THE IMPACT OF CLIMATE CHANGE ON THE TOURISM · PDF fileTHE IMPACT OF CLIMATE CHANGE ON THE...

ECLAC – Project Documents collection Caribbean Development Report, Volume 2

204

THE IMPACT OF CLIMATE CHANGE ON THE TOURISM SECTOR IN SELECTED CARIBBEAN

COUNTRIES

Sandra Sookram

Abstract

Tourism is an important source of economic growth in the Caribbean and one of the most important industrial sectors for some countries in the subregion. The purpose of this study is to estimate the economic impact of climate change on the tourism sector in nine countries in the Caribbean Basin: Aruba, Barbados, the Dominican Republic, Guyana, Jamaica, Montserrat, the Netherlands Antilles, Saint Lucia and Trinidad and Tobago.

A typical tourism demand function, with tourist arrivals as the dependent variable, is used in the analysis. The independent variables employed in the study are: destination country Gross Domestic Product per capita (GDPPC) and consumer price index (CPI), source country Gross Domestic Product (GDP), and oil prices, to proxy transportation costs between source and destination countries. Two climate variables, temperature and precipitation, are used to augment the tourism demand function. In the first stage of the analysis, the baseline is established using annual data for the period 1989-2007.With respect to the climatic variables, the results at this initial stage of the study indicate a negative effect of both the temperature and precipitation variables on tourist arrivals.

In the second stage of the analysis, the cost of climate change to the tourism sector is forecast to the end of the century under three weather scenarios: A2, B2 and Business as Usual (BAU) (the mid-point of the A2 and B2 scenarios). The estimation is undertaken for the impact of changes in temperature and precipitation, sea level rise, extreme events (for example, the frequency and intensity of hurricanes) and the destruction of ecosystems. The total estimated costs to tourism for the Caribbean subregion under the three weather scenarios at the end of the century are all very high, ranging from US$ 43.9 billion under the B2 scenario to US$ 46.3 billion under the BAU scenario.


205

INTRODUCTION

Tourism130 is an important source of economic growth in the Caribbean. Indeed, it is one of the most important industrial sectors in some countries in the subregion.131 Furthermore, its significance to any one country increases in accordance with the quantity of tourism-related services associated with the sector. The World Tourism Organisation (WTO) has declared that international tourism figures increased by 2% during 2007, and has predicted that the number of international tourists will reach the 1.6 billion mark by the year 2020.132

What needs to be considered is how many of those potential tourists would visit the Caribbean, and what impact climate change would have on that figure. There is no doubt that climate is an important influence on the tourism sector. Numerous studies that analyse climate data indicate that our climate is changing; for example, the average global temperature has increased by approximately 0.6⁰C during the twentieth century. More than that, the rate of increase in air temperature in the Caribbean subregion has exceeded the international mean (Mimura and others, 2007).

This study attempts to determine the possible impact of climate change on nine Caribbean countries,133 using tourist arrivals, climate (represented by temperature and precipitation) and other economic data for the 1989-2007 period. A key objective is to estimate the economic impact of climate change on tourism income under two climate change scenarios (A2 and B2). The main objective of this study is to suggest adaptation and mitigation strategies for the tourism sectors of these nine countries.

According to Braun and others (1999), environmental factors are key components when tourists choose a holiday destination. There is convincing evidence to show that the world’s climate will continue to change during this century. Future variations in temperature and other aspects associated with climate change will have differing effects on different regions worldwide. Table 1 shows the major impacts of climate change and their implications for tourism destinations. It is highly likely that most of these direct effects of climate change, and their subsequent indirect effects, would have an impact on the Caribbean subregion.

There are many studies on the demand for tourism and modelling tourism demand. This study uses a typical tourism model and expands the model to include two climatic variables: temperature and precipitation. The model is then used to forecast the likely impact of changes in temperature and precipitation on the countries in the study.

This study is organized as follows: section 2 reviews the literature related to tourism and climate change; section 3 outlines the methodology followed in the study; section 4 presents the results; section 5 examines the Special Report on Emission Scenarios (SRES) emission scenarios, with particular reference to A2 and B2 and BAU scenarios which are used to forecast the impacts of climate change for nine tourist destinations in the Caribbean; section 6 examines some of the mitigation and adaptation strategies; and section 7 concludes.

130 According to the World Tourism Organization, tourism can be defined as “the activities of persons travelling to and staying in places outside their usual environment for not more than one consecutive year.” 131 As an example, income from tourism in Saint Lucia ranged from 36% to 50% of Gross National Income during the period 1989-2007. 132 See WTO website: http://www.unwto.org/index.php 133 The countries are: Aruba, Barbados, the Dominican Republic, Guyana, Jamaica, Montserrat, the Netherlands Antilles, Saint Lucia, and Trinidad and Tobago.


206

TABLE 1 MAIN IMPACTS OF CLIMATE CHANGE AND THEIR IMPLICATIONS FOR TOURISM

Impact Implications for Tourism Warmer temperatures Altered seasonality, heat stress for tourists, cooling costs, changes in: plant-

wildlife-insect populations and distribution range, infectious disease ranges Decreasing snow cover and shrinking glaciers

Lack of snow in winter sport destinations, increased snow-making costs, shorter winter sports seasons, aesthetics of landscape reduced

Increasing frequency and intensity of extreme storms

Risk for tourism facilities, increased insurance costs/loss of insurability, business interruption costs

Reduced precipitation and increased evaporation in some regions

Water shortages, competition over water between tourism and other sectors, desertification, increased wildfires threatening infrastructure and affecting demand

Increased frequency of heavy precipitation in some regions

Flooding damage to historic architectural and cultural assets, damage to tourism infrastructure, altered seasonality (beaches, biodiversity, river flow)

Sea level rise Coastal erosion, loss of beach area, higher costs to protect and maintain waterfronts and sea defences

Sea surface temperature rise Increased coral bleaching and marine resource and aesthetic degradation in dive and snorkel destinations

Changes in terrestrial and marine biodiversity

Loss of natural attractions and species from destinations, higher risk of diseases in tropical-subtropical countries

More frequent and larger forest fires

Loss of natural attractions, increase of flooding risk, damage to tourism infrastructure

Soil changes (such as moisture levels, erosion and acidity)

Loss of archaeological assets and other natural resources, with impacts on destination attractions.

Source: WTO-UNEP-WMO (2008) Climate Change and Tourism: Responding to Global Challenges

REVIEW OF THE LITERATURE According to Scott and others (2004), the interrelationship between the weather and tourism has featured in studies dating from the 1930s. In 1936, for example, Selke wrote on the geographic aspects of the German tourist trade. So far, these studies have been few, and only in recent times has the literature on tourism started to increase. These tourism studies, as stated by Hamilton and Tol (2007), focused mainly on economic factors and did not include climate variables in the modelling process. The studies had short time-horizons, and climate was taken to be a constant variable. However, there is much evidence to show that climate will change in the long run, and that this change is being hastened by human activities.

More recently, researchers have begun to include climatic variables and, in some cases, a tourism climatic index. One of the first studies on climate change and tourism demand employed temperature to estimate the effect of forecasted changes in temperature on the ski industry in Switzerland (Koenig and Abegg, 1997). The study revealed that, under the present conditions, with prevailing temperature and a snow line of 1,200 m,134 there was an 85 % chance that there would be snow to keep the industry functioning. However, if temperatures were to increase by 2⁰C, then only 65% of all Swiss ski areas would be snow reliable. This would clearly have serious implications for the growth of that sector of the industry.

The increasing volume of literature on the impact of climate on tourism demand is due to the recognition that a more precise modelling of tourism demand must include weather and climate, since they are significant influences on the tourism industry. The climatic factors identified as having the most impact on tourism are temperature, sunshine, radiation, precipitation, wind, humidity and fog (Stern, 2006; Hamilton and Lau, 2004). These factors are significant both to the tourist’s assessment of his or her health and well-being, and to the tourism industry. It is therefore essential that these elements be measured and evaluated, since they form an important resource for tourism.

134 In this study it was mentioned that Pfund (1993) illustrated that a minimum altitude of 1,200 m. (the line of snow reliability) is necessary for the ski industry to be a feasible undertaking.


207

The literature associating climate with tourism implies that changes in climate are likely to affect both the length of the season for tourists and the expected environment. The literature has shown that temperature could potentially have positive implications for the length of the season and for the environment, while other studies have found results to indicate that it has negative implications for tourism. Lise and Tol (2002), using cross-section data, undertook a cross-section analysis on tourists originating in Organisation for Economic Cooperation and Development (OECD) countries, and found that the optimal temperature for their destination countries ranged from 21⁰C to 24⁰C. The implication of this finding is that the predicted increasing global temperatures in certain regions of the world would have devastating effects on the tourist industries of those countries. Hamilton and others (2005) used a simulation model to investigate the effects of climate change on international tourism using the A1B scenario.135 They found that international tourism is expected to increase in the coming decades, but may become sluggish later on in the century.

Another study, Berrittella and others (2006), used a computable general equilibrium model to measure the potential effects of climate change. They employed two pathways to capture the impact of climate change, namely, modifications in the composition of final consumption, and international income transfers. The rationale was that spending by visitors has an impact on consumption and income transfers in the domestic economy. The Berrittella and others (2006) study predicted that, at the international level, changes in climate would eventually lead to a loss in welfare, and that that loss would be disproportionately spread across the various regions of the world.

Temperature is considered to be the most important climate variable in the analysis of tourism demand because, outside a certain range, it affects comfort. There is evidence to show that other weather parameters are also important, for example, rain, wind and hours of sunshine (Scott and McBoyle, 2006). If any of these parameters is to be included in the analysis of tourism flows, it must be included as a determinant or in an index. Many studies include both temperature and precipitation to examine the impact of climate on tourism demand (for example, see Scott and McBoyle, 2006).

There have been very few studies on the impact of climate change on tourism demand in the Caribbean. One noteworthy microanalysis by Uyarra and others (2005) examines the significance of environmental characteristics in influencing the choices made by tourists. The study used a self-administered questionnaire on tourists visiting Bonaire and Barbados – 316 from Bonaire and 338 from Barbados. The study established that warm temperatures, clear waters and low health risks were the main environmental attributes important to tourists visiting the islands. The study found that visitors to Bonaire placed additional importance on marine wild life attributes while tourists going to Barbados had a preference for certain beach characteristics. Uyarra and others (2005) examined the impact of climate change by asking respondents about the likelihood of their returning to these islands in the event of coral bleaching and sea level rise. They found that more than 80% of the visitors to Bonaire and Barbados would be expected not to return to the islands in the event of these occurrences. Mather and others (2005) examined the attraction of the Caribbean as a tourist destination for travellers from North America. This study established that the Caribbean subregion is likely to be less attractive to tourists due to factors such as increased temperatures, beach erosion, deterioration of reef quality and greater health risks.

The climate change variables being used in this study (temperature and precipitation) are considered to be significant determinants of tourism in the Caribbean for important reasons. Trenberth and others (2007) have highlighted the fact that temperatures in the Caribbean subregion have been warming at a rate ranging from 0.0⁰C to 0.5⁰C per decade for the period 1971-2000. In a related study, Peterson and others (2002) have reported that in the Caribbean, the percentage of days with cold temperatures has decreased since the 1950s, while the percentage of days with very warm maximum 135 The A1 emission scenario is outlined briefly in table 8. The A1B scenario is a subset of the A1 scenarios and emphasizes the technological element of the A1 scenarios; in particular A1B incorporates a balanced weighting on all energy sources.


208

or minimum temperatures has increased significantly. In relation to precipitation, it was found that the number of heavy rainfall occurrences has been on the increase (Trenberth and others, 2007).

MODELLING TOURISM DEMAND IN THE CARIBBEAN

The tourism demand function: a review of the literature

The literature on the demand for tourism indicates that tourist flows between the destination and source countries can be explained using a demand function. A review of the literature shows that, in order to measure tourism flows, the majority of tourism demand studies use either the number of arrivals to the destination country, or the amount of expenditure by tourists. While some researchers suggest that the dependent variable in the tourism demand equation should be tourist expenditure, Crouch and Shaw (1992) demonstrate that approximately 70% of the studies that estimate tourism demand functions have employed tourist arrivals as the dependent variable. In this study, the number of tourist arrivals has been used as the dependent variable. The literature on tourism demand suggests that a number of explanatory variables can be used to investigate tourism demand. The independent variables used in this study are as follows: Gross Domestic Product per capita (GDPPC) in the destination country (in US$ million), Gross Domestic Product (GDP) in the source country (in constant 1990 United States dollars), the consumer price index (CPI) in the destination country, transportation costs (in United States dollars), temperature in degrees Celsius (⁰C) and precipitation (millimetres).

