which will be repaid to the Treasury over a period of years” (7/24/33, p. 1). The tax increases included in the NIRA were designed merely to cover the interest and amortization on the debt incurred to pay for the increased spending. In his Message to Congress Recommending Enactment of the National Industrial Recovery Act, Roosevelt stated: “In carrying out this program it is imperative that the credit of the United States Government be protected and preserved. This means that at the same time we are making these vast emergency expenditures there must be provided sufficient revenue to pay interest and amortization on the cost and that the revenues so provided

59

must be adequate and certain rather than inadequate and speculative” (5/17/33, p. 1). This view that the tax increases were designed “to meet service charges on the funds borrowed for construction of public works” was seconded by the 1933 Treasury Annual Report (p. 18). Thus, in a fundamental sense, this act included a spending-driven tax increase, but the tax increase was only a tiny fraction of the spending increase. The revenue raised by the tax increases was extremely modest. The President’s Message to Congress said: “at least $220,000,000 of additional revenue will be required to service the contemplated borrowings of the Government” (5/17/33, p. 1). The 1933 Treasury Annual Report estimated that the NIRA taxes would yield $153.7 million in fiscal year 1934 (p. 22). The NIRA included a number of tax changes. It imposed a capital stock tax of $1 per $1000 of the declared value of a corporation’s capital stock, where firms were allowed to set their declared value however they liked. To give firms better incentives, the capital stock tax was paired with an excess profits tax of 5 percent of net income in excess of 12½ percent of the corporation’s declared value. The act also imposed a tax of 5 percent on dividends received, and extended the manufacturers’ excise tax from the Revenue Act of 1932 for one year. It increased the tax on gasoline from 1 cent to 1½ cents per gallon. All of the new taxes took effect immediately. The NIRA called for the new taxes to end when one of two conditions was met: at the close of the first fiscal year when receipts were greater than expenditures, or when the 18th amendment (Prohibition) was repealed. The 18th amendment was in fact repealed December 5, 1933. In keeping with the law, the dividend tax was repealed and the gas tax reverted to 1 cent after December 31, 1933. The capital stock tax was only to apply to the year ending June 30, 1933, and the excess profits tax was not to apply to any taxable year after June 30, 1934. However, the Revenue Act of 1934, approved May 10, 1934, reimposed both the capital stock and excess profits taxes.

Revenue Act of 1934 Enacted May 10, 1934 The Revenue Act of 1934 included a modest tax increase and a number of tax reforms. According to the Ways and Means Committee report on the bill, “The primary purpose of the bill is to increase revenue by preventing tax avoidance” (“The Revenue Bill of 1934,” 73d Congress, 2d Session, House of Representatives Report No. 704, 2/12/34, p. 1). Thus, the bill had two motivations: reducing tax evasion and closing the budget deficit. The interest in reducing tax avoidance certainly continued a theme common in the 1920s. However, whereas the Coolidge administration emphasized reducing tax rates as the key to reducing evasion, the Roosevelt administration and the Democratic Congress emphasized closing loopholes and raising some rates. One sign of the concern about evasion was that “A subcommittee of your Committee on Ways and Means has been engaged in studying tax avoidance and the means of preventing such avoidance since last June” (House of Representatives Report No. 704, p. 1). Many of the changes included in the bill were described by the Committee report as preventing evasion. For example, it discussed “the most prevalent form of tax avoidance practiced by individuals with large incomes is the scheme [in which] an individual forms a corporation and exchanges for its stock his personal holdings in stock, bonds, or other income-producing property. By this means the income from the property pays corporation tax, but no surtax is paid by the individual if the income is not distributed” (p. 11). The Revenue Act of 1934 made such a scheme less attractive by imposing a 35 percent tax on the undistributed adjusted net income of such personal holding companies.

60

Roosevelt’s speeches in 1934 barely mentioned the proposed act. However, he expressed concern about tax evasion in other contexts. At a press conference in March 1934, Roosevelt discussed the Justice Department’s “proceedings against Mellon and others on tax evasion” (Excerpts from the Press Conference, 3/14/34, p. 3). Roosevelt stressed that “There are several hundred cases in exactly the same category” (p. 3). This concern about tax avoidance among the rich is a theme Roosevelt returned to in this discussion of other revenue actions in the mid- and late 1930s. The Ways and Means Committee expressed substantial concern about the deficit. The report discussed the “vital need of the Government for the additional revenue which should be produced by this bill” (House of Representatives Report No. 704, p. 2). It also said: “Your committee is of the opinion that it is of the utmost importance to reduce the deficits estimated for the fiscal years 1934 and 1935 as much as possible and to attain the goal of a balanced Budget in 1936” (p. 4). The Roosevelt administration appears to have been somewhat more sanguine about the deficit. In his Fireside Chat (Recovery Program), Roosevelt stressed the importance of “preserving and strengthening the credit of the United States Government,” which had been damaged by the fact that “For years the Government had not lived within its income” (7/24/33, p. 1). However, Roosevelt adopted a largely “cyclically-adjusted” view of the deficit. The same Fireside Chat said: “The immediate task was to bring our regular expenses within our revenues. That has been done” (p. 1). The 1934 Annual Budget Message defined roughly two-thirds of all federal expenditures as “emergency” (1/3/34, p. 1). Likewise, the 1935 Annual Budget Message spoke of “a Budget for the fiscal year 1936, which balances except for expenditures to give work to the unemployed” (1/3/35, p. 3). Thus, it seems unlikely that deficit reduction was a key motivation for the bill from the administration’s side. The revenue estimates for the Revenue Act of 1934 are moderate. The Ways and Means Committee report gave an estimate of $258 million (House of Representatives Report No. 704, p. 4). Of this number, roughly one-third was due to non-legislated changes in the administration of depreciation allowances urged by the Congress and adopted by the Treasury. Changes in the tax-rate structure accounted for $28 million of the additional revenues; changes in the treatment of capital gains accounted for another $35 million; and the new treatment of personal holding companies accounted for $25 million (p. 4). These numbers are somewhat larger than those given in other sources. For example, the 1934 Treasury Annual Report said changes in the rate structure and capital gains provisions of the Revenue Act of 1934 increased revenues by $26 million in fiscal year 1935 (p. 24). The President’s 1934 Annual Budget Message suggested a possible revenue impact of the proposed law of $150 million (1/13/34, p. 4). As the previous discussion has suggested, the Revenue Act of 1934 made many changes in the tax law. On rates, it changed the normal income tax from a two-tier tax (with rates of 4 and 8 per cent) to a single tax of 4 percent. It compensated for this reduction in the normal tax by raising surtax rates. The top surtax rate rose from 55 to 59 percent. Tax revenues were raised by lowering the exemption for dividends. The act partially reinstated the earned income credit, which had been eliminated by the Revenue Act of 1932. The act included a very large change in the treatment of capital gains. Under existing law, capital gains were taxed at a maximum rate of 12½ percent. The new law subjected capital gains to the sum of the normal and surtax rates (which could be as high as 63 percent). However, the fraction of the capital gain that was taxed varied with how long the asset had been held. The act also raised the estate tax substantially, and imposed a new tax on personal holding companies. Though not officially part of the act, an important administrative change urged by Congress and adopted by the Treasury raised corporate tax payments. The new depreciation guidelines reduced

61

substantially the amount of deductions that were allowed. They also greatly changed the burden of proof required by justifying such deductions. Social Security Act Enacted August 14, 1935 The Social Security Act instituted a wide range of social insurance programs. In addition to the well-known old-age assistance, it also included aid to dependent children, public health initiatives, and a program of federally supported unemployment compensation. To pay for these programs, new taxes were placed on both workers and employers. The motivation for the new spending programs was, in Roosevelt’s words, “to provide sound and adequate protection against the vicissitudes of modern life” (Fireside Chat, 6/28/34, p. 3). Congress described the act as “a more comprehensive and construction attack on insecurity,” and said “[t]he bill is designed to aid the States in taking care of the dependent members of their population, and to make a beginning in the development of measures which will reduce dependency in the future” (“The Social Security Bill,” 74th Congress, 1st Session, House of Representatives Report No. 615, 4/5/35, p. 3). The motivation for the tax increases was clearly to pay for the new spending. The clearest evidence for this is that the tax increases were legislated in the same bill setting up the spending programs. Roosevelt was also explicit that the new Social Security System should be self-financing. In his Message to Congress on Social Security, he said: “the system adopted, except for the money necessary to initiate it, should be self-sustaining in the sense that funds for the payment of insurance benefits should not come from the proceeds of general taxation” (1/17/35, p. 1). Likewise, Congress emphasized that the bill “seeks to reduce dependency and to encourage thrift and self-support” (House of Representatives Report No. 615, p. 16). The Congressional report also said, “Practically no objections have been made to the imposition of the taxes levied in this bill,” and went on to say that such a self-financing social insurance system could ultimately lower the burden of relief on all taxpayers (p. 16). Just shortly after passage of the Social Security Act, a similar measure was passed for workers in the railroad industry. This program focused almost exclusively on old-age assistance. The act, entitled the Carriers Taxing Act of 1935, was approved August 29, 1935. The program was obviously much smaller than the Social Security program, but was fundamentally similar in structure and motivation. In particular, the tax was designed to pay for the promised pension benefits. The act was replaced by the Carriers Taxing Act of 1937. The revenue effects of the Social Security taxes were modest because the tax rates were low and coverage was incomplete. The President’s Annual Budget Message on January 7, 1937 said that the first full year of tax collection under the Social Security Act would result in revenue of $774.8 million in fiscal year 1938 (p. 4). These numbers match fairly well with the retrospective numbers given in the 1940 Treasury Annual Report. The Annual Report said that Social Security revenues were $253 million in fiscal year 1937 (p. 15). It then said that employment taxes increased $502 million in fiscal year 1938, suggesting a total value of $755 million in the first full fiscal year (p. 16). Both of these revenue estimates are noticeably smaller than those given in the Ways and Means Committee report on the bill. That report said that Title VIII of the bill (the taxing provisions related to old-age assistance and social programs) would raise $560.2 million in fiscal year 1938 and Title IX (the taxing provisions related to unemployment compensation) would raise $501.0 million (House of Representatives Report No. 615, p. 15). In general, the report’s estimates for Title VIII are quite close to the retrospective numbers, but the Title IX numbers are dramatically higher.

