RETURNS TO EDUCATION IN THAILAND 1994-2002† Background Paper for the 2004 Thailand Social Monitor

Niels-Hugo Blunch The George Washington University [email protected]

Draft, May 6, 2004. JEL classification: J31, J71. Keywords: Returns to education; human capital; wage gap decomposition; household fixed effects; East Asian Crisis; Thailand. Abstract: This paper examines education returns in Thailand over the period 1994-2002. First, the developments in the stocks of human capital in terms of educational attainment of the workforce wage are examined and wage equations over this period are estimated, addressing several weaknesses of previous studies of education returns in Thailand. The results imply that the returns to education in Thailand over the period are substantial, peaking in 2000. In turn, this indicates that the East Asian crisis did not fully materialize until after the year 2000, at least as far as education returns are concerned. Second, the paper examines the wage gap related to gender and municipal-non-municipal location by providing results for a wide range of alternative wage decompositions in the Blinder-Oaxaca tradition. The wage gaps related to gender and municipalnon-municipal location have decreased over time, while at the same time the share of the gap due to unobservable factors (though remaining large) has declined, as well, indicating that wagediscrimination has decreased in Thailand over the period.

†

This paper was commissioned by the East Asia and Pacific PREM Sector Department, World Bank, for the 2004 Thailand Social Monitor. I thank John Blomquist, Ana Revenga and Khuankaew Varakornkarn for guidance and support. Remaining errors and omissions are my own. The views expressed here are those of the author and should not be attributed to the World Bank or any of its member countries.

1. Introduction As most countries in the region Thailand, too, was seriously hit by the South East Asian economic crisis in 1997-98. The crisis caused massive lay-offs, which, in turn, have affected wages and earnings and therefore individual welfare. But how exactly have wage earners in Thailand fared over the past decade? And how has human capital accumulation in terms of educational attainment in the traditionally1 well-educated workforce in Thailand been affected as a result of this? Only few studies have examined the returns to education in Thailand in the first place, and therefore much less so the impact of the recent crises on education and earnings in Thailand.2 Indeed, only one recent published study has some bearing3 on the focus of the current paper, namely Moenjak and Worswick (2003), which examine the selection into and subsequent return from secondary vocational education relative to general secondary education in Thailand. The authors pool data from the Labor Force Survey for the years 1989 to 1995 inclusive, selecting as sample only those persons who are sons or daughters of the household head and live with their parents at the time of the survey (since one of their key variables, they argue, is parental education and that is only available for this sub-sample). The main findings related to the wage regressions relevant to my paper are that workers with vocational secondary education receive higher wages than workers with general secondary education, a positive wage premium for workers from Bangkok relative to workers from other regions and negative wage premium to workers from

1

See Khoman (1993) for more detail on education policy in Thailand. A search in Econlit under either “Thailand education” or “Thailand earnings” in the title yields only four relevant studies, which has been published in academic journals: Moenjak and Worswick (2003), Chiswick (1983) and Blaug (1974, 1976). Only the first of these is recent enough to be relevant for this study. When moving to working papers and reports from international organizations and the like, there are more studies available but still relatively few as compared with most other countries. Examples here are Grootaert (1986), Hawley (2001) and Yamauchi (2003). 3 Although the data analyzed stop short of the crisis and therefore do not allow examining the impact of the crisis on education returns. 2