Tourists prefer to visit a country with a high per capita income, since it translates into a higher standard of accommodation and better tourist facilities; they also prefer visiting countries where the poverty level is low.

In tourism demand functions, income in the source country is included as a key explanatory variable. Since travel is expensive and considered a luxury good, it is anticipated that high-income countries would generate a higher number of travellers. Although per capita income is used in some studies, a more general income variable (GDP) is employed in this study, since tourist arrivals include both business and holiday visitors.

Many tourism demand studies employ the consumer price index (CPI) of the destination country to reflect the relative prices of foreign goods and services that tourists purchase in the destination country. These relative prices are the costs of goods and services that tourists would pay for items such as accommodation, food, entertainment, and local transportation.

Transportation costs, usually measured by the cost of a return airline ticket between the source country and the destination country, have been used in many tourism demand studies. Other studies have used various proxies for the transportation cost variable, such as the cost of gasoline for a return flight between the source country and major destination countries. In this study, oil prices are used to proxy travel costs, due to the unavailability of travel cost data over the sample period. It is expected that these two variables would be highly correlated.

A priori, both income variables (GDPPC in the destination country and GDP in the source country) are expected to be positively associated with tourism demand. It is anticipated that the CPI variable, oil prices and the two climate variables would have a negative relationship with tourism demand.

Data


209

Several sources were used to collect the data used in the study. Information on tourist arrivals was obtained from the Caribbean Tourism Organization (Sean Smith, Statistical Specialist, pers. comm). The income variables and the CPI were collected from the International Financial Statistics website (http://www.imfstatistics.org/imf/). Oil prices were obtained from the InflationData.com website (http://www.inflationdata.com) and the two climate variables (temperature and precipitation) were obtained from the Department of Geography Center for Climatic Research of the University of Delaware.

Methodology and empirical model

This section outlines the economic framework and methodology used in the paper. Several statistical techniques have been employed to estimate the demand for tourism and to forecast such demand (see Lim, (1999) for a comprehensive review of the various techniques used). Similar to Johnson and Ashworth (1990), Song and Witt (2000) and Bigano and others (2006), a tourism demand model is used to determine the variables that affect tourism demand in the Caribbean.

Where, TAit is the total tourist arrivals from origin country i in period t

is Gross Domestic Product for origin countries

PCit is the per capita income in the destination country

Cpiit is the consumer price index in the destination country

opt is the price of oil

tt is the temperature

pt is the precipitation

Data were collected from nine Caribbean countries on tourist arrivals, GDP per capita, GDP of the source countries (in this case most of the tourists came from either the United States of America or the United Kingdom), the CPI, oil prices, temperature and precipitation. The data can be categorized as panel data since the same information was collected for the nine countries across the period 1989 to 2007. The panel can be defined as strongly balanced.136 Panel data permits the estimation of a richer set of models than those that employ only time, since advantage can be taken of both cross-sectional and temporal variations in the data. However, the estimations also become more complicated, since there is now more heterogeneity in the data. According to Greene (2008, p.183) “… the crucial distinction between fixed and random effects is whether the unobserved individual effect embodies elements that are correlated with the regressors in the model, not whether these effects are stochastic or not.”

The Hausman test was used to determine whether or not the model should be one that takes into consideration fixed effects or random effects of the data. The results indicate that the random effects model should be employed, in other words, an insignificant p-value was obtained (p-value > 0.05). The random effects model is employed, which uses a weighted average of the between- and within- variation in the data, and can accommodate within-unit serial correlation.

The random effects panel data model of tourism demand (log-log specification) employed in this study is assumed to take the following form:

136 A strongly balanced panel dataset is one in which each panel has the same number of observations and the observations for different panels are all made at the same time. (Adapted from definition at: http://www.stata.com/help.cgi?xt_glossary#strongly_balanced).


210

where, is tourist arrivals in the nine destination Caribbean countries (i = country; t =

time)

is the coefficient for the six independent variables

is per capita income for the destination country

is Gross Domestic Product for origin countries

is the Consumer Price Index for the destination countries

is the price of oil (proxy for travel cost)

is the temperature in the destination countries

is the precipitation in the destination countries

is the intercept

is the combined time series and cross-section error components

is the cross-section, or country-specific error component

The double log model, reported by Lim (1999), is one of the more popular model specifications. The variables (in level) used in the model are summarized in table 2. The data indicate that, in the present sample of nine Caribbean countries, the average annual number of tourist arrivals over the period 1989 to 2007 was 166,600 persons. The average price of a barrel of oil was US$ 33.27. The temperature variable is significant: the annual mean temperature experienced is 26⁰C. The study by Lise and Tol (2002) found that the optimal temperature for comfort ranged between 21⁰C and 24⁰C, with the optimal temperature for tourists from the countries of major interest to the

Caribbean (the United States of America or the United Kingdom) being approximately 23⁰C. The

average temperature of the Caribbean is a solid 3⁰C higher than the optimal temperature for tourism of its major source countries.

TABLE 2

SUMMARY OF CORE VARIABLES USED IN THE REGRESSION ANALYSIS

Variable Observations Mean Std. Dev. Minimum Maximum Tourist arrivals 171 1666 3751 7 21285 Per capita income in destination country 171 7133.637 5878.714 477 25253

GDP in source country 171 5,298,579 3,074,121 982,323 9,393,837

CPI in destination country 170 95.68012 36.46744 8.97 263.11

Oil prices 171 33.27053 13.17715 15.52 64.93

Annual mean temperature 161 25.99453 1.407047 21.89 28.22

Annual mean precipitation 162 141.9133 84.14548 24.7 493.45

The results of the model are outlined and analysed in the following section.


211

RESULTS

The results for the double-log random effects model specification are provided in table 3. Robust standard errors were specified to control for heteroskedasticity. The results of the F-test confirm that all the coefficients in the model are different from zero.

TABLE 3

ESTIMATION RESULTS FOR RANDOM EFFECTS MODEL

Variables Tourist Arrivals Per capita income in destination country 0.18718* (1.831) Gross domestic product in origin country -0.62628*** (2.688)

Consumer price index in destination country 0.31431* (1.652)

Oil prices -0.53791 (1.557)

Annual mean temperature -6.61282*** (4.153)

Annual mean precipitation -0.54254*** (4.446) Constant 38.78618*** (8.683) Wald test Chi squared value (probability in parentheses)

108.33 (0.0000)

Observations 160 Robust t-statistics in parentheses * significant at 10%; ** significant at 5%; *** significant at 1%

An examination of the results indicates that the coefficient estimates were generally in agreement with expectations and, of importance, the results obtained for the climate variables were highly significant.

It was found, however, that as the Gross Domestic Product in the origin countries decreases, tourist arrivals increased to the Caribbean subregion. A straightforward explanation for this could be that as income decreases in the origin countries (in this case, the United States and the United Kingdom), it becomes more affordable to visit the Caribbean rather than other, more expensive alternatives, for example Europe or Asia. Table 4 shows the top 10 tourist destinations for 2007. As a matter of note, the Caribbean is not even mentioned in the top 50 tourist destinations.

TABLE 4

TOP TEN TOURIST DESTINATIONS FOR 2007

Country Tourist Arrivals (millions)

France 81.9 Spain 59.2 United States 56.0 China 54.7 Italy 43.7 United Kingdom 30.7 Germany 24.4 Ukraine 23.1 Turkey 22.2 Mexico 21.4

Source: UNWTO World Tourism Barometer 4(2), 2008

It was also found that, as the CPI in the destination country increases, so do tourist arrivals. Again, this could be because the Caribbean, even with increasing prices, provides a still cheaper alternative to other tourist destinations.


212

The expected results were received for the two climate variables: essentially, as temperatures increase, tourist arrivals decrease. Similarly, as rainfall increases, tourist arrivals decrease. This has grim implications for tourist arrivals to the Caribbean due the predicted increases in temperature under the various weather scenarios put forward by the Intergovernmental Panel on Climate Change (IPCC) in their Special Report on Emissions Scenarios (SRES). The results of the model indicate that precipitation is expected to affect tourism to a much smaller degree than temperature (the model yielded a temperature coefficient of -6.61, whereas the precipitation coefficient was -0.54254). Furthermore, according to the IPCC predictions, precipitation is expected to decline in certain Caribbean countries. The literature on tourism demand has pointed to the fact that tourists prefer dry holiday locations rather than wet ones (Lise and Tole, 2002), therefore, according to the results of the model, as the climate changes in some countries and less precipitation is observed, the impact on tourism would be positive.

FORECASTING THE COST OF CLIMATE CHANGE

Forecasted climate change impacts at nine tourist destinations in the Caribbean

The estimated tourism demand model satisfied demand theory, passed the various hypothesis tests, and reported significant results. The next stage of the analysis requires that the demand model be used to make forecasts for the rest of this century for the Caribbean countries under study.

The forecasted tourist arrivals data were used to obtain a cost figure until the year 2100, using A2, B2 and Business as Usual (BAU)137 weather scenarios. The tourism industry is very important for most of the countries under analysis and any decline in the industry would have a negative effect on the GDP of those countries. There are many different emission scenarios and table 5 gives a brief explanation of four SRES storylines, which includes the two that are used in this study, namely A2 and B2.

Table 5

SRES storylines used for calculating future greenhouse gas and other pollutant emissions

Storyline Description A1 Very rapid economic growth; population peaks mid-century; social, cultural and

economic convergence among regions; market mechanisms dominate. Subdivisions: A1F1 – reliance on fossil fuels; A1T – reliance on non-fossil fuels; A1B - a balance across all fuel sources

A2 Self reliance; preservation of local identities; continuously increasing population; economic growth on regional scales

B1 Clean and efficient technologies; reduction in material use; global solutions to economic, social and environmental sustainability; improved equity; population peaks mid-century

B2 Local solutions to sustainability; continuously increasing population at a lower rate than in A2; less rapid technological change than in B1 and A1

Source: Table A.2, page 107 of the United Kingdom Climate Impacts Programme UKCIP02 climate scenarios technical report

The A2 scenario envisages that by the year 2100 the population would have reached a figure of 15 billion, with generally slow economic and technological development. It predicts slightly lower greenhouse gas (GHG) emissions than other scenarios. The B2 scenario forecasts a slower population growth of 10.4 billion by 2100, with a rapidly developing economy and greater stress on environmental protection, so producing lower emissions and less future warming

137 The Business as Usual (BAU) weather scenario in this study is simply the average values, for temperature and precipitation, of the A2 and B2 scenarios until the end of the century. There are no forecasted data for the BAU weather scenario for the Caribbean subregion.


213

Bueno and others (2008) undertook an estimation of the cost of climate change in the Caribbean in the absence of action by these countries to counteract the effects of climate change. They estimate both a low-impact scenario and a high-impact scenario for the years 2025, 2050, 2075 and 2100. The low-impact scenario is the optimistic scenario, where the world takes action in the near future and where emissions are significantly reduced by mid-century, and continue to decrease by the end of the century. The high-impact scenario is pessimistic in nature and one in which business-as-usual (BAU) takes place, meaning that greenhouse gas emissions continue to increase drastically throughout the twenty-first century. Table 6 shows an extract of the table presented in their study.138 The data in the table reveal that, under both the high- and low-impact scenarios, all of the Caribbean countries have much to lose in the tourism industry.

TABLE 6

COST OF LOW-IMPACT AND HIGH- IMPACT SCENARIOS FOR TOURISM IN SELECTED

CARIBBEAN COUNTRIES

Low-impact scenario

(US$ billion)

High-impact scenario

(US$ billion)

Country GDP

(US$

billion) 2025 2050 2075 2100 2025 2050 2075 2100

Aruba 2.35 0.02 0.04 0.06 0.08 0.10 0.20 0.30 0.40

Barbados 2.54 0.02 0.03 0.05 0.07 0.09 0.17 0.26 0.35

Dominican Republic 20.52 0.07 0.14 0.21 0.28 0.36 0.71 1.07 1.43

Jamaica 8.77 0.04 0.07 0.11 0.15 0.18 0.37 0.55 0.74

Montserrat 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

The Netherlands Antilles 2.70 0.02 0.04 0.06 0.07 0.09 0.18 0.28 0.37

Saint Lucia 0.70 0.01 0.01 0.02 0.02 0.03 0.05 0.08 0.11

Trinidad and Tobago 12.61 0.01 0.02 0.02 0.03 0.04 0.08 0.12 0.16

Source: Bueno and others (2008)

Using low- and high-impact climate scenarios,139 and examining the impact of rising temperatures in the subregion (for 12 CARICOM countries), a study by Margaree Consultants Limited (2002) suggests that, on an annual basis, for the low-impact scenario the Caribbean stands to lose US$ 715 million in tourist expenditure, while for the high-impact scenario, losses in tourism expenditure amount to US$ 1,430 million. In terms of the cost to tourist facilities due to sea level rise,140 it was determined that, on an annual basis, it would cost US$ 9 million, and US$ 80 million, to replace hotels due to sea level rise under the low- and high-impact weather scenarios, respectively. An evaluation of the loss in tourism income due to the loss of beaches and ecosystems was also carried out in the same study, by examining the fraction of beach area lost in conjunction with the amount that tourists spend on enjoying the ‘sun, sea and sand.’ At an annual rate, the loss would be US$ 550 million in the low-case scenario and US$ 2.4 billion in the high-case scenario. 138 Guyana was not included in the Bueno and others (2008) study. 139 Figures for temperature were based on the Intergovernmental Panel on Climate Change (IPCC) Third

Assessment Report (2001) - an increase of 2°C for the low-impact scenario and an increase of 3.3°C for the high-impact scenario.

