The Elasticity of Corporate Taxable Income -Evidence from South Africa I

Collen Ledigaa, Nadine Riedela,b,∗, Kristina Strohmaierc

aUniversity of BochumbCESifo Munich

cUniversity of Tubingen

Abstract

How strongly do firms reduce their reported taxable income when corporate tax

rates increase? We add new evidence from an emerging economy to a so far

developed-country-dominated literature.

Keywords: Corporate taxation, developing countries, bunching

1. Introduction

While a growing empirical literature assesses behavioral responses to corporate

taxation in industrialized economies, limited access to corporate micro data has

largely prevented analogous analyses for less developed countries. Our study

helps filling this gap. We draw on the population of corporate tax returns in5

South Africa in 2009–2015 to determine the elasticity of corporate taxable in-

come by exploiting non-linearities in the corporate tax schedule. The analysis

yields large elasticities for firms with incomes around the first kink (∼70,000

South African Rand or e5,000 p.a.), while entities with incomes around the

second and third kink (∼R365,000 and ∼R550,000 p.a.) respond only moder-10

ately to marginal tax increases.

Two further insights emerge. Firstly, we show that firms bunch at zero cor-

porate taxable income. This supports Johannesen et al. (2016) who report a

∗Corresponding author: Nadine Riedel, Email: nadine.riedel@rub.de

Preprint submitted to Elsevier 10th June 2018

similar pattern for multinational entities and interpret it as evidence for inter-

national income shifting. Our data suggests that zero is also a focal point for15

domestic tax avoidance and evasion - despite the fact that, for many firms, the

South African tax provisions imply no, or even disincentives, for bunching beha-

vior. This supports the notion of some taxpayers having imperfect information

on the tax code (Hoopes et al., 2015).

Finally, we follow prior papers and assess how firms manage to bunch at20

kink points. As total assets, sales, costs and other accounting measures show no

major discontinuities at our tax thresholds, bunching firms do not systematically

differ from other entities near the kinks. This contrasts prior findings (see,

e.g., Bachas and Soto, 2016), which suggest that corporate tax responses are

dominated by cost-overreporting.125

2. Institutional Background, Data and Methodology

South Africa is an upper-middle-income economy with a tax-to-GDP ratio of

29% in 2015 (compared to a 34%-OECD-average).2 It levies a “standard” tax

rate of 28% on corporate income and applies a progressive “Small Business

Corporations” (SBCs) tax scheme, where income of eligible businesses is taxed30

at lower rates. Kinks in the SBC-schedule will in the following be used to identify

the corporate taxable income elasticity. Figure 1 depicts the three kinks of the

marginal tax rate schedule for 2015.3

To quantify behavioral responses to taxation, we estimate the elasticity of

taxable income (ETI), which captures all kinds of behavioral responses including35

real production shifts, tax avoidance and tax evasion. We employ the nonpara-

metric approach of Saez (2010), which exploits clustering of firms around tax

1After finalizing the first working paper version, we became aware of Boonzaaier et al.(2017) who equally estimate corporate income elasticities in South Africa. Compared to theirstudy, our data offers the advantage that firms subject to SBC-taxation can be clearly iden-tified, which allows for more precise estimates and may explain differences in result patterns.Contrary to others (including Boonzaaier et al., 2017), we moreover assess bunching at zerocorporate taxable income.

2Information on tax-to-GDP ratios was obtained from OECD revenue statistics.3See http://www.sars.gov.za for tax schedules in all years and eligibility criteria.

2

Figure 1: SBC Tax Schedule 2015

Taxable income

Mar

gin

al

Tax

Rat

e

R70, 700 R365, 000 R550, 000

10%

20%

30%

kinks. To estimate the excess mass of firms around the kink, we determine the

counterfactual density by running a local polynomial regression on binned data,

while excluding bins within the bunching window. The latter is determined40

by an endogenous data-driven procedure (Dekker et al., 2016). To account for

increasing tax thresholds over time when pooling the tax years, the data is res-

caled relative to the threshold and a weighted average value is used. Finally, we

only pool years with identical changes in marginal tax rates at a given kink.