62

Revenue estimates for the Carriers Taxing Act of 1937 are more difficult to find. The President’s 1938 Annual Budget Message said that it had been expected that the taxes on carriers and their employees would produce revenues of $134,552,000 in fiscal year 1937. However, litigation delayed collection, so only $345,088 was collected (1/3/38, p. 3). The Social Security Act imposed three taxes. One was a tax on individuals of 1 percent on wages up to $3000. The second was an equivalent tax on employers on wages paid (again up to $3000). Both of these tax rates were supposed to rise ½ percentage point every three years, reaching a maximum of 3 percent after December 31, 1948. However, the Social Security Amendments of 1939, approved August 10, 1939, prevented the first ½ percent increase that was to have applied in calendar years 1940, 1941, and 1942. The third tax was a tax on total wages paid by employers with 8 or more employees for unemployment compensation. The rate was 1 percent in 1936, 2 percent in 1937, and 3 percent in 1938 and after. The Carriers Taxing Act of 1935 put a tax on employees of 3½ percent of compensation received up to $300 per month. Like the Social Security Act, it put an equivalent tax on employers. These rates were lowered by the Carriers Taxing Act of 1937. The new rate was set at 2¾ percent on compensation up to $300 per month for both employees and employers. The rate was to increase ¼ of a percentage point every three years until it reached 3¾ percent after December 31, 1948. (This act also refunded all taxes paid under the 1937 act for compensation paid prior to January 1, 1937.) The income taxes under both the Social Security Act and the Carriers Taxing Act of 1937 took effect on January 1, 1937. The unemployment compensation tax on large employers included in the Social Security Act took effect January 1, 1936. Revenue Act of 1935 Enacted August 30, 1935 The Revenue Act of 1935 was a modest tax increase motivated largely by notions of fairness. Unlike earlier tax changes, budget balance paid a relatively minor role. According to the January 3, 1935 Annual Budget Message, the deficit at the end of fiscal year 1934 was nearly $4 billion. The Roosevelt administration clearly viewed economic conditions as a temporary emergency that warranted such large budget deficits. The Budget Message said:

[U]nemployment is still large. The States and local units now provide a smaller proportionate share of relief than a year ago and the Federal Government is therefore called upon to continue to aid in this necessary work.

For this reason it is evident that we have not yet reached a point at which a complete balance of the Budget can be obtained. I am, however, submitting to the Congress a Budget for the fiscal year 1936, which balances except for expenditures to give work to the unemployed (p. 3).

Such expenditures for relief and recovery were projected to be approximately $4 billion in fiscal year 1936. In his Annual Budget Message in January 1936, Roosevelt was even more pro-deficit. He said: “Our policy is succeeding. The figures prove it. Secure in the knowledge that steadily decreasing deficits will turn in time into steadily increasing surpluses, and that it is the deficit of today which is making possible the surplus of tomorrow, let us pursue the course that we have mapped” (1/3/36, p. 2). In his Message to Congress on Tax Revision, Roosevelt never mentioned raising revenue as a motivation for the proposed changes. In testimony to the House Ways and Means Committee, Secretary

63

of the Treasury Henry Morgenthau mentioned balancing the budget, but only in the vaguest terms (“Proposed Taxation of Individual and Corporate Incomes, Inheritances, and Gifts,” 74th Congress, 1st Session, Hearings before the Committee on Ways and Means, House of Representatives, July 8, 9, 10, 11, 12, 13, 1935, p. 5). He was pushed on this point by Representative Knutson, who said: “I would say that the Secretary’s position seems to be that this is a revenue measure, and the President’s position seems to be that this is a share-the-wealth program” (p. 12). Roosevelt’s Message to Congress on Tax Revision centered on the fairness argument for the proposed changes. He proposed increasing the tax on inheritances and gifts because: “The transmission from generation to generation of vast fortunes by will, inheritance, or gift is not consistent with the ideals and sentiments of the American people” (6/19/35, p. 1). He also proposed raising tax rates on higher incomes saying:

Social unrest and a deepening sense of unfairness are dangers to our national life which we must minimize by rigorous methods. People know that vast personal incomes come not only through the effort or ability or luck of those who receive them, but also because of the opportunities for advantage which Government itself contributes. Therefore, the duty rests upon the Government to restrict such incomes by very high taxes (p. 2).

Even the proposed changes in the corporate income tax centered on equity. Roosevelt said: “It

seems only equitable, therefore, to adjust our tax system in accordance with economic capacity, advantage and fact. The smaller corporations should not carry burdens beyond their powers; the vast concentrations of capital should be ready to carry burdens commensurate with their powers and their advantages” (p. 3). Secretary Morgenthau also emphasized the fairness motivation in his testimony to Congress. He said:

These proposed taxes rest on the principle of ability to pay. They are devised to draw an accumulation of wealth and income which, for the most part, have been derived from Nation-wide activities. In consequence, their enactment should constitute an important step forward in reshaping our tax structure along sounder and fairer lines (Hearings, July 8, 9, 10, 11, 12, 13, 1935, p. 5).

The Ways and Means Committee report on the bill made it clear that Congress was largely just following the President’s suggestions. It stated: “It is hoped that the enactment of these titles into law will remedy some of the defects of our present tax system, will provide substantial revenue, and will carry out the major policies recommended by the President in his message” (“The Revenue Bill of 1935,” 74th Congress, 1st Session, House of Representatives Report No. 1681, 7/30/35, p. 4). One sign that the Democratic majority may have agreed with the President’s motivation is that it revised the excess profits tax on corporations to be sharply progressive and higher than existing law on all but the lowest profit level. The report stated: “This tax was not specifically mentioned by the President, but your committee feels it is in line with his general policy to tax those with the ability to pay whether they be individuals or corporations” (p. 7). The “Views of the Minority” included in the report suggested that “the Democratic members of the committee were not only indifferent, but actually hostile to his [the President’s] proposals” (House of Representatives Report No. 1681, p. 18). In their view, “That the bill now comes before the House with their approval is further evidence of the fact that the majority are not guided by their convictions but by the orders they receive from the White House” (p. 18). The minority was also quite cynical about why the President proposed the tax changes just six months after saying he did not feel it advisable to raise taxes. They said: “The fact is that the President’s tax message came at a time when the administration’s popularity and prestige were rapidly on the decline, and it served to divert public attention from the

64

criticisms which were being leveled at the President and his policies. It doubtless had a secondary purpose of undermining the increasing political strength of the two chief exponents of the ‘share the wealth’ and ‘soak the rich’ philosophy by making a bid for the support of their large army of followers” (p. 17). The revenue estimates provided by Secretary Morgenthau for the initial proposed changes had a huge range. He said the revenue increase could range between $118 million and $901.5 million (Hearings, July 8, 9, 10, 11, 12, 13, 1935, p. 6). The 1935 Treasury Annual Report gave an estimate of $222 million for fiscal year 1937 (pp. 39-40). This number is also given in the President’s Annual Budget Message (1/3/36, p. 1), which suggests it may be somewhat low because “collections in the fiscal year 1937 from income taxes and the estate tax only partially reflect the Revenue Act of 1935” (p. 6). The Ways and Means Committee report gave an estimate for the revenue increase of $270 million in “a full year of operation, under present improving business conditions.” Of this total, $45 million was from the increased surtaxes, $15 million was from the graduated corporation tax, $100 million was from the increase in the excess-profits tax, and $110 million was from the inheritance and gift taxes (House of Representatives Report No. 1681, p. 4). The Revenue Act of 1935 contained a number of provisions. It raised the surtax on incomes over $50,000. While surtax rates rose just a point or two at incomes around $50,000, the surtax rate at $100,000 rose 6 percentage points, and that on incomes over $1 million rose between 14 and 16 percentage points. The act also raised the estate tax substantially. The specific exemption was lowered from $50,000 to $40,000, and rates were doubled at low levels. The top estate tax was raised from 60 to 70 percent.