140 According to the estimates by Margaree Consultants Limited (2002), the sea level is expected to rise between 0.5 metres (low impact scenario) and 2.0 metres (high impact scenario) by the year 2100.


214

Projections for extreme events

In this section of the paper, projections regarding extreme events are made using three weather scenarios: A2, B2 and Business as Usual (BAU).

Methodology

A tourism demand model was used to estimate the costs of temperature and precipitation to the tourism sector in the nine Caribbean countries. This model and results of the model as relate to the effects of temperature and precipitation were outlined in section 4 of this paper. Costing of the climate change effects required forecasting of the variables in the model: per capita income and CPI of the Caribbean countries included in the study, GDP of the source or origin countries (primarily the United States and United Kingdom), oil prices, and the temperature and precipitation of the nine countries. Forecasts up to the year 2100 of these non-climatic variables are not available for any of the nine countries. Therefore, forecasting techniques were employed to estimate these variables. With respect to the climate variables, forecasts for both variables were received from the Institute of Meteorology in Cuba (INSMET). The predictions from INSMET were obtained from the European Centre Hamburg Model (ECHAM), an atmospheric general circulation model developed at the Max Planck Institute for Meteorology. The annual costs of climate change impacts to 2100 are estimated in United States dollars, using 2007 as the comparator and base year. This method is frequently used in the literature (for example, see Haites, 2002) and is considered standard.

However, apart from temperature and precipitation and its effects on the tourism sector, there are other climate variables that have the potential to affect the tourism sector negatively, in particular, increases in the occurrence of extreme events, sea level rise, and extreme destruction of ecosystems due to ocean acidification. Due to lack of data, the methodology used for this part of the study was adopted from Toba (2009). These costs do not take into consideration the indirect costs (for example, the loss of employment) associated with changes in temperature and precipitation, extreme events, sea level rise and ecosystem destruction. It must be noted that most of the results obtained from existing research on economic effects of climate change in the Caribbean are not directly comparable to each other and to this study, since many variations exist with respect to the number of countries used in the studies, the sectors examined, and the data and methodologies employed.

Results

Table 7 shows the total costs of climate change due to temperature and precipitation changes for the nine countries under the three climate change weather scenarios. The aggregated costing for the Caribbean subregion is shown at four different points in the century – in the years 2025, 2050, 2075 and 2100.

TABLE 7

AGGREGATED COSTING FOR A2, B2 AND BAU SCENARIOS FOR THE CARIBBEAN

SUBREGION: TEMPERATURE AND PRECIPITATION

(Costs in $US million - 2007 dollars)

Year A2 B2 BAU

2025 26.4 29.1 27.8

2050 111.3 116.3 115.5

2075 203.0 209.1 209.7

2100 283.9 292.7 291.8


215

The table shows that, under all three scenarios, the countries under study are poised to lose extensively from the predicted changes in temperature and precipitation under the three weather scenarios. As shown in table 7, in total, by the year 2100, the costs to the nine countries due to changes in temperature and precipitation are estimated to be US$ 283.9 million under the A2 scenario, US$ 292.7 million under the B2 scenario and US$ 291.8 million under the BAU scenario.

Table 8 provides costing under the three weather scenarios for extreme events. This includes losses due to increases in the frequency and intensity of hurricanes and the accompanying windstorms, floods and landslides. Similar to Toba (2009), and as reported by Haites and others (2002), hurricanes in the Caribbean are expected to increase by 27% per annum. Haites and others (2002) used the example of 1995 hurricanes (Hurricane Luis and Hurricane Marilyn) to determine the cost in terms of income loss from the tourism sector, and found that tourism expenditures decreased by about 17%. Therefore, with a 27% increase in hurricanes due to climate change and an estimated 17% decrease in tourist expenditures when a hurricane strikes, it is estimated that tourist expenditures are expected to decrease by 21.6% due to increases in extreme events. The costs to the countries under consideration with regard to extreme events are shown in table 8.

TABLE 8

AGGREGATED COSTING FOR A2, B2 AND BAU SCENARIOS: EXTREME EVENTS


Year A2 B2 BAU

2025 9,896.6 10,057.4 10,155.9

2050 15,503.5 14,950.9 15,982.7

2075 18,394.2 17,797.4 19,181.4

2100 18,466.9 17,870.4 19,255.2

The table indicates that for extreme events occurring in the Caribbean subregion, the total costs under the scenarios for extreme events, aggregated to 2100, are as follows: A2: US$ 18.5 billion, B2: US$ 17.9 billion and BAU: US$ 19.3 billion.

TABLE 9

AGGREGATED COSTING FOR A2, B2 AND BAU SCENARIOS: SEA -LEVEL RISE AND

DESTRUCTION OF ECOSYSTEMS


Year A2 B2 BAU

2025 13,745.2 13,968.6 14,094.8

2050 21,532.7 21,654.6 22,185.8

2075 25,547.5 25,608.2 26,628.4

2100 25,648.5 25,709.5 26,731.0

Table 9 presents the loss to the tourism sector due to the predicted rise in sea level and the destruction of ecosystems due to occurrences such as ocean acidification. Once more, similar to Toba (2009), it is assumed that tourists spend about 30% of their total expenditure on activities related to the sea. With rising sea levels and ecosystem destruction produced by climate change, it is assumed that this amount would be lost due to non-participation in these activities. The costs calculated in table 9 represent the losses that would occur when tourists refrain from sea-related activities. With respect to rising sea levels and the destruction of ecosystems, again the Caribbean subregion is affected negatively throughout the century, culminating by the end of the century in costs under the


216

three weather scenarios as follows: A2: US$ 25.6 billion, B2: US$ 25.7 billion and BAU: US$ 26.7 billion.

A summation of all costs incurred by the tourism sector due to temperature and precipitation change, extreme events and sea level rise and ecosystem destruction under the three weather scenarios is shown in table 10. The figures in table 10 indicate that the tourism sector in the Caribbean countries under examination stands to incur losses in the sums of US$ 44.4 billion (A2), US$ 43.9 billion (B2) and US$ 46.3 billion (BAU) (2007 dollars) in total by the end of the century.

Table 10

Total costs incurred by the Caribbean subregion under A2, B2 and BAU scenarios (Costs in $US million - 2007 dollars)

Year A2 B2 BAU

2025 23,668.4 24,055.2 24,278.6

2050 37,147.7 36,721.8 38,283.9

2075 44,144.8 43,614.6 46,019.7

2100 44,399.5 43,872.6 46,278.7

ADAPTATION AND MITIGATION STRATEGIES

Adaptation Strategies

The UNWTO-UNEP-WMO (2008, p.81) defines adaptation as “….. an adjustment, in natural or human systems in response to actual or expected climatic stimuli or their effects, which moderates harm or exploits beneficial opportunities.” There is little doubt that the tourism sector will be unable to adopt adaptation strategies to cope with changes in climate. UNWTO-UNEP-WMO (2008) maintains that the tourism industry is dynamic and flexible enough to implement measures of an adaptive capacity to deal with climate change. For instance, this is an industry that has had various recent jolts (such as terrorism and severe acute respiratory syndrome (SARS)) and has shown an ability to cope. However, the changing climate must be recognised as such and strategies must be adopted before it is too late. Table 11 shows the projected changes in the temperature and precipitation variables for the period 2010 to 2099.

Table 11 Projected increases in air temperature and changes in precipitation for small island regions

(Percentage) Forecast period Region

2010 to 2039 2040 to 2069 2070 to 2099

(a) Projected increases in air temperature

Mediterranean 0.60 to 2.19 0.81 to 3.85 1.20 to 7.07

Caribbean 0.48 to 1.06 0.79 to 2.45 0.94 to 4.18

Indian Ocean 0.51 to 0.98 0.84 to 2.10 1.05 to 3.77

Northern Pacific 0.49 to 1.13 0.81 to 2.48 1.00 to 4.17

Southern Pacific 0.45 to 0.82 0.80 to 1.79 0.99 to 3.11

(b) Projected changes in precipitation (percentage)

Mediterranean -35.6 to 55.1 - 52.6 to 38.3 -61.0 to 6.2

Caribbean -14.2 to 13.7 -36.3 to 34.2 -49.3 to 28.9

Indian Ocean -5.4 to 6.0 -6.9 to 12.4 -9.8 to 14.7


217

Northern Pacific -6.3 to 9.1 -19.2 to 21.3 -2.7 to 25.8

Southern Pacific -3.9 to 3.4 -8.23 to 6.7 -14.0 to 14.6

Source: Becken and Hay (2007), Tourism and Climate Change

TABLE 12 TOURISM STAKEHOLDERS AND CLIMATE ADAPTATION STRATEGIES

Type of adaptation

Tourism operators/businesses Tourism industry associations

Governments and communities

Financial sector (investors/insurances)

Technical ‐ Snow-making ‐ Slope contouring Rainwater collection and

water recycling systems ‐ Cyclone-proof building

design and structure

‐ Enable access to early-warning equipment (such as radio) operators.

‐ Develop websites with information on adaptation measures.

‐ Reservoirs and desalination plants

‐ Fee structures for water consumption

‐ Weather forecasting and early warning systems

‐ Require advanced building design or material (fire resistant) standards for insurance

‐ Provide information material to customers

‐ Water conservation plans ‐ Low season closures ‐ Product and market

diversification ‐ Regional diversification

in business operations ‐ Redirect clients away

from impacted destinations

‐ Snow-condition reports through the media

‐ Use of short-term seasonal forecasts for planning marketing activities.

‐ Training programmes on climate change adaptation.

‐ Encourage environmental management with firms (such as via certification)

‐ Impact management plans (for example, the Coral Bleaching Response Plan)

‐ Convention/event interruption insurance

‐ Business subsidies (for example, insurance or energy costs)

‐ Adjust insurance premiums or not renew insurance policies

‐ Restrict lending to high risk business operations

Policy ‐ Hurricane interruption guarantees

‐ Comply with regulations (such as building codes)

‐ Coordinated political lobbying for GHG emission reductions and adaptation mainstreaming.

‐ Seek funding to implement adaptation projects

‐ Coastal management plans and setback requirements

‐ Building design standards (such as. for hurricane force winds).

‐ Consideration of climate change in credit risk and project finance assessments

Research ‐ Site location (for example, north facing slopes, higher elevations for ski areas)

‐ Assess awareness of businesses and tourists, as well as knowledge gaps.

‐ Monitoring programmes (such as to predict bleaching or avalanche risk, beach water quality)

‐ Extreme event risk exposure

Education ‐ Water conservation education for employees and guests

‐ Public education campaign

‐ Water conservation campaigns

‐ Campaigns on the dangers of UV radiation

‐ Educate/inform potential and existing customers

Behavioural ‐ Real-time webcams of snow conditions

‐ GHG-emission offset

‐ GHG-emission offset programmes

‐ Extreme event recovery marketing

‐ Good practice in-house.


218

programmes ‐ Water conservation initiatives

Source: World Trade Organization (2008) As demonstrated in table 12, different types of adaptation strategies are utilized by tourism

stakeholders ranging from the technological to the behavioural. These approaches are presently being used to cope with changes in climate in the destination country. As explained by WTO, these approaches are rarely used singly, but are usually combined to deal with the specific climate variation being experienced.