3. Results45

3.1. Bunching at Kinks in the Tax Schedule

Figure 2 shows the distribution of firms at the first kink in the SBC tax schedule.

The data is subdivided in two time periods, 2009–2012 and 2013–2016, as the

marginal tax jump differs between the intervals (being 0 to 10% in the former

case and 0 to 7% in the latter).50

For both time spans, we see clear bunching for firms under the SBC tax

schedule (and no bunching for Non-SBCs). For the years 2010–2012 (2013–

2016), we estimate a highly significant excess mass of taxpayers b = 6.62 (b =

7.48). Using a weighted average for the threshold value4, this translates into an

4We calculate this weighted average using the share of bunchers as weights.

3

Figure 2: Bunching at the First Threshold

050

010

0015

0020

00Fr

eque

ncy

-20000 -10000 0 10000 20000Distance First Threshold

SBCs, 2010-2012b: 6.6214 b_se: 0.2683 e: 0.7913 binwidth: 717

600

800

1000

1200

1400

Freq

uenc

y

-20000 -10000 0 10000 20000Distance First Threshold

Non-SBCs, 2010-2012b: 0.0594 b_se: 0.0646 e: 0.0095 binwidth: 964

050

010

0015

00Fr

eque

ncy

-20000 0 20000Distance First Threshold

SBCs, 2013-2016b: 7.4797 b_se: 0.3589 e: 1.3280 binwidth: 847

500

600

700

800

900

1000

Freq

uenc

y

-20000 0 20000Distance First Threshold

Non-SBCs, 2013-2016b: -0.0727 b_se: 0.0639 e: -0.0179 binwidth: 1176

Notes: Observed distribution: Blue dotted line; Estimated counterfactual: Grey dotted line.

ETI of e = 0.79 (e = 1.33).55

Table 1 reports the results for the other two thresholds. Firms respond

at both income levels. Compared to the lowest threshold, the bunching at

the second and third threshold translates into significantly smaller elasticity

estimates though (ranging between 0.08 and 0.15).

While existing papers report large tax elasticities in less developed country60

contexts (see, e.g., Waseem, 2018), we find mixed evidence for the corporate

income tax in South Africa. Specifically, our results point to moderate tax

elasticities—similar in size to developed country estimates—for larger SBCs at

the upper two kinks; the elasticities at the lower end of the income distribu-

tion are, in turn, sizable and well exceed their developed country counterparts,65

compare e.g. evidence for the UK in Devereux et al. (2014).5 This pattern may

5Devereux et al. (2014) report an ETI for the UK of 0.14–0.18 (0.54–0.57) for companieswith profits around the £300k kink (£10k kink).

4

root in differences in institutional and economic tax evasion constraints faced

by entities of different size.6

Table 1: Results Bunching Analyses

Bunching window binwidth excess mass se elasticity

Second threshold2010–2012 [ -10,362 ; 6,705 ] 1219 6.6147 0.5023 0.12052013 [ -6,090 ; 2,610 ] 1740 4.3389 0.3732 0.08432014–2016 [ -10,230 ; 930 ] 1860 3.8754 0.2484 0.1210

Third threshold2014–2016 [ -3,792 ; 1,264 ] 2528 2.9359 0.3142 0.1454

3.2. Bunching at Zero

On top, one of the most salient features of our corporate taxable income dis-70

tribution is that firms strongly bunch at a taxable income of zero, cf. Figure 3

for active7 SBCs and Non-SBCs in 2009–2015. Strikingly, this aspect has been

largely ignored in the existing literature - except for a recent paper by Johan-

nesen et al. (2016), who show that multinational firms bunch at zero income

and interpret the pattern as evidence for international tax avoidance.75

As SBCs are by definition domestic entities, the results in Johannesen et al.

(2016) are found to carry over to national entities in our analysis, despite the fact

that the South African tax schedule provides no or even disincentives for SBCs

to bunch at zero.8 One possible explanation for the pattern is that taxpayers

have incomplete information on the business tax schedule (see, e.g., Hoopes80

et al., 2015) and zero is a focal point for domestic evasion/avoidance strategies

based on these false perceptions.9

6We tested for differences in response rates along observed dimensions other than size,namely industry affiliation and host region and found no evidence of systematic variation.