The act also made numerous changes to the corporate income tax. For example, it replaced the existing flat corporate income tax of 13¾ percent with a graduated tax of 12½ to 15 percent, based on net income. It also included an increase in the excess profits tax (re-introduced by the National Industrial Recovery Act of 1933). The estate tax provisions were in effect beginning on August 30, 1935. The personal and corporate income tax provisions were scheduled to take effect on January 1, 1936. However, they were superseded by the Revenue Act of 1936, which set the same tax rates (but made some other changes as well). As a result, not personal or corporate income tax returns were actually filed under the Revenue Act of 1935. Revenue Act of 1936 Enacted June 22, 1936 As discussed in the description of the Revenue Act of 1935, the Revenue Act of 1936 was a closely related measure. It made no changes to the surtax rates or estate tax rates included in the Revenue Act of 1935. But, because of its timing, it is the tax bill under which returns were first filed using those higher rates. The main tax changes included in the Revenue Act of 1936 affected the corporate income tax. It lowered the graduated normal corporate income tax slightly, and then added on a scale of surtaxes on undistributed profits, graduated from 7 to 27 percent. The proximate motivation for the bill appears to have been a desire to restore progress toward budget balance following two events. One was a Supreme Court decision invalidating taxes collected under the Agricultural Adjustment Act and the Bituminous Coal Act. The other was the passage of the Adjusted Compensation Payment Act (the Bonus Bill), which greatly accelerated a promised bonus payment to veterans. According to the President’s Supplemental Budget Message to Congress, these two

65

events threatened the “clear-cut and sound policy” of making progress toward reducing the deficit. He therefore declared: “it is incumbent upon us to make good to the Federal Treasury both the loss of revenue caused by the Supreme Court decision and the increase in expenses caused by the Adjusted Compensation Payment Act” (3/3/36, p. 1). The President estimated that $620 million per year of new permanent taxes were necessary; “five hundred million dollars of this amount represents substitute taxes in place of the old processing taxes, and that only one hundred and twenty million dollars represents new taxes not hitherto levied” (p. 1). The President’s message also indicated that another $517 million of additional revenues was needed for temporary purposes, but these revenues could be spread over two or three years (p. 2). The motivation of replacing lost revenue and covering the amortized cost of the veterans’ bonus was reiterated in the President’s Annual Budget Message in January 1937 (1/7/37, p. 3). While budget balance was the proximate motivation for the act, notions of fairness clearly played a key role in determining the form the tax increase took. In his March 3rd message, Roosevelt described an undistributed profits tax as “a form of tax which would accomplish an important tax reform, remove two major inequalities in our tax system, and stop ‘leaks’ in present surtaxes” (3/3/36, p. 1). He also said: “the aim, as a matter of fundamental equity, should be to seek equality of tax burden on all corporate income whether distributed or withheld from the beneficial owners” (p. 1). In a campaign appearance following passage of the bill, he said: “the undistributed profits tax, is merely an extension of the individual income tax law and a plugging-up of the loopholes in it, loopholes which could be used only by men of very large incomes” (Address at Worcester, Mass., 10/21/36, p. 3). In another campaign speech, Roosevelt said the tax was part of the fight against monopolies because it made it “harder for big corporations to retain the huge undistributed profits with which they gobble up small business” (Address at Boston, Mass, 10/21/36, p. 2). As with other tax actions in the mid-1930s, Congress largely parroted the President’s motivation. The Ways and Means Committee report on the bill said: “The need for such a bill was called to the attention of the Congress by the President” (“The Revenue Bill of 1936,” 74th Congress, 2d Session, House of Representatives Report No. 2475, 4/21/36, p. 1). After reprinting the President’s message, the report said: “The President requests the Congress to raise 620 million dollars of additional revenue annually … [and] suggests some form of undistributed profits tax. Your committee recognizes the fact that the greatest defect in our present system of taxation lies in the fact that surtaxes on individuals are avoided by impounding income in corporate surpluses” (p. 3). It then went on to detail the major purposes of the undistributed profits tax, citing, in addition to preventing tax avoidance, the removal of inequities between different forms of business organization and between large and small shareholders (p. 3). The estimates of the expected revenue effects are fairly consistent. The Ways and Means Committee report said the undistributed profits tax would “produce an average of at least 620 million dollars in additional revenue annually” (House of Representatives Report No. 2475, p. 4). This was despite the fact that the plan eliminated some other taxes affecting corporations. Indeed, one way that Congress raised an additional $173 million of temporary revenues was to continue the capital stock tax at one-half the previous rate for one last year. The President’s Statement on the Summation of the 1937 Budget said that it would produce annual revenue of $652 million at current business conditions (9/2/36, p. 1). As described above, the key tax changes brought about by the act were those affecting corporations. Under previous law, the corporate income tax rate was slightly graduated, from 12½ percent to 15 percent. There was an excess profits tax of 6 percent of net income in excess of 10 percent of the declared value of the capital stock and 12 percent of net income in excess of 15 percent of declared value. The new law eliminated the excess profits tax. It set a normal tax on corporate net income graduated from 8 to 15 percent. It then imposed a tax on undistributed net income graduated from 7 to 27

66

percent. The new corporate tax took effect after December 31, 1935. As a result, it superseded the changes in the corporate tax included in the Revenue Act of 1935 before any taxes were paid under that law. The Revenue Act of 1936 included a one-time tax on unjust enrichment. The idea was that some sales had occurred under the assumption that the excise taxes declared unconstitutional would be continued. As a result, firms may have already passed along the tax, and so received a windfall from the Supreme Court’s decision (see 1936 Treasury Annual Report, pp. 29-32, for a description of the acts provisions). The Revenue Act of 1936 had no effect on individual normal or surtax rates; it merely recodified the rates contained in the Revenue Act of 1935. The most notable change it made in the individual income tax was to subject dividends to the normal tax. Revenue Act of 1937 Enacted August 26, 1937 The Revenue Act of 1937 was a fairly minor action taken to prevent certain methods of tax avoidance and evasion. It had no effect on tax rates. The impetus for the act came from the President and the Department of the Treasury. On June 1, 1937, Roosevelt sent a message to Congress, the centerpiece of which was a letter from Treasury Secretary Henry Morgenthau. Morgenthau stated that lower-than-expected revenues led the Treasury to investigate individual income tax returns. The preliminary investigation found several devices being used by high-income people to avoid taxes. Among them were personal holding companies being set up in places such as the Bahamas and Panama; domestic personal holding companies; incorporating yachts and country estates; and creation of multiple trusts for relatives and dependents. Roosevelt concluded that: “it seems to me that the first duty of the Congress is to empower the Government to stop these evil practices, and that legislation to this end should not be confused with legislation to revise tax schedules” (Message to Congress on Tax Evasion Prevention, 6/1/37, p. 6). Congress embraced the President’s call to action. According to the Ways and Means Committee report on the bill, the special Joint Committee on Tax Evasion and Avoidance was formed on June 11, 1937 (“The Revenue Bill of 1937,” 75th Congress, 1st Session, House of Representatives Report No. 1546, 8/13/37, p. 1). The joint committee concluded that “legislation should be enacted in regard to the following subjects, with respect to which it has been shown that certain serious loopholes exist” (p. 2). It also urged that “legislation along the lines recommended be enacted at the earliest possible moment in order to protect the revenue, and in order that all may bear their fair share of the tax burden” (quoted in House of Representatives Report No. 1546, p. 2). The Ways and Means Committee concurred with the joint committee’s recommendations (p. 2). No sources give estimates of the possible revenue effects of the action. The 1937 Treasury Annual Report stated: “the preventative tax evasion and avoidance provisions of the Revenue Act of 1937 will tend to prevent revenue losses which might otherwise occur” (p. 30). Secretary Morgenthau, in a letter quoted by Roosevelt, said: “if tax evasion and tax avoidance can be promptly stopped through legislation and regulations resulting from a special investigation a very large portion of the deficiency in revenues will be restored to the Treasury” (Message to Congress on Tax Evasion Prevention, 6/1/37, p. 1). In a Message to Congress on April 20, 1937, the President said that “income taxes will produce $267,200,000 less than the former [January] estimate for the fiscal year 1937” (Message to Congress on Appropriations for Work Relief for 1938, 4/20/37, p. 1). Thus, even if a large portion of this underestimate were due to the loopholes addressed by the act, the revenue effects would be fairly small.

67

As discussed above, the Revenue Act of 1937 had no effect on individual income tax rates, including no effect on capital gains taxes or the estate and gift taxes. One of the key changes included in the act was to raise the surtax on undistributed adjusted net incomes of personal holding companies. Under previous law, the rate ranged from 8 to 48 percent, depending on net income. The Revenue Act of 1937 raised the rate to 65 percent on the first $2000, and 75 percent on all income over $2000. The act also changed the treatment of trusts, and the ability to incorporate yachts and homes. The changes were effective after 12/31/36. Revenue Act of 1938 Enacted May 28, 1938 The Revenue Act of 1938 was an extensive revision of the corporate income tax and the tax on capital gains. It was designed to be roughly revenue-neutral. The act became law without the President’s approval, suggesting that Congress’s motivation is the key one to consider. The Ways and Means Committee report on the bill said: “The purpose of the bill, as reported, is to improve our existing revenue system, to remove inequities, to equalize the tax burden, and to stimulate business activities, and to accomplish this without reducing the revenue which would be obtained by existing law under present conditions” (“The Revenue Bill of 1938,” 75th Congress, 3d Session, House of Representatives Report No. 1860, 3/1/38, p. 2). Of these motivations, the last is the most unusual. The macroeconomic effects of tax actions were rarely mentioned in the interwar era. Perhaps as recognition of the fact that the economy was in the midst of a severe recession, the committee singled out these possible effects saying: “Finally, and most important, it is believed that there will be very substantial stimulation to business by the enactment of the bill into law which will bring into being a well-balanced tax system, improved with respect to certainty and equity” (House of Representatives Report No. 1860, p. 2). The further motivation for revising the corporate tax was the sense that the undistributed profits tax imposed by the Revenue Act of 1936 had caused “a substantial number of cases of hardship” (House of Representatives Report No. 1860, p . 3). Among the perceived problems with the tax was that it discouraged business expansion and thus hurt employment, it was particularly hard on small, financially weak corporations, and that it penalized corporations which found it necessary to use current earnings to pay debts (pp. 3-4). The committee felt that “the principle of the undistributed-profits tax is sound and should be retained. However, it is believed that it should be substantially modified” (p. 4). President Roosevelt agreed that some modification of the law to help small corporations might be useful, but he objected to the other changes, such as a relatively flat corporate tax, that Congress was considering (Letter on the Tax Bill, 4/13/38, p. 2). Another motivation for the bill had to do with the capital gains tax. The Ways and Means Committee report said of this tax: “It is claimed that the present tax is so high, especially in the case of taxpayers subject to high surtax rates, that assets become frozen and few transactions tax place” (House of Representatives Report No. 1860, p. 7). At the same time, the Committee did not want to do anything that would encourage or benefit short-term speculation (p. 7). The President expressed a similar concern in his Message to Congress Recommending Legislation in November 1937. He said: “Nor should we extend tax privileges to speculative profits on capital where the intent of the original risk was speculation rather than the actual development of productive enterprise” (11/5/37, p. 2). More generally, Roosevelt seemed less sympathetic to the claims of hardship from the existing capital gains tax and emphasized that “capital gains should be taxed at progressive rates” (Letter on the Tax Bill, 4/13/38, p. 2).