2

municipal areas. Lastly, the returns to both education and experience are higher for females than for males, conditional on gender (only sub-sample regressions across gender are reported). Moving to working papers and reports, there are more studies available although not as many as for most other countries. Hawley (2001), examining data from Thailand’s National Labor Force Survey from 1985 to 1998, finds evidence of increases in educational attainment (in terms of the number of years of schooling) and a decline in the gap related to educational attainment between rural and urban areas between 1985 and 1998. Additionally, the returns to education is found to have increased from 1985 to 1995 and then decreased from 1995 to 1998. Both of these studies have a few shortcomings, which are quite common in the earnings equations literature. First, by estimating by means of standard ordinary least squares4 the studies (implicitly) assume exogeneity of the education variables. If this maintained hypothesis is not satisfied, the estimated parameters will be biased and inconsistent. Second, for the case of Hawley (2001) the returns to education are computed based on the “raw” de-exponentialization of the dummy variable, not taking into account the impact of the precision of the estimate, which may potentially seriously bias the estimated impact (see Kennedy, 1981 for a detailed discussion of this), while the Monjak and Worswick paper simply reports the dummy variable parameter estimate directly. For females in the selectivity corrected regression, for example, Moenjak and Worswick report a relative return of 49.4 percent, whereas the calculated return using Kennedy’s formula is 58.7 percent. Third, for the case of the Hawley paper, returns are estimated for the various years separately without formally testing for this, for example by means of a Chow-type test for exclusion of year interactions in a pooled regression model. Moenjak and Worswick, on the other hand, pools the data for all seven years, thereby implicitly assuming constant returns to education and experience over the entire period, allowing only for a change in the intercept after 4

And self-selection corrected for the case of the Moenjak and Worswick paper.

3

1991 Fourth, a variable for an individual’s marital status is included into the regressions. While it is certainly true that married status may affect the reservation wage of an individual, it does not seem reasonable that it should affect the observed wage, i.e. market wages. In the best case, this is an unnecessarily included explanatory variables. In the worst case, however, the inclusion of this—potentially endogenous—variable may create biased (and inconsistent) parameter estimates. Fifth, while Moenjak and Worswick attempts to control for self-selection, their insistence on using parental education as an explanatory variable (rather than for example allowing for household specific fixed effects, which would seem to capture socioeconomic background, as well) in effect generates sample selection by excluding a very large part of the sample. Indeed, their subsequent sensitivity analyses indicate that the sons and daughters indeed are a select sample of the general population. Additionally, in terms of examining the impact of the East Asian crisis (the Hawley paper), 1998 is probably “too early” to see the full impact of the crisis (indeed, one might conjecture that the full impact of the crisis will not be fully reflected in the data for a long time, due to the dynamic inter-linkages of human capital accumulation, wages/earnings and economic welfare). This paper examines the returns to education in Thailand over the period 1994-2002, while trying to address the potential problems of previous studies raised above. Also, by analyzing a longer time span than previous studies, I am in a better position to capture the impact of the crisis on returns to education. The analyses consist of two main parts. First, I examine the developments in the stocks of human capital in terms of educational attainment of the workforce and estimate the returns to education over time in Thailand. This part of the analyses addresses the following questions: (1) How have the returns to education changed overall in Thailand over the period 1994-2002, including the changes within each level or type of education over time and the changes in the relative returns between different levels or types of education over time for (1) 4

the full sample (2) females and males? (3) across municipal and non-municipal areas and (4) across regions. Second, I examine in more detail the earnings gaps related to gender and municipal-non-municipal location. This parts of the analyses investigates how much of the earnings differentials for the female-male and municipal-non-municipal sub-groups are due to differences in observed skills, including education and experience, and how much is due to unobserved factors (in the literature typically interpreted as “discrimination”). The paper is structured as follows. The methodology of the paper is discussed in the next section. Section three presents the data examined in this paper, while section four provides descriptive analyses of wages and educational attainment in Thailand. The econometric analyses follow in section five, while section six concludes and provides implications for policy.

2. Methodology This section reviews the methodology applied in this paper. The first section presents a simple economic model in the human capital tradition, followed by a discussion of the econometric techniques applied in the empirical analyses in the next section.