TABLE 13 POSSIBLE ADAPTATION MEASURES FOR TOURISM IN SMALL ISLAND COUNTRIES

AND BARRIERS TO IMPLEMENTATION

Adaptation measures Relevance to tourism Barriers to implementation

Measures to remove barriers

Mainstreaming adaptation in planning

Currently adaptation is not mainstreamed in tourism planning

Lack of information on which to base policy initiatives

Improve targeted information, for example, climate-risk profile for tourism

Include climate risk in tourism regulations, codes

Currently such risks are not reflected in tourism-related regulations

Lack of information on which to base regulatory strengthening

Improve information, such as climate-risk profile for tourism

Institutional strengthening Shortfall in institutional capacity to coordinate climate responses across tourism-related sectors

Lack of clarity as to the institutional strengthening required to improve sustainability of tourism

Assess options and implement the most appropriate strategies

Education/awareness raising Need to motivate and mobilise tourism staff and also tourists

Lack of education and resources that support behavioural change

Undertake education./awareness programmes

Shade provision and crop diversification

Additional shade increases tourist comfort

Lack of awareness of growing heat stress for people and crops

Identify, evaluate and implement measures to reduce heat stress

Reduce tourism pressures on coral

Reefs are a major tourist attraction

Reducing pressures without degrading tourist experience

Improve off-island tourism waste management

Reduce tourism pressures on other marine resources

Increased productivity of marine resources increases well-being of tourism-dependent communities

Unsustainable harvesting practices and lack of enforcement of regulations and laws

Strengthen community-based management of marine resources, including land-based issues

‘Soft’ coastal protection Many valuable tourism assets at growing risk from coastal erosion

Lack of credible options that have been demonstrated and accepted

Demonstration of protection for tourism assets and communities

Improved insurance cover Growing likelihood that tourists and operators will make insurance claims

Lack of access to affordable insurance

Ensure insurance sector is aware of actual risk levels and adjust premiums

Desalination, rainwater catchments and storage

Tourist resorts are major consumers of fresh water

Lack of information on future security of freshwater supplies

Provide and ensure utilization of targeted information, based on climate risk profile.

Drainage and pumping systems

Important services for tourist resorts and for tourism-dependent communities

Wasteful practices; lack of information to design adequate systems

Provide and ensure utilization of targeted information, based on climate risk profile.

Enhanced design and siting standards

Many valuable tourism assets at growing risk from climate extremes

Lack of information needed to strengthen design and siting standards.

Provide and ensure utilization of targeted information.

Tourism activity/product diversification

Need to reduce dependency of tourism on ‘sun, sea and sand’

Lack of credible alternatives that have been demonstrated and accepted

Identify and evaluate alternative activities and demonstrate their feasibility.


219

Source: Becken and Hay (2007), Tourism and Climate Change

The literature on adaptation strategies shows a wide range of measures that Caribbean countries could adopt. Adoption of such measures would depend on the different climate change impacts due to factors such as increasing temperatures, changes in precipitation, increasing intensity of hurricanes and other extreme events, and sea level rise. There is a range of climate change adaptation strategies that Caribbean countries could utilize to tackle the varying effects of climate change. Becken and Hay (2007) outline some possible adaptation measures, along with the barriers to implementation in small island countries (see table 13).

The Stern Review (2006) has emphasized that it is more cost-effective to implement techniques that are proactive rather than reactive, and to support no-regrets measures. In the event that there is no major change in climate, the proactive, no-regrets strategies will still be valuable and economical. As an example, the literature on climate change risk assessment of tourism operators (Elsasser and Burki, 2002; Scott and others 2002; Becken, 2004) has revealed that they have minimal knowledge of climate change and that there is a subsequent lack of long-term planning in the event of future climate changes. This indicates that there is an urgent need to educate and ensure that tourism policymakers, who formulate policies for both the private and public sectors, are aware that the climate is changing and that the tourism industry has to adapt to the change or face decline.

An estimation of the cost of adaptation is a complex one which depends significantly on the determinants of the adaptive capacity of countries in the Caribbean subregion. The IPCC (2001), drawing from Smit and others (1999), categorized determinants of adaptive capacity, including issues such as the availability of technological resources, the organization of essential institutional and decision-making bodies, the stock of human and social capital, information management, and public perception.

Mitigation Strategies

Rogner and others (2007) have asserted that adaptation and mitigation can complement each other, act as substitutes for, or be independent of one another. A discussion of mitigation measures to cope with climate change must of necessity include technological, economic and social changes and substitutions that can be employed to attain a reduction in GHG emissions (UNWTO-UNEP-WMO 2008; Hall and Williams, 2008). The IPCC report has asserted that human activity has been a major contributor to climate change, which may have started as early as the mid-1700s. Carbon dioxide (CO2) emission is just one of many GHG emissions. However, CO2 emissions become important when released in large quantities, as can happen due to human activity such as in the burning of solid waste, wood and wood products, and fossil fuels such as oil, natural gas, and coal. Figure 1 shows the per capita CO2 emissions (metric tons of carbon) for the nine Caribbean countries under study in this paper. The data reveal that Aruba, the Netherlands Antilles and Trinidad and Tobago have the highest per capita CO2 emissions among the countries under study.

A framework for mitigation strategies must outline the mitigation tools and techniques, policies and measures which go along with the various climate change scenarios. In addition to the mitigation process, the potential exists for approaches in the areas of transport and accommodation that tour operators, consumers, and destination countries can take to cope with – and perhaps alter – the path of climate change.


220

FIGURE 1

PER CAPITA CO2 EMISSIONS (METRIC TONS OF CARBON) FOR SELECTED CARIBBEAN COUNTRIES (1950-2005)

Data Source: USAID Development Statistics for Latin America and the Caribbean

UNWTO-UNEP-WMO (2008) has outlined four key mitigation measures that can be used to deal with GHG emissions from tourism:

(1) Reducing energy use/focusing on energy conservation: by changing transport behaviour (for example, using more public transport, shifting to rail and coach instead of car and aircraft, choosing less distant destinations), and by changing management practices, for example by introducing videoconferencing for business tourism.

(2) Improving energy efficiency: the use of new and innovative technology to decrease energy demand (usually by carrying out the same operation with a lower energy input).

(3) Increasing the use of renewable or carbon neutral energy: substituting fossil fuels with energy sources that are not finite and that cause lower emissions, such as biomass, and hydro-, wind- and solar energy.

(4) Sequestering CO2 through carbon sinks: CO2 can be stored in biomass (for example, through afforestation and deforestation), in aquifers or oceans, and in geological sinks (such as depleted gas fields). Indirectly, this option can have relevance to the tourism sector, considering that most developing countries and small island developing States (SIDS) that rely on air transport for their tourism-driven economies are biodiversity-rich areas with important biomass CO2 storage function. Environmentally-oriented tourism can play a key role in the conservation of these natural areas.

An estimation of the costs associated with mitigative action (for example, abatement costs) depends critically on the potential of the tourism sector to implement processes associated with mitigation. Auton (2008) summarized four phases that the tourism sector can implement to mitigate tourism sector impact on climate, as follows:

(1) The elimination GHG emissions by circumventing activities that can be avoided without significant change to the tourist experience.

(2) The reduction of GHG emissions by focusing on energy efficiency.

(3) The substitution of practices that accounts for large GHG emissions with practices that have a lower carbon footprint.


221

(4) The offsetting of any remaining emissions so as to achieve full carbon neutrality.

Some of these mitigation actions can include measures such as:

• use of fuel-efficient cars by tour operators, hotels and resorts

• employing channels such as effective ventilation and appropriate roofing to reduce temperature in buildings

• new buildings designed with the conservation of energy in mind

• the use of more efficient lighting and energy-efficient appliances (less electricity use).

CONCLUSION

In this study, tourism demand was modelled for nine Caribbean countries. The model was then used to predict the impact of climate change on the tourism sector under three weather scenarios (A2, B2 and BAU) until the end of the century. The results of this research provide proof that the tourism sector of the Caribbean subregion would be affected profoundly by climate change. It is therefore very important that Caribbean countries adapt and mitigate against impending climate change to promote and sustain growth in the tourism sector. To undertake this task, governments in the subregion must come together to formulate policies which would ensure that the sector remains sustainable.

The estimation of costs undertaken in this study would have benefited greatly from country-specific data for extreme events and sea level rise due to climate change. Until such data is available, costing figures from existing studies would have to be used to ensure that climate variables are taken into consideration and included in any study that examines the impact of climate change on the tourism sector in the subregion.

It is essential that Governments and policymakers become involved in the process, perhaps at a very early stage, since climate change is an environmental, and ultimately a developmental problem. One clear advantage of participation by these stakeholders is the recognition and acknowledgement that for researchers to generate more accurate estimation of costs, data must be made available to these practitioners in the field.

It is necessary that further work in this area involve a thorough investigation of adaptation and mitigation strategies and the costs of implementing of such strategies in the Caribbean subregion. With the formulation of mitigation and adaptation strategies and the appropriate policies in place, the tourism sector can play a key role in dealing with climate change and encouraging sustainable growth in the sector.


222

Bibliography

Auton, G.P. (2008), “Mitigating risk in adaptation strategies for climate change and sustainability in the tourism industry: a global perspective”, presented at the Jeju Forum for Global Sustainability, Korea.

Available at: http://74.125.93.132/search?q=cache:JND55K5Q808J:www.icleikorea.or.kr/board/bbs/download.php%3Fbo_table%3Ddoc_data%26wr_id%3D5%26no%3D1+Mitigating+Risk+in+Adaptation+Strategies+for+Climate+Change+and+Sustainability+in+the+Tourism+Industry:+A+Global+Perspective&cd=1&hl=en&ct=clnk

Becken, S. and J. Hay (2007), Tourism and Climate Change – Risks and Opportunities, Channel View Publications, Cleveland, Ohio, United States of America.

Becken, S. (2004), “How tourists and tourism experts perceive climate change and forest carbon sinks”, Journal of Sustainable Tourism 12 (4): 332-345.

Berrittella, M. and others (2006), “A General Equilibrium analysis of climate change impacts on tourism”, Tourism Management 27: 913-924.

Bigano, A. and others (2006), “Economy-wide estimates of the implications of climate change” in A joint analysis for sea level rise and tourism, Nota di lavaro 135.06, Fondazione Eni Enrico Mattei.

Braun, O.L. and others (1999), “Potential impacts of climate change effects on preferences for tourism destinations. A psychological pilot study”, Climate Research 11: 247–254.

Bueno, R. and others (2008), “The Caribbean and climate change: the cost of inaction”, Stockholm Environment Institute (US Center) and Global Development and Environment Institute (Tufts University).

Crouch, G.I. and R.N. Shaw, (1992), “International tourism demand: A meta-analysis integration of research findings”, in Choice and Demand in Tourism, Ed. P. Johnson and B. Thomas, pp. 175-207, Mansell, London.

Elsasser, H. and R. Burki (2002), “Climate change as a threat to tourism in the Alps”, Climate Research 20: 253-257.

Greene, W.H. (2008), Econometric Analysis (6th ed.), Prentice Hall, U.S.A.

Hall, C.M. and A.M. Williams (2008), Tourism and Innovation, London, Routledge.

Haites, E. and others (2002), “Assessment of the impact of climate change on CARICOM countries”, Environmentally and Socially Sustainable Development – Latin America and Caribbean Region, The World Bank, available at http://www.margaree.ca/reports/ClimateChangeCARICOM.pdf

Hamilton, J.M. and R.S.J. Tol (2007), “The impact of climate change on recreation and tourism”, Working Paper FNU52, Hamburg University Centre for Marine and Climate Research, Hamburg.

Hamilton, J.M. and M.A. Lau (2004), “The role of climate information in tourism destination choice”, Working Paper FNU56, Hamburg University Centre for Marine and Climate Research, Hamburg.


223

Hamilton, J.M., D.J. Maddison and R.S.J. Tol (2005),” Effects of climate change on international tourism”, Climate Research 29: 245-254.

Intergovernmental Panel on Climate Change (IPCC) (2001), Climate change 2001: Synthesis Report, Intergovernmental Panel on Climate Change, Geneva.

Johnson, P. and J. Ashworth (1990), “Modelling tourism demand: A summary review”, Leisure Studies, 9 (2): 145-161.

Koenig, U. and B. Abegg (1997), “Impacts of climate change on winter tourism in the Swiss Alps”, Journal of Sustainable Tourism 5 (1): 46 – 58.

Lim, C. (1999), “A meta-analysis review of international tourism demand”, Journal of Travel Research 37: 273-284.

Lise, W. and R.S.J. Tol (2002), “Impact of climate on tourist demand”, Climate Change 55: 429-449.

Mather, S., D.Viner and G.Todd (2005), “Climate and policy changes: their implications for international tourism flows”, in M.C. Hall and J. Higham (Eds), Tourism, Recreation and Climate Change, Channel View Publications, Great Britain.