68

Congress, in the end, did not produce a bill that allayed Roosevelt’s concerns about equity. As a result, Roosevelt took the somewhat unusual step of allowing the bill to become law without his signature. In his Address at Arthurdale, West Virginia, he said: “By taking this course, I am calling the definite attention of the American people to those unwise parts of the bill that I have been talking to you about today—one of them which may restore in the future certain forms of tax avoidance of the past, and of continued concentrated investment power, which we in Washington had begun to end; and the other feature, a definite abandonment of a principle of tax policy long ago accepted as part of our American system” (5/27/38, p. 5). This principle was that taxes should be set “in proportion to ability to pay” (p. 4). The Revenue Act of 1938 was fundamentally a reform measure not a revenue raising measure. The committee report stated: “According to the best information the committee has been able to secure, from the Treasury Department and other sources, it appears reasonably certain that the revenues of the Government will be as great under the bill as under existing law” (House of Representatives Report No. 1860, p. 2). This is consistent with the 1938 Treasury Annual Report which stated: “The effect of these statutory changes on income tax liabilities for relatively low income years such as calendar years 1938 and 1939 is decidedly less important in determining the income tax receipts than are the changes in the business situation” (p. 35). It is also consistent with the President’s charge in November 1937 that “Nor can we at this time accept a revision of our revenue laws which involves a reduction in the aggregate revenues” (Message to Congress Recommending Legislation, 11/5/37, p. 2). The Revenue Act of 1938 included fundamental changes in the corporate income tax. Under the Revenue Act of 1936, all corporations were subject to a normal tax graduated from 8 to 15 percent, and a surtax on undistributed profits graduated from 7 to 27 percent. The Revenue Act of 1938 removed the undistributed profits tax on firms with net incomes of less than $25,000. Such small firms paid a tax graduated from 12½ to 16 percent (1938 Treasury Annual Report, p. 28). According to the Ways and Means Committee report, this change exempted 88 percent of corporations from the undistributed profits tax (House of Representatives Report No. 1860, p. 4). For firms earning more than $25,000, a tentative tax of 19 percent was imposed on adjusted net income. To retain the spirit of the undistributed profits tax, the tentative tax “is reduced by the sum of (a) 16½ percent of the credit for dividends received and (b) 2½ percent of the dividends paid credit, but not to exceed 2½ percent of the adjusted net income” (Treasury Annual Report, pp. 28-29). The Revenue Act of 1938 also effected substantial changes in the capital gains tax. Under the Revenue Act of 1936, individuals paid the sum of their normal and surtax rates on capital gains. The fraction of capital gains included in the calculation of the tax varied according to how long the asset had been held. The new law divided capital gains into three categories depending on how long the asset had been held. Short-term capital gains were assets held for less than 18 months. All of these short-term gains were included as income and taxed at the applicable normal and surtax rates. Long-term capital gains were divided into two categories, those for assets held between 18 and 24 months, and those for assets held for more than two years. For the first of these two groups, two-thirds of the gain was counted as income; for the second, one-half was counted. Long-term capital gains were taxed at the lower of the taxpayer’s normal plus surtax rate and 30 percent. The changes in both the capitals gains tax and the corporate income tax took effect after December 31, 1937. Revenue Act of 1939 Enacted June 29, 1939 The Revenue Act of 1939 was a revenue-neutral tax action that extended a number of excise taxes and revised the corporate income tax. The corporate tax change was designed to continue the move toward simplification begun in the Revenue Act of 1938.

69

The extension of the excise taxes was straightforward. Roosevelt had recommended it in his 1939 Annual Budget Message. He said: “I am recommending the reenactment of the excise taxes which will expire in June and July of this year, not because I regard them as ideal components of our tax structure, but because their collection has been perfected, our economy is adjusted to them, and we cannot afford at this time to sacrifice the revenue they represent” (1/5/39, p. 3). The Ways and Means Committee report on the bill concurred with this assessment, saying: “If the temporary taxes … are permitted to lapse at this time, a loss in revenue of over $600,000,000 will occur” (“The Revenue Bill of 1939,” 76th Congress, 1st Session, House of Representatives Report No. 855, 6/16/39, p. 14). It then parroted the President’s statement that the government could not afford the loss of revenue (p. 15). The revision of the corporate income tax was motivated by a desire to make the tax calculation less burdensome, and, by doing so, to stimulate economic activity. The Ways and Means Committee report stated that the first objective of the bill “is to remove from the existing corporate income-tax structure such business deterrents and tax irritants as may be possible to consider at this time” (House of Representatives Report No. 855, p. 1). It said that following passage of the Revenue Act of 1938, “further inequities have become evident and attention has been drawn to a number of instances where our existing tax laws act as a deterrent to a free flow of business activity. In addition, our existing law contains certain tax ‘irritants’ which are relatively unimportant from the point of view of revenue but are burdensome and irritating to taxpayers” (p. 2). Chief among these perceived irritants was the undistributed profits tax component of the corporate income tax, which the committee suggested had “acquired prominence as a psychological irritant largely because of the widespread emotional criticism which has been directed against it” (p. 8). This motivation appears to have been at least partially suggested by the administration. The committee report stated that at a hearing on May 27, “the Secretary of the Treasury appeared and stressed the desirability of making certain changes in our corporate income-tax structure, which, it was thought, would encourage business activity and a freer flow of capital into productive enterprise” (House of Representatives Report No. 855, p. 2). At the same time, the President seemed somewhat cool to the changes. In his Address Before the American Retail Federation in May 1939, he emphasized that “especially in view of the unbalanced budget, … we ought not to raise less money from taxation than we are doing now,” and he suggested that “it would be bad for business, to shift any further burden to consumer taxes” (5/22/39, p. 4). Therefore, he argued that any change to reduce deterrent taxes on corporations needed to be replaced by other taxes on corporations. He was somewhat dismissive of what he referred to as the “great hullabaloo for the repeal of the undistributed earnings tax,” which he said raised only $20 million of revenue (p. 4). He said that he was willing to have this tax repealed subject to two conditions: that the same revenue be raised through another tax on corporations earning more than $25,000 a year and that the loophole the undistributed profits tax sought to close did not re-emerge (pp. 4-5). By its nature, the extension of the excise taxes, which had been first introduced in 1932 and had already been renewed three times, did not raise revenues relative to the previous year. Likewise, all of the discussion suggests that the corporate tax revision was designed to keep revenues the same. For example, the Ways and Means Committee report said: “While one purpose of the bill as reported is to stimulate business activity, the committee has sought to accomplish this without endangering the productivity of the existing tax structure. According to the best information the committee has been able to secure, it appears reasonably certain that the revenues of the Government will not be reduced appreciably under the present bill” (House of Representatives Report No. 855, p. 3). The key change that the Revenue Act of 1939 brought about was a change in the income tax rate on corporations earning more than $25,000. Under the Revenue Act of 1938, such firms paid a rate ranging from 16½ to 19 percent, depending on their distribution of dividends. The differential between

70

the rate paid and 16½ percent is what was referred to as the continuation of the undistributed profits tax. This graduated rate was replaced by a flat rate of 18 percent. The income tax rates on small corporations were unchanged. The Revenue Act of 1939 made many other small administrative changes designed to be more taxpayer-friendly. For example, one method of calculating the normal profit involved the declared capital stock used as the base for the capital stock tax. Firms were given the option of increasing the declared valuation in both 1939 and 1940. The Revenue Act of 1939 was to take effect after December 31, 1939. However, according to the 1941 Statistics of Income, “The rates of tax provided by the Revenue Act of 1939 were never in effect, being superseded by those of the Revenue Acts of 1940” (Part 2, p. 315, fn. 31). Revenue Act of 1940 Enacted June 25, 1940 The Revenue Act of 1940 was a widespread tax increase motivated by an increase in defense spending and concern about the related deficit. It included a mixture of temporary and permanent tax changes and affected almost all existing tax rates. The proximate cause for the tax increase was the increase in defense spending necessitated by the deteriorating international situation. Roosevelt detailed the need for more defense spending in both his January 3, 1940 Annual Message to the Congress, and, more forcefully, in his May 16, 1940 Message to Congress on Appropriations for National Defense. The May message emphasized the rapidity of modern warfare and said: “The clear fact is that the American people must recast their thinking about national protection” (p. 1). Roosevelt asked for nearly $1 billion in extra appropriations for national defense. The President asked for new taxes to pay for the increased expenditures because he felt economic conditions did not warrant a large increase in the deficit. In his January 3, 1940 Annual Budget Message, Roosevelt defended the large budget deficits of the mid-1930s on strikingly Keynesian grounds. He said: “The deliberate use of Government funds and of Government credit to energize private enterprise—to put purchasing power in the hands of those who urgently needed it and to create a demand for the products of factory and farm—had a profound effect both on Government and on private incomes” (p. 1). Likewise, the President defended allowing the budget to deteriorate again during the 1938 recession, saying: “The experience of 1938-1939 should remove any doubt as to the effectiveness of a fiscal policy related to economic need” (p. 1). However, Roosevelt felt that by 1940 the situation had changed substantially. He said: “we are achieving the highest levels of production and consumption in our history,” though he emphasized that unemployment was still quite high (Annual Budget Message, 1/3/40, p. 1). The President concluded that:

Against this background of aims substantially but not fully attained, I propose in the field of fiscal policy that we adopt the following course: We should count upon a natural increase in receipts from current taxes and a decrease in emergency expenditures, and we should try to offset the unavoidable increase in expenditures for national defense by special tax receipts, and thus hope to secure, for the over-all picture, a gradual tapering off, rather than an abrupt cessation, of the deficit (pp. 1-2).