2.1 Conceptual framework The main contribution of this paper is empirical. Hence, I will limit the theoretical discussions to the minimum necessary to provide a satisfactory theoretical underpinning of the empirical analyses. The theoretical framework for the analysis is standard human capital theory, where an individual builds up knowledge and skills via education and experience (specific on-the-job, as well as general experience) (Becker, 1993; Mincer, 1974). Formally, the economic model may be derived from the theory of individual demand for schooling, which views education as an investment in human capital (Becker, 1993). In the traditional human capital literature earnings 5

are determined by education and other individual characteristics. In the subsequent applied earnings equations literature variables have often been added in a more or less ad hoc fashion, for example occupation and industry variables. Since this paper specifically focuses at education returns in Thailand, the bare minimum of variables required to make valid inference on these returns are included so as to be sure to estimate the impact of education on earnings, thereby avoiding mixing returns to education with returns to occupation or industry, say. These considerations lead the following simple model: Wi = W(Ii, Hi, Ci)

(2.1)

where W (wages) of individual i is the dependent variable, I is a vector of individual characteristics: gender, educational attainment, age and age squared, the latter to capture possible non-linearities, proxying general experience. H captures household characteristics, for example related to ability (assortative mating)5 or the networks available to the household. C is a vector of community variables in the form of geographical location: municipal-non-municipal location and region.

2.2 Econometric framework The workhorse in the applied human capital earnings equations literature undoubtedly has been ordinary least squares (OLS). While intuitive and straightforward to use, this approach has some problems, however. Most importantly, the method assumes that all regressors are exogenous. Additionally, all unobserved heterogeneity, including ability, is subsumed in the residual. In the event that the explanatory variables are not exogenous, this will lead to inconsistent estimates. For example, while it is commonly assumed in the human capital literature that completed education

5

See Park and Smits (2003), which find that Thailand have relatively low levels of homogamy in spousal education examining data for ten East and South-East Asian countries for four marriage cohorts (1950-64, 1965-74, 1975-84 and 1985 or later).

6

may be treated as exogenously given since it is effectively predetermined, it could be claimed that education should rather be treated as endogenous or, alternatively, that ability needs to be controlled for. The reason is that ability is unobserved so that if more able people obtain more education, the error term and the explanatory variables will be positively correlated, leading to biased and inconsistent estimates. One way to deal with this, if having panel data, is to extract the individual (fixed) effect from the data, whereby the resulting estimates in the earnings equation for education, experience and so on will indeed be valid. While I do not have panel data available in the present case, it is possible to still do something about the potential endogeneity (unobserved heterogeneity). Specifically, a household specific (rather than an individual-specific) effect may be included in the regressions— this will be identified, while an individual-specific effect will not—effectively capturing the “Hi”term in (2.1) above. The rationale behind the household-specific effect is that it will partially capture ability, at least at the household level, rationalized for example by the presence of assortative mating. It will also capture factors related to social and professional networks, which will commonly be shared within households. While it is not clear whether one should specify the empirical counterpart of (2.1) using fixed or random household specific effects, the hunch would be to specify the models using fixed effects (due to the possibility of endogeneity/omitted variables bias). This may be tested, however. Indeed, in order to examine the estimated model(s) and formally evaluate the different models so as to find the “best” model a battery of specification tests are carried out. These include Hausman (1978) specification tests for discriminating between random and fixed effects models, tests for statistical significance of the fixed effects in the fixed effects specification, and finally a test for sample split of the models across years, gender, municipal-non-municipal location and region (see Appendix H for details). 7

This model is then subsequently estimated separately across gender, municipal-nonmunicipal location and region so as to examine the structural differences in education returns along these dimensions. So as to provide a benchmark and since this model is the one typically employed in this type of analyses, the results using standard OLS is included, as well. Lastly, while the results from the earnings regressions will provide an initial indication of the presence and direction of earnings gaps related to various characteristics, this will be taken a bit further by examining various decompositions of the Oaxaca-Blinder type (Oaxaca, 1973; Blinder, 1973). Specifically, the earnings gaps related to gender and municipal-non-municipal location will be decomposed. Formally, the estimable counterpart of (2.1) given by: Yi = β 0 + β 1 Di + β 2 X i + ε i