Mimura, N. and others (2007), “Small islands. Climate change 2007: impacts, adaptation and vulnerability”, Contributions of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, M.L. Parry, O.F. Canziani, J.P. Palutikof, P.J. van der Linden and C.E. Hanson, (eds.), pp. 786-716, Cambridge University Press, Cambridge, United Kingdom.

Rogner, H. and others (2007), “Introduction”, in Climate Change 2007: Mitigation, contribution of Working Group III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, B. Metz, O.R. Davidson, P.R. Bosch, R. Dave, L.A. Meyer (eds.), Cambridge University Press, Cambridge, United Kingdom..

Scott, D. and G. McBoyle (2006), “Climate change adaptation in the ski industry”, Mitigation and Adaptation Strategies to Global Change 12: 1411-1431.

Scott, D., G. McBoyle and M. Schwartzentruber (2004), “Climate change and the distribution of climatic resources for tourism in North America”, Climate Research 27 (2) pp. 105-117.

Scott, D. and others (2002), “The vulnerability of winter recreation to climate change in Ontario’s Lakelands tourism region”, Department of Geography Publication Series Occasional Paper 18, University of Waterloo, Waterloo, Ontario.

Smit, B. and others (1999), “The science of adaptation: a framework for assessment”, Mitigation and Adaptation Strategies for Global Change, 4: 199 – 213.

Song, H. and S.F. Witt (2000), “Tourism demand modelling and forecasting”, Modern Econometric Approaches, Oxford, Pergamon.

Stern, N. (2006), “The economics of climate change”, The Stern Review, Cambridge University Press, Cambridge, United Kingdom.

Toba, N. (2009), “Potential economic impacts of climate change in the Caribbean Community”, in W. Vergara (ed), Assessing the potential consequences of climate destabilization in Latin America, Latin America and Caribbean Region Sustainable Development Working paper 32, The World Bank, Washington, D.C.


224

Trenberth, K.E. and others (2007), “Observations: surface and atmospheric climate change”, Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, pp. 235-336, S. Solomon, D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M. Tignor and H.L. Miller, (eds.), Cambridge University Press, Cambridge, United Kingdom.

United Nations World Trade Organization (UNWTO), United Nations Environment Programme (UNEP) and World Meteorological Organization (WMO) (2008), “Climate change and tourism: responding to global challenges”, (prepared by Scott, D. and others), UNWTO, Madrid, and UNEP, Paris.

Uyarra, M.C. and others (2005), “Island-specific preferences of tourists for environmental features: implications of climate change for tourism-dependent States”, Environmental Conservation 32 (1): pp 11–19.


225

Is the payoff to overeducation smaller for Caribbean immigrants? Evidence from

hierarchical models in the United States labour market

Dillon Alleyne

ABSTRACT

This paper examines the overeducation/undereducation and required (OUR) returns to education hypothesis among Caribbean immigrants in the labour market of the United States of America using the IPUMS 5% sample.140 The results show that Caribbean immigrants in the United States receive lower returns than the native born despite being in the United States for some time. The methodology employed is somewhat different to the traditional approach, as a hierarchical model is used to account for the nesting of individuals within occupation, or for the presence of fixed effects due to differences in occupation.

The results suggest that overeducation, though common to both the native born and to Caribbean immigrants, is rewarded less among immigrants and represents an underutilization of resources. It was also observed that Caribbean immigrants have higher levels of undereducation relative to the native born.

The paper suggests ways in which immigration policies could be crafted to improve the return to education of immigrants who bring considerable pre-immigration experience to the labour market, to create win-win situations for both sending and receiving countries. With respect to sending countries, it is argued that improved returns to overeducation add to immigrant wealth which allows those who wish to return to the Caribbean, often with improved skills and expertise, to do so much

140 I thank Robin Kraft at the Centre for Global Development, Washington DC, for invaluable research assistance. I also wish to thank colleagues in the Department of Economics, University of the West Indies, Mona, Jamaica, Karoline Schmidt of the ECLAC Subregional Headquarters, Port of Spain and an anonymous referee for helpful comments.


226

earlier. Secondly, this translates to a possible increased flow of remittances and other resources to the Caribbean. For receiving countries, improved returns to immigrant overeducation reduce the misallocation of resources at the level of the labour market.

INTRODUCTION

There has been significant movement of labour globally despite rigidities imposed through immigration laws and regulations, as workers move to host countries in anticipation of improved economic opportunities. In the case of the Caribbean, there has been significant emigration since the 1970s, especially to the United States of America. Official statistics show that there were some 1.2 million immigrants to the United States from the Caribbean by 1980, 1.9 million by 1990, and 2.95 million by 2000.141 The Caribbean- born accounted for almost 10% of the 31.1 million foreign-born population by 2000, and Cubans accounted for 29% of the overall Caribbean-born population in 2000, followed by the Dominican Republic accounting for 23.3 %, Jamaica for 18.8%, and Haiti for 14.2%. The Caribbean diaspora in the United States of America is varied, with migrants coming from the English-, Spanish-, Dutch- and French-speaking Caribbean, and with major concentrations in Florida, New York, New Jersey, Massachusetts and California. Despite a relatively long period of migration, with well-established networks, there is anecdotal evidence of overeducation among Caribbean migrants in the United States labour market.

Overeducation occurs when employed individuals have more education than is required for the job. This is a form of underemployment or labour underutilization, which imposes costs both on the individual and on the economy. In an environment in which there is increasing competition for skilled labour (Alleyne, 2008), barriers to the full utilization of skills of migrants represent a waste of resources and a limitation on the ability of migrants to reach their goals quickly and efficiently.

This paper investigates the extent of overeducation and undereducation among Caribbean immigrants to the United States relative to native workers. It also points to the factors which explain such a phenomenon, and seeks to identify policy initiatives that might be pursued based on the assumption that overeducation is a major problem for immigrants.142 While some work has already been done by Chiswick and Miller (2005) in examining overeducation among Caribbean migrants, the focus has been on a broader set of countries which incorporated the Caribbean.

This paper employs the large Integrated Public Use Microdata Series data set (IPUMS 5 % USA) to examine this issue, and distinguishes between returns to the native born, the Caribbean-born residing in the United States for more than five years, and the Caribbean-born residing in the United States for five years or less. This paper is organized as follows: Section II briefly reviews the relevant literature. Section III discusses various approaches to measuring overeducation. Section IV examines the modelling approaches to the OUR hypothesis versus the standard returns to education model. Section V examines the actual model used and relevant descriptive statistics. Section VI discusses the results from the hierarchical models used and section VII concludes with some policy recommendations.

141 These are the only official statistics. See Campbell Gibson and Kay Jung (2006). 142It is also recognized that overeducation is also a feature of the internal labour market of many industrialized countries. For example, in the United Kingdom, some studies have found the incidence of overeducation to be about 30%, while in the United States of America, it is about 45% (Groot, W and H. Maassen van den Brink, 2000). In the case of Australia, the estimates of overeducation were between 10% and 30% (Green and others, 2005; Linsey, 2005).


227

REVIEW OF THE LITERATURE

There is no general theory of overeducation in the literature but there are a number of explanations, each drawing on the theory of labour market behaviour (Hartog, 2000). An important aspect of this problem is whether overeducation is a temporary or a permanent feature of the labour market, whether it is due to labour market inefficiency, or discrimination, and how it affects individual earnings and productivity. In this section, various explanations for overeducation, among them the human capital theory, the career mobility approach, the job competition approach and the theory of discrimination, are discussed briefly.

THE HUMAN CAPITAL THEORY The explanations build on the basic human capital theory of Becker (1964), Becker and Cheswick (1966) and the particular context in which the phenomenon is observed. The human capital model approach to investment in education assumes that some optimal investment is made with the expectation of positive returns to that education. As a result, earnings increase with the level of education, as the individual is willing to forego earnings and incur education investment costs in the short run, in return for increased future benefits. The model implicitly assumes that all education acquired by a worker is acquired to perform the duties expected in the worker’s job.

Since human capital also consists of labour market experience and firm-level training, workers who are overqualified would be more likely to be less experienced and to have less job training. Thus, in the case of immigrants, overeducation is evidence of a compensation for other kinds of training and skills acquired in the labour market. This may reflect the lack of host country labour market skills which, over time, as immigrants remain in the host country, would be compensated for by increased experience. The prediction is that there should be less overeducation143 when migrants hail from jurisdictions that are similar to that of the host country, in terms of education system, language and the institutional workings of the labour market.144

According to this approach, overeducation reflects a lower quality of formal education obtained by migrants in their home country and therefore, migrants who acquire additional training in the host country should have better matches between skills and education.145 To push the argument further, migrants from countries that have a lower standard of education should have lower matches between job requirements and education level, and with increasing assimilation there should be better matches. In addition, education of immigrants may contain large elements of region-specific skills that are less easily transferrable across national boundaries. Of course, when the cultural and institutional differences are large, the effects are greater.

The human capital model sees overeducation as arising from an increase in educational attainment of workers causing relative wages of highly-skilled workers to fall. Thus, within the human capital framework, overeducation is not a permanent phenomenon. If there is an increase in the number of educated workers relative to labour market requirements, employers can substitute away from poorly-educated to better-educated workers as the wages of educated workers fall. The model predicts a positive relationship between returns to overeducation and earnings.

143 This would mean that the degree of overeducation should be similar to that of host country workers. 144 Caribbean economies are strong exporters of qualified labour and family ties; geographical proximity and the use of the same language make the United States of America and Canada and, to a lesser extent, the United Kingdom, preferred destinations for migrants. 145 Bratsberg and Terell (2002), using 1980 and 1990 United States census data, show that the income effect of education is positively related to indicators of education quality in the home country.


228

THE CAREER MOBILITY APPROACH The career mobility approach is another spin off from human capital theory, in which overeducation is seen as a temporary phenomenon largely experienced by immigrants new to the labour market. According to this approach, newly-arrived immigrants are willing to take up positions below their level of education while acquiring experience and the specific training required for future advancement in the labour market (Linsley, 2005). Of course, this assumes that immigrants can be in occupations which are in line with their training, and that the additional experience acquired helps to advance their careers. While some immigrants can make this leap, others may be stuck in areas totally removed from their professional training and may demonstrate very little advancement along their previous career path. As Nielsen (2008) points out, the validity of the career mobility theory would require that the overeducated worker make the leap to a better-matching job within a reasonable period of time. In fact, many migrants fail to make this transition (Büchel, Felix and A. Mertens, 2000).

THE JOB COMPETITION THEORY The approaches set out so far examine the labour market from the supply side. On the demand side, job competition theory argues that firms compete for highly-productive workers who represent training costs to the firm. Since training costs should be lower for more educated workers, firms match these workers to high-paying jobs. At the same time, if there is a significant increase in the education-level of workers, lower-skilled workers are forced down the queue, to be replaced by more highly-educated workers. Because of the oversupply, educated workers are prepared to accept a mismatch rather than be unemployed. Thus, if overeducation is due to structural factors, skilled employees will continue to invest in education in order to maintain their position in the queue. The implication is that overeducation is not temporary, and causes overinvestment in education.146

THE THEORY OF DISCRIMINATION Theories of discrimination also explain the degree of overeducation, and the difference in returns to education between the native born and immigrants in the labour force. One approach is often termed statistical discrimination, according to which employers have little or no knowledge of the skills and productivity levels of applicants and, therefore, proceed on the basis of notions of ethnicity and gender, to name a few, in making hiring decisions. The idea is that employers feel that such characteristics are correlated with performance. In the face of discrimination, immigrants may find it more difficult to get a job equivalent with their skills and may then prefer to take other jobs. Thus, the expectation is that there ought to be much more overeducation among immigrants relative to native workers. If there is statistical discrimination, this should be reduced over time, as employers receive more information about the productivity of immigrants.

Overeducation may also occur due to higher search- and mobility- costs among immigrants, who may not have much local support and may take time to get relevant labour market information. With time, immigrants would gather such information, thus reducing the monopolistic influence of employers in the labour market. Nonetheless, if employers have a taste for discrimination against immigrants in the labour market, and are willing to maintain such discrimination, then overeducation would persist. The impact of product market competition over time, however, should cause a decline in such practices.

ISSUES ARISING

146 The assignment approach is also set out in the literature, in which both demand- and supply- side factors are examined. It assumes that not all workers with similar education are equally productive, so that there is a mismatch of skills which adjusts over time (Sattinger, 1993).