He reiterated this stance in his Annual Message to the Congress the same day, saying: “Therefore, in the hope that we can continue in these days of increasing economic prosperity to reduce the Federal deficit, I am asking the Congress to levy sufficient additional taxes to meet the emergency spending for national defense” (1/3/40, p. 4).

71

In addition to his view of the desirable path for the deficit, Roosevelt also stressed the notion that temporary defense expenditures should be covered by dedicated taxes. In his Annual Budget Message, he said: “I believe that it is the general sense of the country that this type of emergency expenditure be met by a special tax or taxes. Moreover, this course will make for greater assurance that such expenditures will cease when the emergency has passed” (1/3/40, p. 3). The Ways and Means Committee report stressed similar motivations for the bill. It began by stating that: “Recent developments in the European War have reminded us forcefully of the inadequacy of our means of defense against modern weapons of aggression” (“The Revenue Bill of 1940,” 76th Congress, 3d Session, House of Representatives Report No. 2491, 6/10/40, p. 1). The committee fully supported the President’s program for increased defense spending. It then discussed the fiscal situation and said that the increased defense expenditures would result in a large deficit without the proposed tax increase (p. 2). The committee did not invoke views about the appropriate size of the deficit, but instead acted as if it was obvious that keeping the deficit from ballooning was desirable. The concrete reason it gave for the tax increase and increased authorization for borrowing was that it “will give the Treasury that flexibility in its financing which [is] so necessary for the effective management of the public finances in times like these and thus obviate the payment of higher interest rates” (p. 2). The President and Congress both expressed views on the desirable nature of the tax changes. The President said that he hoped “the Congress will follow the accepted principle of good taxation of taxing according to ability to pay and will avoid taxes which decrease consumer buying power” (Annual Budget Message, 1/3/40, p. 3). At a Press Conference in late May, he stressed the desirability of simply increasing all taxes by 10 percent (5/28/40, p. 4). Congress stressed its desire to “enable a larger proportion of our citizens to participate in the responsibility of providing an adequate national defense” (House of Representatives Report No. 2491, 6/10/40, p. 3). It was also concerned about the creation of “war millionaires,” but postponed until later in the year the creation of a new excess profits tax in order to reach closure on the current bill (p. 3). The House report on the bill said that it would yield additional annual revenues of $1,004 million (House of Representatives Report No. 2491, p. 2). Of these additional revenues, $322 million came from permanent tax increases and $682 million from temporary taxes expected to last for only five years (p. 2). The nature of the tax changes was both extensive and highly varied. For individuals, the Revenue Act of 1940 permanently lowered the personal exemption by 20 percent (from $2500 to $2000 for married couples). It also raised surtax rates on net incomes between $6000 and $100,000. At some levels the increase in rates was quite dramatic: for example, the marginal surtax rate on net incomes of $50,000 rose from 27 percent to 40 percent. The rates on incomes above $100,000 remained at their already very high marginal rates: the top marginal rate (on incomes greater than $5,000,000) was 75 percent. On top of these permanent tax increases, the act imposed a temporary “defense tax” equal to 10 percent of essentially all regular taxes. Thus, the effective normal and surtax rates were 10 percent higher than the stated rates. The Revenue Act of 1940 permanently raised the normal corporate tax by 1 percentage point at each level of income. As with the individual income tax, these rates were then raised by 10 percent. Thus, the effective top rate rose to 22 percent (from 19 to 20 percent, plus 10 percent more). The act was effective after December 31, 1939. However, no corporate tax was filed under this act because later acts superseded it for the 1940 tax year.

72

Second Revenue Act of 1940 Enacted October 8, 1940 The Second Revenue Act of 1940 was a tax increase aimed primarily toward corporations. Its key feature was the introduction of a new excess profits tax. Like the Revenue Act of 1940, the Second Revenue Act of 1940 was passed in the context of rapid increases in defense expenditures. However, the deficit and the need for revenues received little attention in the discussion. Rather, concern about fairness played a central role. In his short Message to Congress on a Steeply Graduated Excess Profits Tax, Roosevelt said: “We are asking even our humblest citizens to contribute their mite. It is our duty to see that the burden is equitably distributed according to ability to pay so that a few do not gain from the sacrifices of the many. I, therefore, recommend to the Congress the enactment of a steeply graduated excess profits tax” (7/1/40, p. 1). Likewise, at a Press Conference on August 27, 1940, the President stated: “it is up to the Congress to pass whatever excess profits tax Congress thinks should be put on, in order to prevent the creation of another crop of American millionaires” (p. 4). Congress had expressed similar sentiments in its discussion of the Revenue Act of 1940, but had postponed creation of the excess profits tax. The Ways and Means Committee report on the Second Revenue Bill of 1940 summarized this discussion saying: “your committee expressed the desire that the rearmament program should furnish no opportunity for the creation of new war millionaires or the further substantial enrichment of already wealthy persons” (“Second Revenue Bill of 1940,” 76th Congress, 3d Session, House of Representatives Report No. 2894, 8/28/40, pp. 1-2). The report went on to say that “Your committee is still of this opinion,” but that it felt some incentives were necessary “to stimulate the cooperation of private enterprise in the defense program” (p. 2). For this reason, Congress coupled a steeply graduated excess profits tax with a suspension of the profit limitations of the Vinson-Trammel Act covering construction of military ships and aircraft. The President had asked for an excess profits tax “to be applied to all individuals and all corporate organizations without discrimination” (Message to Congress on a Steeply Graduated Excess Profits Tax, 7/1/40, p. 1). However, the Ways and Means Committee decided that individual and partnership income were already subject to heavy surtaxes (House of Representatives Report No. 2894, p. 2). For this reason, it limited the tax to corporations. The committee report estimated that the excess profits tax would yield $305 million in calendar year 1940 (p. 3). The report also estimated that the revenue would be over $700 million per year once the defense program was fully operative (p. 3). The Second Revenue Act of 1940 made no changes to the individual income tax. One provision was to permanently raise the normal income tax rate on corporations with incomes greater than $25,000. The rate was raised to 22.1 percent. The temporary defense tax was retained as an additional 10 percent of the tax called for by the Revenue Act of 1940 (which was 19 percent). Thus, the total tax on corporations at this income level was 24 percent. The Second Revenue Act of 1940 added a new excess profits tax, in addition to the declared-value excess profits tax which had been in effect since June 30, 1933. The tax was collected on net income over the specific exemption of $5000, plus a credit. The credit could be calculated in either of two ways: 95 percent of average base period net income (where the base was the five years 1936-1940), or 8 percent of invested capital. The rates rose from 25 percent on the first $20,000 of excess profits to 50 percent on excess profits over $500,000. The act took effect after December 31, 1939.