(2.2)

, where Yi is ln(earnings) and Di is a female dummy variable (or municipal-non-municipal location dummy, as case may be). For simplicity only one explanatory variable, X, is included but this could be thought of as a vector of characteristics, including experience and education. Estimating (2.2) separately for females and males, the resulting estimated wage gap may then be decomposed as: Y m −Y

f

= ( X m − X f ) βˆ m + X f ( βˆ m − βˆ f )

(2.3)

The first term on the right hand side of (2.3) is the part of the gender wage gap that may be attributed to endowments in terms of education and experience, while the second term is what is typically termed “discrimination” in the literature. The reason for this is that this term effectively is a premium above and beyond that resulting from formal qualifications in terms of for example education and experience. It is possible to go one step further, decomposing the two terms on the right hand side of (2.3) into the contribution from the different explanatory variables (see Oaxaca, 1973 and Blinder, 1973 for details). This will be done in the subsequent empirical analyses. 8

While the focus in the empirical analyses will be on the results from the Oaxaca-Blinder decompositions, results from the Reimers (1983), Cotton (1988) and Neumark (1988) decompositions will be offered as alternative wage differential decompositions to check for the robustness of the results from the Oaxaca-Blinder decompositions.

3. Data The data are from the Social Economic Survey, which is an annual household survey carried out by the Thailand National Statistical Office since 1957. The data analyzed in this paper include five rounds of the SES: 1994, 1996, 1998, 2000 and 2002. The survey contains information on individual characteristics such as age, gender, and educational attainment and earnings and household background information such as consumption expenditures and assets. Additionally, there are special additional modules for some years. The earnings measure is (log) monthly wages and salaries in real terms (deflated by CPI, 1995=100), so that the sample essentially consists of wage or formal sector workers. The variables from the survey selected as explanatory variables for the analyses in this paper are specifically chosen so as to focus on the (gross) returns to education, while at the same time ensuring that inference on these returns will be valid. Hence, a “core” gender, age and age squared (to capture potential general experience, which may potentially be non-linear), educational attainment, municipal-non-municipal location and region (to capture locational earnings differences) is chosen as the main focus of the estimations. Again, including marital status as an explanatory variable in the wage regression as has sometimes been done (Hawley, 2001, for example) appears misplaced. This variable, while probably affecting reservation wages, would seem unlikely to affect observed, i.e. market, wages and therefore should not be included. Various other candidates for inclusion as explanatory variables such as occupation and industry. 9

While possibly interesting and illuminating in their own right these variables are not included here since this paper specifically is aimed at examining the returns to education, that is the gross returns, as these are the measures most directly relevant for policy. The set of additional control variables included here—age, age squared (to capture potential general experience), gender and the household-specific fixed effects—seems to be the “minimum set” of variables which will ensure valid inference of the education returns parameters.

4. Descriptive Analysis of Wages and Educational Attainment Descriptive statistics for the regression samples are shown in Appendix B. So as to better highlight the developments in educational attainment, however, these variables are graphed below in individual graphs for the full sample, as well as broken up into sub-samples related to gender, municipal-non-municipal areas and region (Figures 4.1-4.10 below). From these figures several issues appear. First, for the period as a whole, the number of individuals who have not attended school has remained roughly constant at about 4 percent for the sample as a whole and about 5 and 3 percent for females and males and about 5 and 1 percent for non-municipal and municipal areas (Figures 4.1-4.5). However, there has been changes in the composition of educational attainment over the period towards higher levels of educational attainment for the people who do go to school. This is most apparent in the change in the fraction of individuals who complete only lower primary education: starting from about 44 percent in 1994, it declined to about 31 percent in 2002. The levels of upper primary educational attainment remains fairly stable at about 25 percent , while the increases are mainly concentrated within lower and upper secondary and university graduates. (Figure 4.1). These same patterns in educational attainment developments hold for female and males, as well (Figures 4.2 and 4.3). Interestingly, the amount of university drop-out remains 10