229

Two issues arise here. The first is that there is no recognized body that properly compares host country skills to those of immigrants, for various levels of education in the United States labour market, except for a few professions. The second is that skills and experience acquired abroad are often not easily transferred to the host country. Thus, imperfect information and the unwillingness to incur search costs by employers may be important factors in explaining relative overeducation among immigrants. The current immigration policy of many developed countries also favours immigrants with host-country training over those who have been trained abroad, especially in developing countries.147 Whether there are greater mismatches among Caribbean-born workers versus native-born workers in the United States, is an empirical question. If there is a mismatch due to quality considerations, then one should see lower mismatches among Caribbean migrants who have acquired some education in the United States, relative to recent arrivals. These several explanations, though useful, offer only a partial explanation of the reasons for overeducation, and their predictions are not consistent across explanations.

MEASURING OVEREDUCATION IN THE MODIFIED HUMAN CAPITAL MODEL

The view of education in the labour market along the lines of Becker (1964) and Mincer (1974) has come under challenge by proponents of the Overeducated/Undereducated/Required (OUR) hypothesis. Under this approach, each job is perceived as having a required level of education and training that is necessary for its satisfactory completion. At the same time, in each job there are workers with levels of education greater or less than that reference level. The former are called overeducated and the latter are called undereducated. While the concepts are straightforward, the measurement issues are not.

There are three approaches to measuring these, which are: job analysis (JA), worker self-assessment (WSA), and realized matches (RM) procedures. Job analysis uses evaluations of the required level of education for the job titles in an occupational classification by job analysts, and compares these with the actual educational attainment of the worker. Despite its apparent objectivity in using and evaluating job classifications, job analysis has several shortcomings (Hartog, 2000). Usually, occupational categories in the regular data sets are broad, giving rise to heterogeneity among workers who are in the same occupational category. In many instances, job classifications may be dated, providing a mismatch between such classification and workers’ educational level. Nielsen (2007, p.13) also argues that it may be difficult to translate job requirements into years of schooling, since many jobs do not require significant formal schooling. Also, such assessments may reflect the characteristics of workers currently in jobs, rather than characteristics needed to perform tasks required in the job.

In the worker self-assessment (WSA) approach, surveys are used to ask workers about the level of education and schooling that is required to perform their job. The responses are then employed to compare the actual level of education with the workers’ perceived requirement for the job. While this would incorporate current information, it is very subjective, especially since individuals will tend to exaggerate reported education requirements for their job.

The realized matches (RM) procedure is not subject to bias and is based on the actual educational attainment of each worker in the sample under consideration. A number of approaches have been used in the literature, and the most popular employ the mean and standard deviation (Verdugo and others, 1989) or the modal education level of workers as the reference level. Thus, workers whose education level is above one standard deviation of the mean are overeducated, while those whose level is one standard deviation below are undereducated. Those whose education level is within that range of plus or minus one standard deviation of the mean years are termed adequately 147 This is the case for Canada, Australia and New Zealand.


230

educated.148 In the case of the mode, workers whose actual level of education is the modal level for that occupation are adequately educated for the job. Those whose education is above the mode are overeducated, and those below are undereducated.149 The RM procedure imposes strong symmetry on the matching distribution.

While it is argued that empirical results are not sensitive to the approach used, the methods are very different (Chiswick and Miller, 2005; Hartog, 2000) and are often based on data availability. This must be taken into account when drawing conclusions from such analyses.

OVEREDUCATION AND ITS DETERMINANTS

This paper adopts the RM approach to matching, under which workers whose years of education are one standard deviation above the mean occupational level are categorized as overeducated, while workers whose educational attainments are one standard deviation below the mean are categorized as undereducated. Those workers whose educational attainments are within one standard deviation below and above the mean are categorized as having the required education. It is recognized that either the use of the mode or the mean imposes strong symmetry conditions on the data: however, there is no information in the data set that allows for any other approach to the analysis. The estimation procedure employs a specification which modifies the standard human capital model along the lines referred to as the (OUR) or Overeducated/Undereducated/Required model.

In the standard Mincer approach to human capital investment, the natural logarithm of annual earnings per worker )Y(ln i is regressed on a constant, and a linear term in actual years of schooling

)s( i and linear and quadratic terms in proxies for years of experience such as tx and 2tx , respectively.

Additional variables employed are the personal characteristics of workers, hours worked and occupation effects, plus a random error term.

The equation for this specification is as follows:

i2i2i1

2i2is0i ....xxssYln ε++β+β+α+α+α= ………………....... (1)

where ),0(d.i.i~ 2i σε

This basic equation with modifications has been important to understanding the return to education for a range of countries and situations. On the one hand, although derived from a clear theoretical perspective, namely, human capital theory, its premise lies in some simplifying assumptions. The OUR framework, on the other hand, disaggregates the education variable to incorporate the issues of overeducation, undereducation and those workers who have the required level of education for the job, as given in equation (2).

i2i2i1

2i2

Ui3

Oi2

Ri10

i2i2i1

Ui

Oi

Ris0i

...xxssss

)2(.......................xx)sss(Yln

ε++β+β+α+α−α+α+α=

ε++β+β+−+α+α=

148 Some authors, for example, Nielsen (2007), employ a modified RM procedure; thus, an individual is adequately employed as follows:

( )

( )Dev.Std**decile%decile%

Mediandecile%

MedianSDev.Std**decile%decile%

decile%MedianMedian actual

22575

75

22575

25

⎟⎠⎞

⎜⎝⎛

−−

+≤≤⎟⎠⎞

⎜⎝⎛

−−

−

149 See Cohn and Khan (1995), Kiker and others (1997)


231

In this equation, the education variable is disaggregated such that

UOR ssss −+=

where Rs is required schooling or the years of education required for the job Os is over-schooling, or years of education in excess of that required for the job and Us is under-schooling.

The equation collapses to Rss = for adequately matched workers

OR sss += for overeducated workers and UR sss −= for undereducated workers.

Unlike the standard ordinary least squares (OLS) formulation in equation (2) however, the equation estimated in this paper accounts for the occupation effects, especially since Caribbean migrants are concentrated in particular occupations. The 2000 United States census data suggest, for example, that 24.6% of Caribbean-born workers were in sales and office occupations, as against 19.9% for the foreign born, and 24.4% were in service occupations as against 20.0% for other groups. To capture occupational effects, a hierarchical model is employed (Bryk, A.S and Raudenbush, 1992). The model is set out as follows:

)3(.......................xx)sss(Yln ijoj2ij2ij1

Uij

Oij

Rijs0ij ε+μ++β+β+−+α+α=

),0(N~,),0(N~ oooj2

ij τμσε

where i represents the returns to individuals, and j refers to the occupation of individuals. Thus oju accounts for shifts in the regression line at the occupation level (fixed effects) and ijε is the random error term at the individual level. Hierarchical models allow for nested data structures, and correct for biases in parameter estimates due to the clustering of the data structure. The model allows for the estimation of the individual level )( 2σ and the occupational level )( 00τ variances which can be used to determine how much of the total variation in returns can be accounted for by individuals or by their occupation.150

MODEL FORMULATION AND DESCRIPTIVE STATISTICS

The discussion in the literature suggests that there are no clear-cut explanations for overeducation among migrants relative to native workers. The usual approach is to employ a set of variables based on expectations about the factors that influence the degree of overeducation in the labour market among immigrants and native workers. Among the usual variables employed are experience in the

150 The variation (interclass correlation) accounted for at the individual level is )ˆˆ/(ˆ 2

0000 σ+ττ


232

labour market, age, language skills and source of education, whether acquired in the host country or elsewhere, to capture a notion of the quality of education. The issue of heterogeneity among workers is also important, as workers with similar education-levels may display varying efforts to achieve job matches (Chiswick and Miller, 2008). Such motivations may be unobservable and, in addition, some workers may have better networks than others, which may make it easier to access information about job opportunities in the host country.

It is not clear whether overeducation may reflect the desire to have just any employment or whether it is due to lack of ability to find suitable employment. Furthermore, there may be structural factors which allow for greater overeducation among immigrants relative to native workers. Chiswick and Miller (2005), using the 1% Integrated Public Use Microdata Series (IPUMS), found that the mean years of schooling for employed males for the native born and Caribbean migrants were 13.5 years and 12.4 years respectively. The percentage of overeducation, undereducation and correctly-matched individuals were 11.9%, 8.04% and 80.04% for the native born, while it was 8.06%, 18.16% and 73.77% for Caribbean-born immigrants. Thus, the degree of overeducation was lower and the degree of undereducation was much higher, than for the native born. The results also suggested that the Caribbean-born workers were among the highest of the correctly matched individuals, and among the lowest of the overeducated of any region.

In the light of these results, workers of Caribbean origin were divided into two groups. In the first were Caribbean-born who had been in the United States of America for more than five years and, in the second, were those who had migrated to the United States within the last five years. The minimum age for the second group was 20 years of age, on the assumption that such individuals would have acquired most of their education in the Caribbean. This would provide ample opportunity to test for quality differences in education levels.

Equation (4) sets out the general formulation, where N is for native-born, C is for the Caribbean-born who have lived in the United States for more than five years, while RC represents Caribbean nationals who have arrived within the last five years at age 20 or older. For Caribbean-born workers, the linear and quadratic terms on years in the United States was also employed to capture the impact of experience acquired abroad. Following Chiswick (1997), the weeks and hours worked have both been placed on the right hand side to avoid misspecification of the income equation. The model to be estimated within the hierarchical modeling framework is as follows:151

151 Experience - potential experience defined as: (age-years of schooling -5) Gender- a dummy variable with 1 for males, and zero otherwise. Married- is a dummy variable with 1 for being married and zero otherwise. South- a dummy variable with 1 for living in a Southern state and zero otherwise. Speaks no English-a dummy variable with 1 if an individual does not speak English and zero otherwise. Speaks poor English- a dummy variable with 1 if the individual speaks English poorly and zero otherwise. Veteran-a dummy variable with 1 for veterans and zero otherwise. Metropolitan- a dummy variable with 1 for living in a metropolitan area and zero otherwise. Black is 1 if the individuals belongs to the black race and 0 otherwise.


233

)4..(..............................workedhoursofLogworkedweeksofLogYSM

)YSM(migrationcesinYearsBlacktanMetropoliVeteranPoorlyEnglishSpeaksEnglishnoSpeaksSouthMarried

Gender100/ExperienceExperiencesssYln

ijoj

ij11ij102

12

1110ij9ij8

ij7ij6ij5ij4

ij32ij2ij1

Uij3

Oij2

Rij10)RC,C,N(ij

ε+μ

+β+β+β

+β+β+β+β

+β+β+β+β

+β+β+β+α−α+α+α=

In the model estimation, the years since migration and the square of that variable were restricted to zero in the equation for native workers.152 The data set is the 5 % IPUMS for both males and females from 20 to 65 years of age. This is a relatively large data set, with over 4 million observations for the native born, 57,000 observations for persons of Caribbean origin residing in the United States in excess of five years, and 6,650 for Caribbean-born persons who have migrated within the last five years.

Figure 1 below shows the distribution of income for each of the three groups. It is clear that the distribution of income for recent arrivals is very much to the left of the other two groups. The native-born distribution is slightly to the right of the Caribbean-born living more than five years in the United States. Thus, immigrants from the Caribbean who have been in the United States in excess of five years have a somewhat similar distribution relative to the native born, but they are not identical.

0.2

.4.6

.8Den

sity

4 6 8 10 12 14Log of annual earnings

Native bornCaribbean born > 5 years in USCaribbean born <=5 years in US

Figure 1: Kernel Density Estimates

On the assumption that returns to education might vary by the level of education, which reflect different labour market segments, as was found by Bratsberg and Ragan Jr. (2002), figure 2 plots the mean education and log of mean earnings for the groups at various education levels. The results are striking, for they suggest non-linearities in the relationship between earnings and education, especially for Caribbean immigrants in the United States five years or less. Thus, differences in education level may be an important source of differences in earnings, as shown above.

152 In the standard formulation, an additional variable, the mean occupational wage, was also used. See table 8.


234

9.5

1010

.511

9.5

1010

.511

5 10 15 20

5 10 15 20

Native Born Caribbean Born >5 years in US

Caribbean born <= 5 years in the US

Log

of M

ean

Ear

ning

s

Mean Years of Education

Figure 2: Mean Earnings and Mean Years of Education

Table 1 reports the percentage distribution of employed individuals by occupational groups, after removing missing values and zero-income categories.