7=firms with positive sales and deductions8Above zero, income is exempted from taxation; below zero, indefinite loss-carryforward

provisions in the South African tax schedule provide an implicit tax subsidy (see, e.g., Domarand Musgrave, 1944), making the budget set convex or linear. For Non-SBCs, the budget setis linear around zero if they expect positive future income.

9A quantitative interpretation (yielding an ETI) is not straightforward at zero, given that

5

Figure 3: Bunching at Zero

020

0040

0060

0080

00Fr

eque

ncy

-20000 0 20000Distance Zero Threshold

b: 11.8494 b_se: 0.4539 e: - binwidth: 446SBCs

050

000

1000

0015

0000

2000

0025

0000

Freq

uenc

y

-20000 0 20000Distance Zero Threshold

b: 29.7389 b_se: 1.0126 e: - binwidth: 471Non-SBCs

3.3. Anatomy of Response

Finally, we follow previous papers and plot the median of total assets, sales

and total costs of sales against taxable income around the kinks to determine85

how firms bunch (cf. Figure 4). We cannot detect any clear anomalies except

for a small jump in costs for the second threshold. These findings contrast

recent evidence (see, e.g., Bachas and Soto, 2016), which suggests that firms’

tax responses are dominated by tax evasion through cost-overreporting. Our

results are, in turn, consistent with evasion by a proportional underreporting of90

sales, deductions and assets (with evading firms “mimicking” taxpayers at the

threshold) and with real responses - hence underpinning that response margins

may differ across countries and institutional contexts.

4. Conclusion

Three conclusions follow: First, the pattern of identified tax elasticities suggests95

that progressive business taxation may be optimal from an efficiency perspective.

Second, if zero serves as a focal point for (domestic) tax evasion/avoidance

strategies, variation in “zero-bunching” across time and across firm-subgroups

may be used to identify correlates of evasion/avoidance behavior. Third, our

there are no (clear) jumps in the marginal tax rates and relative income responses cannot becalculated. The estimated excess mass for SBCs (Non-SBCs) is 11.8 (29.7).

6

Figure 4: Anatomy of Response

(a) Total Assets (log)

(b) Sales (log)

(c) Total Costs (log)

Notes: The panels show the medians of total assets, sales and costs around the first and secondthreshold for SBCs (filled, blue dots) and Non-SBCs (non-filled, grey dots).

7

analysis stresses that tax response margins may depend on institutional contexts100

and that findings for one country may not carry over to others.

References

Pierre Bachas and Mauricio Soto. Not(ch) your average tax system: Corporate

taxation under weak enforcement. mimeo, 2016.

Wian Boonzaaier, Jarkko Harju, Tuomas Matikka, and Jukka Pirttila. How105

do small firms respond to tax schedule discontinuities? Evidence from south

african tax registers. VATT Working Papers, 2017.

Vincent Dekker, Kristina Strohmaier, and Nicole Bosch. A data-driven proced-

ure to determine the bunching window - an application for the netherlands.

CPB Discussion Paper, 2016.110

Michael P. Devereux, Li Liu, and Simon Loretz. The elasticity of corporate

taxable income: New evidence from UK tax records. American Economic

Journal: Economic Policy, 6(2):19–53, May 2014.

E. Domar and R.A. Musgrave. Proportional income taxation and risk-taking.

Quarterly Journal of Economics, 58:388–422, 1944.115

Jeffrey L. Hoopes, Daniel Reck, and Joel Slemrod. Taxpayer search for inform-

ation: Implications for rational inattention. American Economic Journal:

Economic Policy, 7(3):177–208, 2015.

Niels Johannesen, Thomas Tørsløv, and Ludvig Wier. Are less developed coun-

tries more exposed to multinational tax avoidance? Method and evidence120

from micro-data. WIDER Working Paper, 2016.

Emmanuel Saez. Do taxpayers bunch at kink points? American Economic

Journal: Economic Policy, 2(3):180–212, 2010.

Mazhar Waseem. Taxes, informality and income shifting: Evidence from a

recent pakistani tax reform. Journal of Public Economics, 157:41–77, 2018.125

8