73

The excess profits tax component of the Second Revenue Act of 1940 was modified slightly by the Excess Profits Tax Amendments of 1941, enacted March 7, 1941. The amendments were designed to “deal with the effect of certain abnormal situations upon the excess profits tax liability of corporations” (1941 Treasury Annual Report, p. 59). For example, the method of computing the base period net income was adjusted to aid companies whose earnings in the second half of the base period were higher than in the first half. The amendments were made retroactive to tax years after December 31, 1939. Revenue Act of 1941 Enacted September 20, 1941 The Revenue Act of 1941 was an enormous tax increase affecting both individuals and corporations. While the act was fundamentally driven by the increase in defense expenditures related to developments in Europe, macroeconomic conditions and notions of fairness also played a role. In his 1941 Annual Budget Message, Roosevelt outlined the tremendous increase in defense expenditures and the resulting rise in the budget deficit (1/3/41, pp. 2, 4). He then went on to say: “There is no agreement on how much of such an extraordinary defense program should be financed on a pay-as-you-go basis and how much by borrowing” (p. 4). The President, however, clearly felt that conditions warranted a substantial move toward higher taxes. He said: “We cannot yet conceive the complete measure of extraordinary taxes which are necessary to pay off the cost of emergency defense and to aid in avoiding inflationary price rises which may occur when full capacity is approached. However, a start should be made this year to meet a larger percentage of defense payments from current tax receipts” (p. 4). Roosevelt was clearly walking a fine line in his proposal. He noted that “Economic activities and national income are rising to record heights” (p. 3). But, at the same time, he was “opposed to a tax policy which restricts general consumption as long as unused capacity is available and as long as idle labor can be employed” (p. 4). His goal was a “policy aimed at collecting progressive taxes out of a higher level of national income” (p. 4). As with so many Roosevelt-era tax changes, concern about equity was a substantial concern. The Budget Message said: “The additional tax measures should be based on the principle of ability to pay,” and “it is the fixed policy of the Government that no citizen should make any abnormal net profit out of national defense” (1/3/41, p. 4). Likewise, the President’s Annual Message to Congress on January 6, 1941 said: “I shall recommend that a greater portion of this great defense program be paid for from taxation than we are paying today. No person should try, or be allowed, to get rich out of this program; and the principle of tax payments in accordance with ability to pay should be constantly before our eyes to guide our legislation” (p. 5). In May 1941, when the President recommended that “three and one-half billion of additional taxes should be levied during the coming year to defray in part the extraordinary defense expenditures,” he again emphasized fairness concerns. He wanted a law “so devised that every individual and every corporation will bear its fair share of the tax burden,” and one “which will convince the country that a national defense program intended to protect our democracy is not going to make the rich richer and the poor poorer” (both quotations from Recommendation for Additional Taxes, 5/1/41, p. 1). The Ways and Means Committee report on the bill made it clear that rising defense expenditures were the fundamental motivation for the tax increase. It stated: “The bill is unprecedented in the amount of revenue it is designed to provide. It lays a substantially increased burden upon the American people. But there is convincing evidence that this burden will be borne cheerfully in the light of the overwhelming importance of national defense” (“The Revenue Bill of 1941,” 77th Congress, 1st Session, House of Representatives Report No. 1040, 7/24/41, p. 2). Like the President, Congress said it aimed “to distribute the additional tax burden as equitably as possible” (p. 2). Congress, however, gave more

74

prominence to macroeconomic conditions. The report talked of the tax increase “supplying a needed restraint upon inflationary tendencies” (p. 2). This more direct focus on inflation could reflect differences in timing of the statements on the tax increase between the President and Congress. In late July 1941, the President also became very concerned about inflation and proposed price controls (see, Message to Congress on Price Control Legislation, 7/30/41, p. 1). The President asked for a tax increase of $3.5 billion (Recommendation for Additional Taxes, 5/1/41, p.1). The Ways and Means Committee report said that the proposed bill would yield at least that amount in a full year of operation (House of Representatives Report No. 1040, p. 2). The Revenue Act of 1941 increased a wide variety of taxes. For individuals, the act raised surtax rates and integrated the temporary 10 percent defense tax into the permanent tax structure. A key change was that the surtax, which used to begin at a net income of $4,000 with a rate of 4 percent, now started at a net income of zero with a rate of 6 percent. Surtax rates, in general, increased dramatically. The rate at $10,000 increased from 10 percent to 25 percent. The top marginal rate (at a net income of $5 million) rose from 75 percent to 77 percent. The personal exemption was reduced (from $2000 to $1500 for a couple) and the $400 credit for the first dependent was eliminated for the head of family. The act left the normal tax on corporate income largely unchanged: the rate on corporations earning more than $25,000 remained at 24 percent. However, the act added a surtax of 6 percent on the first $25,000 of surtax net income, and 7 percent on income above $25,000. The act raised the excess profits tax rates from a range of 25 to 50 percent to 35 to 60 percent. The act also tightened up the “invested capital” method for determining the base from which excess profits were calculated. Instead of the base being calculated as 8 percent of total invested capital, it was reduced to 7 percent of invested capital in excess of $5 million.

75

APPENDIX 2 TECHNICAL DETAILS OF COMPUTATION OF CHANGES IN ORDINARY TAXABLE

INCOME AND CHANGES IN MARGINAL RATES

This appendix describes how we construct data for taxable income exclusive of capital gains and losses and for marginal rates on non-capital-gains income from the figures in the Statistics of Income.

A. Overview Obtaining the income data we need for taxpayers in a given range of net income involves two

steps. First, and most important, we need to remove capital gains and losses. Second, because we do not want to include income changes that resulted from changes in how taxable income was defined, we need to correct for changes in the definition of taxable income. Throughout our sample period, taxable non-capital-gains income was very similar to net income excluding capital gains and losses, and the definition of the non-capital-gains components of net income did not change. We therefore use net income excluding capital gains and losses—which we refer to as “ordinary taxable income”—as our income measure throughout.

To obtain the marginal rate on non-capital gains income faced by households in a given range of

net income, we need to exclude any portion of their net income that was either untaxed or taxed separately. For example, in most of the 1920s capital gains income was taxed at a separate rate, and in much of the 1930s a portion of capital gains income was excluded from taxable income. We can then find the marginal rate that applied to the relevant level of taxable income. The personal income tax in the interwar era had two components: a “normal” tax and a “surtax.” Normal tax rates were low, typically on the order of 4 percent, relatively stable, and only slightly graduated. Surtax rates, in contrast, were often very high, volatile, and extremely progressive. In all of our analysis, we look at the combined effects of the two components to measure marginal rates. In cases where legislation changed the tax code retroactively, we use the definition of taxable income and tax rates that were in effect during the year, not the rates that were applied ex post.

As described in the text, a key input into our analysis is the change in marginal rates that was the

result of policy (rather than of economic developments changing households’ incomes, and so moving them into different tax brackets). To find the policy-induced change in marginal rates in year t, we compute marginal rates on year t – 1 income using the definition of taxable income and the tax rates that were in effect in year t and compare them with the marginal rates implied by the definition of taxable income and the tax rates that were in effect during year t – 1.

The income categories in the Statistics of Income do not correspond exactly to the groups we want to use in our statistical work. For example, our top percentile group in some year might include the filers in all income categories over $200,000 plus a certain number of filers in the $150,000–$200,000 category. To estimate the total ordinary taxable income of the top percentile group, we would therefore need to estimate the division of the taxable income in the $150,000–$200,000 group between the taxpayers who are in the top percentile group and those who are not. Similarly, everyone in the $90,000–$100,000 range in some year might have been in our second percentile group, but the marginal rate might have changed within this range. To estimate the average marginal rate faced by the second percentile group, we would therefore need to estimate the fraction of the overall income of the $90,000–$100,000 that was taxed at each relevant marginal rate.

The Statistics of Income report the number of households in each range of net income. We follow

76

the standard practice of modeling high incomes as following a Pareto distribution. We fit a Pareto distribution to the ranges of net income at the top of the income distribution for each year, and assume that incomes within each range follow this distribution. This allows us to find the total ordinary taxable income of each percentile group and group’s weighted average log after-tax share.

The remainder of this appendix describes the specifics of how we construct the figures we need for our analysis. B. Capital Gains Corrections The Statistics of Income report some data on capital gains by income category throughout our sample period. But because both the tax code and the data on capital gains and losses in the Statistics of Income changed over our sample, our procedure for subtracting capital gains (and adding capital losses) from the reported figures for overall net income is slightly different in different years. 1918–1921. We estimate ordinary taxable income by subtracting “Profits from sales of real estate, stock, bonds, etc.” from net income. The Statistics of Income for these years do not report data on the net capital losses of taxpayers who had net losses. As a result, although taxpayers could deduct these losses in computing ordinary net income, we are unable to add them back into the net income figures. Thus, our estimates for this period correspond to ordinary taxable income minus net capital losses. In later years when data on net capital losses are available, they are only about 5 percent of net income for high-income taxpayers. Other studies of tax responsiveness also neglect net capital losses (for example, Gruber and Saez, 2002). 1922–1923. The Statistics of Income break capital gains into short-term and long-term. We subtract both from reported net income. As with 1918–1921, we are unable to add net capital losses back into the net income figures. 1924–1925. We again subtract both short-term and long-term capital gains from net income. However, the resulting concept is slightly different than in earlier years. Starting in 1924, long-term net capital losses could no longer be claimed as a deduction in computing net income, but instead could be claimed as a 12½ percent tax credit. Thus, when we subtract capital gains from net income, the result is ordinary net income less net short-term capital losses (rather than ordinary taxable income less all net capital losses), which is closer to what we want conceptually.

This change means that there is a conceptual discontinuity in our income measure from 1923 to

1924. To prevent it from affecting our results, when we compute the percentage change in income from 1923 to 1924, we use our 1924 income figures minus eight times the 12½ percent tax credit for long-term capital losses. As a result, we are finding the change in a consistent series (ordinary taxable income minus all capital losses). 1926–1933. Starting in 1926, the Statistics of Income include data on net short-term capital losses (which continued to be deductible in computing net income). We therefore subtract both short-term and long-term capital gains from net income as before, and add short-term capital losses. The resulting measure corresponds to taxable income excluding all capital gains and losses. This change again introduces a discontinuity in our measure. To prevent it from affecting our results, when we compute the percentage change in income from 1925 to 1926, we do not add short-term capital losses to the 1926 income figures. 1934–1937. We subtract “net capital gain” from net income and add “net capital loss.” As with

77

our figures for 1926–1933, the resulting figures correspond to ordinary taxable income. 1938. Beginning in 1938, some assets were no longer classified as capital assets, and gains and losses on them were treated differently than other capital gains and losses. However, data on these gains and losses are reported in the Statistics of Income. We therefore subtract gains on all assets from net income and add losses on all assets (other than short-term losses on assets classified as capital assets, which could not be deducted in computing net income). Again, the resulting figures correspond to ordinary taxable income. 1939–1941. Starting in 1939, short-term losses on assets classified as capital assets from the previous year could be carried forward and deducted against the current year’s capital gains. Since these losses are subtracted in the computation of net income, we add them back in. The remainder of the calculation of ordinary net income is the same as for 1938.