TABLE 1 PERCENTAGE DISTRIBUTION OF EMPLOYED BY THREE-DIGIT OCCUPATION GROUPS

Codes in IPUMS

Native born Caribbean born with

more than 5 years in US

Caribbean born 5 years or less in US

1-43 Management, business and financial operations 10.78 7.72 3.62 50-73 Business operations specialists 2.55 1.89 1.34 80-95 Financial specialists 2.65 2.95 1.47 100-124 Professional and related occupations 2.73 1.86 1.74 130-156 Architecture and engineering occupations 2.39 1.65 1.44 160-196 Life physical and social science occupations 1.14 0.67 0.51 200-206 Community and social services occupations 1.92 1.99 1.35 210-215 Legal occupations 1.58 0.94 0.32 220-255 Education, training and library occupations 7.44 4.58 3.29 260-296 Arts, design, entertainment, sports and media

occupations 2.12 1.29 1.54

300-354 Healthcare practitioners and technical occupations

5.65 7.67 3.54

360-365 Healthcare support occupations 1.77 6.92 6.79 370-395 Protective service occupations 2.15 2.29 2.34


235

400-416 Food protection and servicing occupations 2.93 3.13 5.86 420-425 Building and ground cleaning and maintenance


430-465 Personal care and service occupations 2.30 3.28 3.97 470-496 Sales occupations 10.28 9.00 10.01 500-593 Office and administrative support occupations 15.27 15.66 12.28 600-613 Farming, fishing and forestry occupations 0.38 0.21 0.47 620-676 Construction, extraction and maintenance


680-694 Extraction workers 0.10 0.01 0.01 700-762 Installation, maintenance and repair workers 3.94 4.10 4.07 770-896 Production occupations 7.32 6.69 11.06 900-975 Transportation and material moving occupations 5.32 6.61 8.40 980-983 Military occupations 0.29 0.22 0.08

Source: 5 % IPUMS USA. There is a striking similarity between the distribution of workers among the native-born and

the Caribbean-born living for more than five years in the United States, by occupational category. Interestingly, in some occupations, the percentage of Caribbean-born workers arriving within the last five years is higher than for the native born. These major categories are health support occupations, protective services, food protection and servicing, building and ground cleaning and maintenance, personal care and service, transport and material moving occupations, and production operations, many of these being lower level occupations. The highest concentration of recently-arrived immigrants was in production operations, 11.06%, sales, 10.01%, and office and administrative support, 12.28%.

TABLE 2 MEAN YEARS OF EDUCATION BY OCCUPATIONAL GROUP

Codes in IPUMS

Natives Caribbean-born with

more than 5 years in US

Caribbean-born with 5 years or

less in US

1-43 Management, business and financial operations

14.47 14.21 13.93

50-73 Business operations specialists 14.38 14.18 13.57 80-95 Financial specialists 15.06 15.01 15.67 100-124 Professional and related occupations 14.85 14.77 15.36 130-156 Architecture and engineering occupations 14.85 15.02 15.07 160-196 Life, physical and social science

occupations 16.30 15.99 15.84

200-206 Community and social services occupations

15.64 15.13 14.50

210-215 Legal occupations 16.89 16.16 15.24 220-255 Education, training and library

occupations 15.95 15.69 15.28

260-296 Arts, design, entertainment, sports and media occupations

14.65 13.82 13.61

300-354 Healthcare practitioners and technical occupations

15.24 14.90 14.53

360-365 Healthcare support occupations 12.51 11.83 12.20 370-395 Protective service occupations 13.24 12.61 11.93 400-416 Food protection and servicing

occupations 12.25 11.01 11.30


236

420-425 Building and ground cleaning and maintenance occupations

11.85 10.55 10.81

430-465 Personal care and service occupations 12.77 11.75 11.75 470-496 Sales occupations 13.37 12.63 12.72 500-593 Office and administrative support

occupations 12.96 12.81 12.61

600-613 Farming, fishing and forestry occupations 11.79 9.44 9.59 620-676 Construction, extractions and

maintenance occupations 12.08 11.33 11.27

680-694 Extraction workers 11.88 10.82 12.00 700-762 Installation, maintenance and repair

workers 12.33 12.02 11.90

770-896 Production occupations 12.15 11.04 11.51 900-975 Transportation and material moving

occupations 12.10 11.48 11.67

980-983 Military occupations 13.48 13.27 12.44 Overall mean 13.53 12.72 12.28

Table 2 examines the mean years of education by occupational group. The overall mean years of education was higher for the native born and for persons of Caribbean origin with more than five years in the United States, but slightly lower for those who had come in the last five years. The differences in the overall mean years of education were statistically significant at the 5% level. The mean years of education for natives varied between 11.79 and 16.89 years of education while, for persons of Caribbean origin with more than five years in the United States, the mean years of education varied between 9.4 and 16.1 years and, for recent arrivals, the mean was about the same. Interestingly, recent arrivals had a slightly higher average age of education relative to other groups in the occupational groups of professional and related, architect and engineering, and extraction workers.

The overall results suggest no striking difference between the two Caribbean groups. Thus, significant differences in returns for new immigrants would reflect some discounting of local experience and education, plus problems of integrating quickly into the local labour market.

TABLE 3 DISTRIBUTION OF LEVEL OF EDUCATION FOR NATIVE BORN, CARIBBEAN-BORN IN THE UNITED STATES MORE THAN FIVE YEARS, AND CARIBBEAN-BORN IN THE

UNITED STATES FIVE YEARS OR LESS 11th grade

or less 12th grade no

diploma, some College-no

diplomas, high school graduate

or GED

Associate degree

Bachelor’s

degree

Masters’ degree,

professional degree, or Doctorate

Natives 5.81 56.20 8.15 19.62 10.12 Caribbean-born with more than five years in US

14.89 54.77 9.11 13.44 7.78

Caribbean-born and less than five years in US

20.11 56.74 5.75 10.22 7.18

Caribbean born with five years or less in the United States Cuba 16.59 54.34 4.75 11.99 12.32 The Dominican Republic 34.29 47.67 3.53 7.96 6.55 Haiti 19.66 65.00 5.50 7.30 2.54 Jamaica 18.91 58.41 7.17 10.76 4.75 Antigua and Barbuda 15.86 58.14 2.92 13.91 9.18


237

Bahamas (the) 7.41 46.69 16.88 22.40 6.62 Barbados 8.50 50.44 19.15 14.43 7.48 Dominica 18.09 61.04 6.34 11.04 3.49 Grenada 19.27 62.34 8.08 9.50 0.08 Saint Kitts and Nevis 13.68 65.01 3.37 14.03 3.91 Saint Lucia 11.24 66.37 12.43 9.96 0.00 Saint Vincent and the Grenadines 13.10 73.26 12.03 0.00 1.60 Trinidad and Tobago 7.88 68.14 5.61 11.35 7.02 Guyana 18.04 59.81 9.03 9.78 3.33 Source: IPUMS USA, 5 % Sample.

Table 3 examines the percentage of individuals by education level, for each of the three major groups and for individual Caribbean countries. The results are interesting, showing that 5.81% of native born workers, 14.89% of Caribbean–born workers living more than five years in the United States, and 20.11% of Caribbean-born workers living for five years or less in the United States, have a level of education to eleventh grade or below. Except for Associate degrees, the native born had significantly higher shares than the Caribbean-born in the higher degree categories. In terms of professional and other degrees, the percentage of Caribbean-born workers in the United States for more than five years was 7.78%, which was much closer to the 7.18% for recent arrivals. The country-level data are reported for individuals from several Caribbean countries that were living in the United States for a maximum of five years. Cuba, the Dominican Republic and Haiti had the highest percentage of persons in the lower educational levels. In terms of the Caribbean populations, it appears that, on arrival in the United States, individuals upgrade their skills by acquiring higher levels of education.153

Table 4 examines the percentage of overeducation, undereducation and required levels of education, using the RM procedure. The overall results show that, among the employed, native born workers have the highest levels of required education (76.36%), and recent arrivals from the Caribbean have the lowest (66.81%). The percentage of Caribbean-born workers in the United States for more than five years having the required levels of education was 70.79%, which means that Caribbean nationals did not catch up completely over time.

TABLE 4 PERCENTAGE DISTRIBUTION OF OVEREDUCATED, REQUIRED AND

UNDEREDUCATED WORKERS Percentage

overeducated

Percentage with required

education

Percentage undereducated

Native-born 13.91 76.36 9.37 Caribbean-born with more than 5 years in United States 11.77 70.79 17.44 Caribbean-born with less than five years in United States 13.19 66.81 20.00 Caribbean-born with five years or less in the United States by country of origin Cuba 20.95 61.93 17.12 The Dominican Republic 12.72 54.76 32.51 Haiti 8.60 73.30 18.10 Jamaica 8.29 74.78 16.93 Antigua and Barbuda 7.51 75.80 16.69 Bahamas (the) 16.25 78.31 5.44 Barbados 10.45 73.16 16.39

153 The results may vary by age and gender.


238

Dominica 17.09 69.37 13.53 Grenada 6.66 72.56 20.78 Saint Kitts and Nevis 3.91 79.93 16.16 Saint Lucia 4.54 80.97 14.50 Saint Vincent and the Grenadines 10.16 68.45 21.39 Trinidad and Tobago 8.65 78.44 12.91 Guyana 7.14 72.02 20.83

Interestingly, both Caribbean-born groups had lower shares of overeducation relative to the native born, but a higher share of undereducation. This may reflect the heavy concentration in such areas as sales, production occupations, and transport and material moving operations. Similar results are found in Chiswick and Miller (2005).

For individual Caribbean countries, the highest percentages of overeducated workers are to be found among Cubans, with a share of 20.9%, the Bahamas with 16.25%, Dominica with 17.0%, and the Dominican Republic with 12.7%. Interestingly, Cuba and the Dominican Republic are Spanish-speaking countries, and Chiswick and Miller (2005) found that English proficiency was important for labour-market performance.


239

TABLE 5 INCIDENCE OF OVEREDUCATION ACROSS POPULATION SUBGROUPS, AS A

PROPORTION OF ALL PERSONS IN POPULATION SUBGROUP (Percentage)

Age Group Native born Caribbean-born residing more than 5 years in the

US

Caribbean-born residing

5 years or less in the

US 20 to less than 30 12.19 10.23 10.93 30 to less than 40 13.90 12.39 14.67 40 to less than 50 14.72 12.75 13.43 50 to less than 60 15.03 10.73 13.67 Greater than 60 years 13.27 10.58 11.89

In Table 5, the incidence of overeducation is examined by age groups with a range of ten years between age groups 20 to 60 years of age. In the case of the native born, the shares rose from 12.19% to 15.03%, and then fell to 13.27%. With respect to persons of Caribbean origin residing more than five years in the United States, the percentages were much lower with increasing age than those of natives. In the case of persons of Caribbean origin who have been five years or less in the United States, the percentage rose steeply, from 10.93% in the first age group to 14.67% in the second, and then fell to 11.8% in the last group. This means that the Caribbean-born who have migrated within the last five years have a higher share of overeducation, by age group, relative to those that have migrated prior to that period.

TABLE 6 INCIDENCE OF UNDEREDUCATION ACROSS POPULATION SUBGROUPS, AS A

PROPORTION OF ALL PERSONS IN POPULATION SUBGROUP (Percentage)

Age group Native born Caribbean-born, residing more than 5 years in

the US

Caribbean-born, residing 5 years or less in the US

20 to less than 30 8.40 11.81 15.33 30 to lese than 40 9.05 14.15 18.22 40 to less than 50 9.40 17.95 22.53 50 to less than 60 11.82 24.28 32.53 Greater than 60 years 16.00 30.88 43.91

Table 6 examines the incidence of undereducation by the same age groups reported in table 5. At each reported age group, undereducation among the native born is much lower than for Caribbean-born workers who are residing in the United States in excess of five years, or those who have been in the United States for five years or less. For the last group, the percentage of undereducation rose from 15.33% to 43.91%. In summary, workers who are Caribbean-born and residing five years or less in the United States, have higher levels of undereducation than the other two groups, at all age levels.

Some studies have found that English proficiency has a considerable impact on the ability of migrants to integrate, in labour markets for which the language of communication is English. Given the significance of many Spanish-, Dutch- and French-speaking groups of Caribbean origin, it would be useful to examine the degree of over- or undereducation for different levels of English proficiency.


240

TABLE 7 DISTRIBUTION OF OVEREDUCATION, UNDEREDUCATION AND REQUIRED

EDUCATION FOR CARIBBEAN-BORN, BY DEGREE OF FACILITY IN SPEAKING ENGLISH

Caribbean-born, more than five years in the United States

Caribbean-born, five years or less in the United States

Over-educated

Under-educated

Required Over-educated

Under-educated

Required

Does not speak English 4.51 56.38 39.11 9.97 40.03 50.18 Speaks only English 11.93 13.72 74.34 8.24 17.30 78.77 Speaks English very well 14.76 11.48 73.76 16.51 15.02 78.47 Speaks English well 10.13 18.90 70.97 17.76 11.57 70.66 Speaks English, but not well

6.84 38.23 54.93 18.08 19.92 62.01

The results in table 7 show that, for recent arrivals, only 50.18% of workers who speak no English had the required education for their jobs, relative to 39.11% of those Caribbean-born workers residing for more than five years in the United States.154 Interestingly, those who spoke English but not well had the next highest percentage of individuals with the required education.

EARNINGS AND JOB MATCHING

The models were estimated for the full sample of Caribbean immigrants but, because of the large size of the native-born sample, a random sample was taken to reduce the sample size for the native born.155 An initial, multilevel unconditional or random intercept- only model156 (the results are not reported) was first run for each sample to get initial values for the variability in wages between the occupation and individual- level variances, after which equation (4) was run, with the explanatory variables added but accounting only for returns to actual education.

With respect to the null model, the proportion of the variance in returns due to occupations was 18.4% for the native born, 17.7% for the Caribbean-born residing for more than five years in the United States, and 13.0% for recent arrivals.157 This suggests that there is some clustering of returns within occupations, and that OLS would lead to misleading results.158

154 It is important to note that workers who did not speak English, or spoke English but not well, were a relatively small percentage of the population. English proficiency is important, since Cuba, the Dominican Republic and Haiti constitute 63.45% of Caribbean-born workers residing for five years or less in the United States. 155 The sample size was reduced to 74,061 observations. 156 The null model is as follows: ij000ij ry +μ+γ= where 00γ is the grand occupation mean income, j0μ is

the mean occupation income and ir is the individual error term. 157 When the decline in the variance due to adding explanatory variables is computed as [ ] )1Model(ˆ/)2Model()1Model(ˆ 000000 ττ−τ , where model 1 is the null model, the results suggest that the decline in the occupational variance was 98% for natives, 95% for Caribbean-born residing for more than 5 years in the United States, and 90% for recent arrivals. 158 The variance component among individuals is 6 times the variance component among occupations for recent arrivals.


241

In table 8,159 the returns to an additional year of education were 9.1% for the native born, which were higher than the returns found by Chiswick and Miller (2005),160 after controlling for occupation effects using OLS. The returns to education for the Caribbean-born residing for more than five years in the United States were 5.5%, and for recent arrivals, 2.8%.161

With respect to the experience variable, the linear and quadratic variables were significant in all the equations. For the native born, when evaluated at 10 years of experience, the increment to earnings with experience was 2.0%, while for recent arrivals it was 1%.

TABLE 8 REGRESSION ESTIMATES OF EARNING EQUATION, NATIVE-BORN, CARIBBEAN-BORN

WITH MORE THAN FIVE YEARS IN THE UNITED STATES, AND CARIBBEAN-BORN RESIDING FOR FIVE YEARS OR LESS IN THE UNITED STATES.

Native born

Caribbean-born >5 years

in United States

Caribbean-born

=<5years in United States

Native born with linear

Spline

Caribbean-born

>5years in United

States with linear Spline

Caribbean-born =<5 years in United

States with linear Spline

Years of education 0.091*** (70.88)

0.055*** (44.77)

0.028*** (8.83)

Experience 0.032*** (39.87)

0.020*** (20.86)

0.017*** (5.88)

0.034*** (41.56)

0.023*** (24.4)

0.019*** (6.36)

Experience2/100 -0.049*** (-28.38)

-0.031*** (-16.85)

-0.029*** (-4.68)

-0.054*** (-30.55)

-0.040*** (-21.28)

-0.033*** (-5.32)

Gender

0.270*** (46.19)

0.190*** (30.32)

0.203*** (10.96)

0.264*** (45.21)

0.186*** (29.79)

0.202*** (10.91)

Marriage

0.090*** (17.51)

0.072*** (12.81)

0.037* (2.20)

0.089*** (17.41)

0.069*** (12.43)

0.036* (2.14)

South

-0.061*** (-12.30)

-0.122*** (-22.32)

-0.106*** (-6.44)

-0.065*** (-13.11)

-0.119*** (-21.93)

-0.104*** (-6.33)

Does not speak English

-0.042 (-0.30)

-0.161*** (-9.75)

-0.142*** (-5.38)

-0.164 (-1.15)

-0.210*** (-12.74)

-0.148*** (-5.60)

Speaks English poorly

-0.029 (-0.79)

-0.139*** (-14.17)

-0.110*** (-5.09)

-0.048 (-1.31)

-0.161*** (-16.48)

-0.113*** (-5.24)

Veteran

-0.013 (-1.80)

0.016 (1.30)

0.133* (2.57)

-0.005 (-0.68)

0.019 (1.56)

0.136** (2.62)

159 The table reports the well known Bic (Bayesian information criterion), where smaller values reflect better model specifications. The Wald Chi(k) test is a “F” test in a non-linear setting with (k) representing the number of coefficients being estimated. In all cases, the models are well specified, as the null hypothesis of no relationship is comprehensively rejected. N is the number of observations. 160 Chiswick (2007) found returns of 5.8% for the native born and 2.3% for the foreign born. 161 By setting H1:α1 = α2 = - α3 in the OUR model, the traditional Mincer equation is obtained. This restriction is rejected, which suggest that the OUR formulation holds.


242

Metropolitan

0.185*** (35.32)

0.094*** (5.42)

0.071 (1.41)

0.183*** (34.97)

0.098*** (5.66)

0.068 (1.35)

Black

-0.043*** (-5.17)

0.030*** (5.03)

0.084*** (4.57)

-0.041*** (-4.89)

0.035*** (6.03)

0.092*** (4.95)

Log of weeks worked

0.846*** (127.16)

0.803*** (121.58)

0.860*** (59.65)

0.850*** (127.92)

0.804*** (122.41)

0.860*** (59.67)

Log of hours worked

0.871*** (123.99)

0.650*** (80.20)

0.589*** (27.18)

0.869*** (123.87)

0.644*** (79.84)

0.588*** (27.16)

Years since migration (YSM)

0.011*** (9.37)

0.01 (0.41)

0.010*** (8.75)

0.009 (0.35)

YSM2

-0.000 (-1.47)

0.003 (0.62)

-0.000 (-0.97)

0.003 (0.69)

Years of education <= 11 years

0.015*** (3.60)

0.008** (3.29)

0.009 (1.53)

Years of education > 11 years

0.102*** (72.32)

0.082*** (49.80)

0.042*** (8.92)

Constant

1.772*** (32.21)

3.316*** (54.23)

3.668*** (28.99)

2.561*** (37.18)

3.749*** (61.18)

3.836*** (28.87)

00τ̂ .0427*** .043*** .061*** .041*** .036*** .056*** 2σ̂ 0.40924*** 0.3997*** 0.4247*** 0.4073*** 0.3956*** 0.4239***

* p<0.05 ** p<0.01 *** p<0.001

N 74061 57158 6650 74061 57158 6650

Bic 144420.3 110251.4 13506.38 144093 109677.8 13507.69

Wald Chi (k) _62159.7 36426.22 57916.65 62801.35 37411.47 5819.61

Bic: Bayesian information criterion

The linear spline function confirms that workers constitute two heterogeneous groups, as the returns to education at eleven years or less was much lower than the returns above eleven years of education. More important, however, is the fact that the differences in returns persist for the Caribbean groups at both levels of education.

Among the other variables of interest, gender is positive and significant, reflecting higher returns for males, while speaking English poorly has a significant negative effect on earnings. In addition, the coefficients with respect to those who do not speak English, or do not speak it well, are also negative and significant for the Caribbean-born. Chiswick (2007) has shown that, after controlling for occupation, some of the earnings disadvantage of immigrants with limited proficiency is due to this deficiency, placing them in lower-earning occupations.

For those Caribbean-born who have been in the United States for five years or less, the coefficient on the linear and quadratic terms on years since migration is not statistically significant, while the coefficient on the linear terms is significant, though small, for the other Caribbean group.


243

This result is important, since it means that, for recent arrivals, there is no significant relationship between experience gained abroad and income.

For the native born, the elasticity of earnings with respect to weeks worked was 0.84 while for hours worked, it was 0.87. These are likely to have different effects, especially for recent immigrants, since weeks worked may differ from the number of hours worked. Individuals may work similar hours in a week but fewer weeks, for example. The elasticities with respect to weeks worked were similar for all groups, but the returns to hours worked were much less for the Caribbean-born relative to the native born.

When account is made of occupational status, the impact of experience on such status is negative for immigrants (Chiswick, 2007). The results of a quantile regression (not reported here) run with the average income for each occupation used as the status variable, show that the returns to experience, though small, decline as incomes rise. These results suggest that the impact of non-transferability of human capital skills is more important for those who are higher up in income distribution; however, the impact is fairly small. Chiswick (2007) pointed out that foreign workers with experience tend to be channelled into lower-paying occupations (captured by mean occupational earnings) but, within that occupation, receive a relatively high rate of pay.

Table 9 reports the results for the OUR model, which show that the variance due to occupation effects still accounts for 3.1% of the variation for the native born, 2.1% for the Caribbean-born residing for more than five years in the United States, and another 2.3% for recent arrivals. With respect to the native born, the returns to an additional year of required education were 15%, which is some 5.9% higher than that obtained when the actual years of education variable is used in the standard Minceran equation. This difference is accounted for by the fact that the returns to actual years of education are a combination of earning increments to correctly matched job requirements, and to years of education that are not matched with job requirements. Thus, once mismatches are accounted for, the returns to years of schooling tend to be much higher.

The returns to required education for the Caribbean-born with more than five years in the United States, and those who have been in the United States for less than five years, were basically 13% and 2% lower than returns to the native born.

The table also reports two types of mismatches, overeducation and undereducation. In terms of the native born, the return to overeducation was 9.2%, while for the Caribbean-born with more than five years in the United States, it was 5.3% and, for more recent arrivals, the return was 2.0%. Thus, there are wide variations between the groups in terms of rewards to overeducation, even when occupation effects are included. Hence, while the percentage of overeducated workers is higher among the native born, the returns are also greater. In contrast, the Caribbean-born living five years or less in the United States are least rewarded for their overeducation. One other important result is that, in the models for the Caribbean, people of African descent receive a positive return relative to other ethnic groups while, for the host country model, they receive a negative return relative to such groups.

The results are in line with other findings that the returns to required schooling are higher than the returns to actual education. This follows from a comparison between tables 8 and 9. In addition, the returns to overeducation are positive, but smaller than the returns to required education, while the returns to undereducation are negative (Hartog, 2000).

The results also suggest that the labour market discounts experience gathered elsewhere, as the increment to experience for the native born was 3.3%, for the Caribbean-born living more than five years in the United States it was 2.2% and, for recent arrivals from the Caribbean, 1.9%.


244

TABLE 9 REGRESSION ESTIMATES OF EARNING EQUATION, NATIVE-BORN, CARIBBEAN-BORN

WITH MORE THAN FIVE YEARS IN THE UNITED STATES, AND CARIBBEAN-BORN, RESIDING FOR FIVE YEARS OR LESS IN THE UNITED STATES, FOR THE OUR

EDUCATION MODEL BASED ON THE MEAN LEVEL OF EDUCATION

Native born

Caribbean-born living more than

5 years in the United States

Caribbean-born living 5 years or

less in the United States

Experience 0.033*** 0.022*** 0.019***

Experience2/100 -0.052*** -0.037*** -0.033***

Overeducation 0.092*** 0.053*** 0.020*

Undereducation -0.085*** -0.039*** -0.022***

Required education 0.150*** 0.133*** 0.128***

Gender 0.271*** 0.186*** 0.200***

Marriage 0.089*** 0.070*** 0.039*

South -0.065*** -0.121*** -0.105***

Does not speak English -0.138 -0.200*** -0.143***

Speaks English poorly -0.036 -0.152*** -0.095***

Veteran -0.006 0.018 0.118*

`

Black -0.043*** 0.037*** 0.087***

Log of weeks worked 0.848*** 0.803*** 0.858***

Log of hours worked 0.866*** 0.643*** 0.583***

Years since migration(YSM) 0.010*** 0.018

YSM2 0 0.001

Constant 0.988*** 2.261*** 2.346***

.031*** .021*** .023***

σ2 .406*** .394*** .418***

* p<0.05,

** p<0.01

*** p<0.001

N 74061 57158 6650

Bic 143873.1 109518.2 13426.56

Wald Chi(k) 63262.41 37752.56 60239.93

00τ̂

THE IMPACT OF CLIMATE CHANGE ON THE TOURISM · PDF fileTHE IMPACT OF CLIMATE CHANGE ON THE...

Documents

Transcript of THE IMPACT OF CLIMATE CHANGE ON THE TOURISM · PDF fileTHE IMPACT OF CLIMATE CHANGE ON THE...