C. Actual and Policy-Induced Changes in Marginal Tax Rates

Knowing a household’s capital gains income and its net income exclusive of capital gains is almost, but not quite, enough to know what its tax liability was, and hence the marginal rate it faced on non-capital-gains income. A household’s computation of both its normal tax and its surtax began with its net income (sometimes, as described above, excluding some or all of capital gains), which equaled gross income less deductions. However, the steps from net income to tax due were slightly different for the two taxes.

For the normal tax, there were several items other than capital gains that received special

treatment. A personal exemption and a credit for dependents were subtracted from net income; until 1936, dividends were excluded; and from 1934 to 1941, 10 percent of the first $14,000 of earned income was also subtracted. Finally, from 1924 to 1931, the normal tax on earned income was reduced by a credit of 25 percent of the normal tax the taxpayer would have had to pay if his or her unearned income was zero. The amount of earned income eligible for the credit varied between $10,000 and $30,000. Fortunately for our purposes, however, normal tax rates were low, and the maximum marginal normal rate was reached at relatively low levels of income. We therefore neglect these complications and assume that all taxpayers at the income levels we are considering paid the top marginal normal rate. For the surtax, the computation of the tax was simpler. The relevant taxable income was either non-capital-gains income (in the years when capital gains were taxed separately) or non-capital-gains income plus the taxable portion of capital-gains income (in years when some or all of capital gains were taxed with other income).

There were only two minor complications with the surtax. First, starting in 1934, the personal

exemption and credit for dependents were subtracted from net income for purposes of the surtax as well as for the normal tax. We have figures on personal exemptions and credits by income range for each year, so we can subtract these from income before finding marginal rates. For high-income taxpayers, the deductions were small relative to income, and so the effects of this adjustment are minor.

Second, from 1924 to 1931, the same 25 percent tax credit on earned income up to some limit that

applied to the normal tax also applied to the surtax. Because most high-income taxpayers were beyond the limit, and because the credit reduced marginal rates by no more than a few percentage points for the others, we neglect this complication.

When legislation changed only tax rates and not the computation of taxable income, finding the

78

policy-induced change in marginal rates at a given level of taxable income is straightforward: the policy-induced change is just the change in the marginal rate at that level of income. When legislation changed how taxable income was computed from year t – 1 to year t, the situation is slightly more complicated. Consider a household with a given level of taxable income in year t – 1. We need to estimate what the household’s taxable income would have been using the year t definition, and then find what the marginal rate would have been at that level. We discuss each case where the definition of taxable income changed in turn. 1921 to 1922. Beginning in 1922, the normal and surtax rates applied to income excluding capital gains, and capital gains were taxed separately. Thus, to know what the relevant taxable income of a 1921 taxpayer would have been under 1922 rules, we should subtract long-term capital gains from the taxpayer’s 1921 income. Unfortunately, the 1921 Statistics of Income do not separate long-term and short-term capital gains. We therefore subtract all capital gains, times the proportion of all capital gains in 1922 for the relevant income group that were long-term. Because capital gains were only a few percent of income in 1921, the effects of this correction are small. 1923 to 1924. Beginning in 1924, long-term capital losses could no longer be deducted from taxable income, but instead resulted in a separate tax credit. The taxable income of a 1923 household under 1924 law therefore equaled its 1923 taxable income plus any long-term capital losses. Since we do not have data on 1923 capital losses, we assume that long-term capital losses as a share of net income for each income group were the same in 1923 as in 1924. We then add the resulting estimates of long-term capital losses in 1923 to the reported 1923 incomes for each group to obtain an estimate of what their taxable income would have been under 1924 law. The effects of this correction are small. 1933 to 1934. There were two changes to how taxable income was calculated in 1934. First, the treatment of long-term capital gains and losses was changed. Rather than being taxed separately, a portion of these gains and losses was included in taxable income, with the fraction varying by the holding period. In addition, the deduction for capital losses (net of any gains) was capped at $2000 per return. In 1933, short-term capital losses and short-term gains (both of which were included in the computation of income subject to the surtax) were similar in magnitude, and long-term losses were much larger than long-term gains. In 1934, reported capital gains income and deductions for losses were similar in magnitude. That is, in both years capital gains and losses on net had little impact on income subject to the surtax. We therefore make no adjustment for the change in the treatment of gains and losses. Second, starting in 1934 the personal exemption and credit for dependents could be deducted from income subject to the surtax. In finding the taxable incomes for the purposes of the surtax that 1933 taxpayers would have had under 1934 law, we therefore subtract their exemptions and dependent credits. 1937 to 1938. Starting in 1938, capital gains and losses on assets held more than 18 months were again taxed at a separate rate. Gains on assets held less than 18 months, however, were now entirely included in taxable income, and none of current-year net losses could be deducted in computing net income. The 1937 Statistics of Income do not separate capital gains income by holding period. The 1938 Statistics of Income, however, separate it according to whether the holding period was more or less than 18 months. To approximate the effect of the change on the relevant taxable income a 1937 taxpayer would have had under the 1938 code, we assume that this division for a given income range was the same in 1937 as in 1938. We assume that half of the long-term capital gains were already excluded in 1937 (the actual fraction varied from 20 to 70 percent depending on the holding period), and that none of the short-term gain was excluded. Since all long-term gains were taxed separately in 1938, this allows us to estimate how much lower a taxpayer’s relevant taxable income would have been under the 1938 rules. We also add back in net losses, since these were no longer deductible.

79

1939 to 1940. The Revenue Act of 1940 lowered all personal exemptions by 20 percent. To find the taxable incomes that 1939 taxpayers would have had under 1940 law, we therefore add back in 20 percent of their personal exemptions. The effects of this adjustment are minor. 1940 to 1941. In 1941, personal exemptions were reduced by an additional 25 percent for joint filers and 6 percent for other taxpayers. 57 percent of the value of all personal exemptions in 1940 was claimed by joint filers (1940 Statistics of Income, p, 121). To find the taxable incomes that 1940 taxpayers would have had under 1941 rules, we therefore add back in 17 percent of their personal exemptions. The effects of this adjustment are again minor. D. Retroactive Changes

If a change to the tax code was enacted at the end of the year or after the end of the year, our baseline measure of tax rates uses the rates that were in effect during the year, not the rates that were applied ex post. For the one case where a change was enacted after mid-year but well before year-end, we try both approaches.

There are five cases of retroactive changes enacted after mid-year. Tax bills enacted in 1919,

1924, and 1926 changed taxes for the previous year; a Congressional resolution enacted on December 16, 1929 changed 1929 taxes; and the Revenue Act of 1941, enacted on September 20, 1941, changed 1941 taxes. Our baseline measures of marginal rates and policy-induced changes in marginal rates use the tax code in effect during 1918, 1923, 1925, and 1929, and ignore the retroactive changes. The 1941 change, however, was in effect for a non-trivial part of the year, and taxpayers likely knew before the bill was passed that tax rates would be raised. Our baseline measure therefore uses the rates specified by the 1941 act in computing marginal rates. However, we also consider the effects of coding this as no change in rates in 1941. The treatment of 1941 has no important effect on our results. In addition, as described in the text, we also consider series for marginal rates that use the rates that were applied ex post in all cases. E. Interpolation and Aggregation to Construct Data for Income Percentile Groups

To construct figures for different percentile groups rather than for the income ranges in the Statistics of Income, we sometimes need estimates of the breakdown of income within a given income range. For example, if some but not all of the filers in the $150,000–$200,000 range are in the top percentile group in some year, we need to estimate the fraction of the ordinary taxable income of the filers in that range that went to the filers in the top percentile group. Similarly, suppose all households in the $90,000–$100,000 range are in the second percentile group in some year, but the marginal rate changed within this range. Then we need to estimate the fraction of the income of this group accruing to households facing each marginal rate.

The highest income ranges in the Statistics of Income usually have fewer than a hundred

households. And, the other income ranges that are relevant to our analysis are generally narrow, such as $90,000–$100,000. As a result, our estimates are not sensitive to the details of our interpolation procedure. The specific approach we use is to fit a Pareto distribution for each year to the income categories at or above the category that includes the return at the 99.95th percentile of the income distribution in that year. Because the treatment of the very top income categories in the Statistics of Income varies over time, we aggregate the taxpayers with incomes over $1.5 million into a single category. This group always includes less than 1/1000th of 1 percent of households. The Pareto distribution function is

80

(A1) 𝐹(𝑌) = 1 − �𝑘𝑌�𝜃

for 𝑌 ≥ 𝑘.

Let Li and Hi denote the bottom and top of income category i, and assume that Li > K. The probability that a return falls in category i is

(A2) 𝑃𝑖 = �𝑘𝐿𝑖�𝜃

− �𝑘𝐻𝑖�𝜃

.

Thus, the likelihood function is

(A3) 𝐿 = ��𝑃𝑖𝑁𝑖𝑀

𝑖=1

�𝑁!

𝑁1!𝑁2! …𝑁𝑀!,

where M is the number of income categories, Ni is the number of returns in category i, and N is the total number of returns in the sample we are considering. The log likelihood function is therefore

(A4) ln 𝐿 = 𝐾 + �𝑁𝑖 ln𝑃𝑖

𝑀

𝑖

,

where 𝐾 ≡ ln(𝑁!) − ∑ ln(𝑁𝑖!)𝑀𝑖=1 . Note that K does not depend on the parameters of the distribution.

We estimate the model by maximum likelihood for each year. The number of income categories

in the sample varies from 13 to 18. The estimates of θ (which is the parameter relevant to the interpolation) range from a low of 1.41 in 1929 to a high of 2.01 in 1920. These estimates are similar to other estimates for income distributions. The estimates are extremely precise: the standard error for the estimate of θ is always less than 0.001.

We then use the Pareto parameters to construct the data that we need on ordinary taxable incomes

by percentile group. Suppose, for example, that 40 percent of the filers in the $150,000–$200,000 range are in the top percentile group in some year, and that the estimate of θ for that year is 1.5. Then the assumption that incomes follow a Pareto distribution implies that 43.5 percent of the income of the filers in this range went to those in the top percentile group.

Similarly, we use the Pareto parameters to estimate each percentile group’s income-weighted

average log after-tax share and the policy-induced change in a group’s income-weighted average log after-tax share.8

8 To see why the change in the log of the taxable income of a group should be related to the change in the group’s income-weighted log after-tax share, suppose the taxable income of the household at percentile i of the income distribution in year t is given by ln𝑦𝑖𝑡 = 𝛼𝑖 + 𝛽𝑡 + 𝛾 ln 𝑆𝑖𝑡 + 𝜀𝑖𝑡 , where Sit is the household’s after-tax share. Then ln yi,t+1 − ln𝑦𝑖𝑡 = 𝛽�𝑡+1 + 𝛾(ln 𝑆𝑖,𝑡+1 − ln 𝑆𝑖𝑡) + 𝜀�̃�,𝑡+1 (where 𝛽�𝑡+1 ≡ 𝛽𝑡+1 − 𝛽𝑡 , 𝜀�̃�,𝑡+1 ≡ 𝜀𝑖,𝑡+1 − 𝜀𝑖𝑡). This in turn implies 𝑦𝑖,𝑡+1 − 𝑦𝑖𝑡 ≅ 𝑦𝑖𝑡[𝛽�𝑡+1 + 𝛾�ln 𝑆𝑖,𝑡+1 − ln 𝑆𝑖𝑡) + 𝜀�̃�,𝑡+1�. Thus, summing over members of the percentile group being considered, and letting Yt be the total taxable income of the group, we have:

∑ (𝑦𝑖,𝑡+1 − 𝑦𝑖𝑡𝑖 )𝑌𝑡

≅ ��𝑦𝑖𝑡𝑌𝑡�

𝑖

𝑦𝑖𝑡[𝛽�𝑡+1 + 𝛾�ln 𝑆𝑖,𝑡+1 − ln 𝑆𝑖𝑡) + 𝜀�̃�,𝑡+1�

= 𝛾�𝑦𝑖𝑡𝑌𝑡𝑖

�ln 𝑆𝑖,𝑡+1 − ln 𝑆𝑖𝑡� + 𝜈𝑡+1,

For example, consider again a case where all households in the $90,000–$100,000 range

where 𝜈𝑡+1 ≡ 𝛽�𝑡+1 + ∑ �𝑦𝑖𝑡 𝑌𝑡� � 𝜀�̃�,𝑡+1.𝑖 This in turn implies the posited relationship:

81

are in the second percentile group. Suppose that 5 percent of the income of the filers in this range was either untaxed or taxed separately, that one marginal rate applied to $80,000–$90,000 and a higher one to $90,000–$100,000, and that the Pareto parameter for the year is 1.5. Then our assumptions imply that 49.3 percent of the ordinary taxable income of the filers in this range was taxed at the lower marginal rate and 50.7 percent was taxed at the higher rate. This would be one part of the overall weighted average for this percentile group. Similarly, to find the policy-induced change from one year to the next, we find the marginal rate at each level of income under each year’s tax code, weight using the first year’s income distribution, and find the difference.

The calculations described in this appendix are clearly not exact. Most importantly, we assume

that quantities that we need to subtract from net income, such as capital gains, are a constant proportion of income within each income range. As Barro and Sahasakul (1983) observe in a different context, the aggregates will be reasonably accurate either if the quantities we need to subtract from net income do not vary greatly as a share of income among members of the group or if the log after-tax share is approximately linear in taxable income over the relevant range. In our case, because the adjustments involve only a moderate fraction of net income, and because the log after-tax share fell fairly steadily with income, the approximation error is likely to be small. And because the actual changes in marginal rates in this period were so large, even moderate errors would have little impact on our estimates.

As a check on our calculations, we have computed the implications of our assumptions for the

amount of taxes paid by the households in selected income ranges for certain years. We find that the calculations match actual taxes paid quite well, sometimes remarkably so.9

ln𝑌𝑡+1 − ln𝑌𝑡 ≅ 𝛾 ��𝑦𝑖𝑡𝑌𝑡

ln 𝑆𝑖,𝑡+1𝑖

− �𝑦𝑖𝑡𝑌𝑡𝑖

ln 𝑆𝑖𝑡� + 𝜈𝑡+1.

9 The largest discrepancies we have found involve overpredictions of the normal tax before 1936. The discrepancies appear to stem from the fact that dividends were exempt from the normal tax until 1936, and some high-income households had sufficiently high deductions and low non-dividend income that excluding only a portion of their dividend income was enough to reduce their normal tax liability to zero. Thus, our estimates appear to overstate average marginal rates for these years. But, since the marginal normal tax rate was low and most high-income households paid some normal tax, the errors appear small.

82

REFERENCES Auten, Gerald, and Robert Carroll. 1999. “The Effect of Income Taxes on Household Income.” Review

of Economics and Statistics 81 (November): 681-693. Barro, Robert J., and Chaipat Sahasakul. 1983. “Measuring the Average Marginal Tax Rate from the

Individual Income Tax.” Journal of Business 56 (October): 419-452. Brownlee, W. Elliot. 2000. “Historical Perspectives on U.S. Tax Policy toward the Rich.” In Joel

Slemrod, ed., Does Atlas Shrug? The Economic Consequences of Taxing the Rich (New York: Russell Sage Foundation), 29-73.

DeLong, J. Bradford and Lawrence H. Summers. 1991. “Equipment Investment and Economic Growth.”

Quarterly Journal of Economics 106 (May): 445-502. Evans, George Herberton, Jr. 1948. Business Incorporations in the United States, 1800-1943 (New

York: National Bureau of Economic Research). Feldstein, Martin. 1995. “The Effect of Marginal Tax Rates on Taxable Income: A Panel Study of the

1986 Tax Reform Act.” Journal of Political Economy 103 (June): 551-572. Giertz, Seth H. 2007. “The Elasticity of Taxable Income over the 1980s and 1990s.” National Tax

Journal 60 (December): 743-768. Goolsbee, Austan. 1999. “Evidence on the High-Income Laffer Curve from Six Decades of Tax

Reform.” Brookings Papers on Economic Activity, no. 2, 1-64. Goolsbee, Austan. 2000. “What Happens When You Tax the Rich? Evidence from Executive

Compensation.” Journal of Political Economy 108 (April): 352-378. Gordon, Roger H., and Joel Slemrod. 2000. “Are ‘Real’ Responses to Taxes Simply Income Shifting

between Corporate and Personal Tax Bases?” In Joel Slemrod, ed., Does Atlas Shrug? The Economic Consequences of Taxing the Rich (New York: Russell Sage Foundation), 240-280.

Gruber, Jon, and Emmanuel Saez. 2002. “The Elasticity of Taxable Income: Evidence and

Implications.” Journal of Public Economics 84 (April): 1-32. Kopczuk, Wojciech. 2005. “Tax Bases, Tax Rates and the Elasticity of Reported Income.” Journal of

Public Economics 89 (December): 2093-2119. Lindsey, Lawrence B. 1987. “Individual Taxpayer Response to Tax Cuts: 1982-1984, with Implications

for the Revenue Maximizing Tax Rate.” Journal of Public Economics 33 (July): 173-206. Lipsey, Robert E. and Doris Preston. 1966. Source Book of Statistics Relating to Construction. (New

York: National Bureau of Economic Research). Moffitt, Robert A., and Mark O. Wilhelm. 2000. “Taxation and the Labor Supply Decisions of the

Affluent.” In Joel Slemrod, ed., Does Atlas Shrug? The Economic Consequences of Taxing the Rich (New York: Russell Sage Foundation), 193-234.

83

Piketty, Thomas, and Emmanuel Saez. 2001. “Income Inequality in United States, 1913-1998.” National Bureau of Economic Research Working Paper No. 8467 (September).

Romer, Christina D., and David H. Romer. 2009. “A Narrative Analysis of Postwar Tax Changes.”

Unpublished paper, University of California, Berkeley (June). Smiley, Gene, and Richard H. Keehn. 1995. “Federal Personal Income Tax Policy in the 1920s.”

Journal of Economic History 55 (June): 285-303. U.S. Board of Governors of the Federal Reserve System. 1943. Banking and Monetary Statistics

(Washington, D.C.: Board of Governors of the Federal Reserve System). U.S. Bureau of Internal Revenue, United States Department of the Treasury. Statistics of Income.

Various years. U.S. Congress, House of Representatives Reports. Various numbers. U.S. Congress, Senate Reports. Various numbers. U.S. Department of the Treasury. The Annual Report of the Secretary of the Treasury on the State of the

Finances. Various years. Woolley, John, and Gerhard Peters. The American Presidency Project (www.presidency.ucsb.edu).