extremely low over the period , at less than half a percent. In turn, this indicates that once an individual enters tertiary education, one is highly likely to complete, as well (or at least to obtain a degree). Second, Thailand is characterized by substantial differences in the stocks of human capital of individuals from municipal and non-municipal areas as well as across regions. The gender gap in education well known from most other developing countries is not as evident for Thailand. While males generally dominate at lower and upper primary, lower and upper secondary and vocational education, for university this is reversed: here, females consistently have at least as great a share as do males over the entire period 1994-2002 (Figures 3.2 and 3.3). The human capital stock in Thailand appears extremely skewed towards municipal areas and the Bangkok region (Figures 4.44.10). This, of course, is likely due to a mix of supply and demand factors: the supply of education institutions is higher in urban areas but the demand for skilled labor is higher in urban areas, as well. Lastly, from Tables B1 and B2, Appendix B, there appear to exist substantial earnings gaps related to gender and geographical location, the mean earnings of males being higher than those of females and the earnings of individuals from municipal areas being higher than those of individuals from non-municipal areas. Similarly, the earnings for other regions lag behind those of Bangkok (apart from in 1994). While the previous preliminary analyses of developments in educational attainment seems to suggest that at least parts of the earnings differential may be explained by the gaps in the human capital stocks, especially across municipal and non-municipal location, these earnings gaps will be examined a bit more rigorously applying Oaxaca-Blinder type earnings decompositions in the multivariate analyses below.

11

Figure 4.1 Educational Attainment in Thailand 1994-2002, Full Sample (Percent).

0.400 1994 1996 1998 2000 2002

0.300 0.200 0.100

Pr im

N on

e ar y, Pr lo im w ar y, up Se pe co r nd ar Se y, co lo nd w ar U y ni ,u ve pp rs er ity ,a U tte ni nd ve ed rs ity ,d eg re e Vo ca tio na l Te ch ni ca l

0.000

Figure 4.2 Educational Attainment in Thailand 1994-2002, Females (Percent).

0.400 1994 1996 1998 2000 2002

0.300 0.200 0.100

ar y, Pr lo im w ar y, up Se pe co r nd ar Se y, co lo nd w ar U y ni ,u ve pp rs er ity ,a U tte ni nd ve ed rs ity ,d eg re e Vo ca tio na l Te ch ni ca l

Pr im

N on

e

0.000

12

Figure 4.3 Educational Attainment in Thailand 1994-2002, Males (Percent). 0.500 0.400

1994 1996 1998

0.300 0.200

2000 2002

0.100

ar y, up Se pe co r nd ar Se y, co lo nd w ar U y ni ,u ve pp rs er i ty ,a U tte ni nd ve ed rs i ty ,d eg re e Vo ca tio na l Te ch ni ca l

ar y, lo w

Pr im

Pr im

N on e

0.000

Figure 4.4 Educational Attainment in Thailand 1994-2002, Non-Municipal Areas (Percent).

0.500 1994 1996 1998

0.400 0.300

2000 2002

0.200 0.100

ar y, up Se pe co r nd ar Se y, co lo nd w ar U y ni ,u ve pp rs er i ty ,a U tte ni nd ve ed rs i ty ,d eg re e Vo ca tio na l Te ch ni ca l

ar y, lo w

Pr im

Pr im

N on e

0.000

13

Figure 4.5 Educational Attainment in Thailand 1994-2002, Municipal Areas (Percent). 0.300 1994 1996 1998

0.200

2000 2002

0.100

ar y, up Se pe co r nd ar Se y, co lo nd w ar U y ni ,u ve pp rs er i ty ,a U tte ni nd ve ed rs i ty ,d eg re e Vo ca tio na l Te ch ni ca l

ar y, lo w

Pr im

Pr im

N on e

0.000

Figure 4.6 Educational Attainment in Thailand 1994-2002, Bangkok Region (Percent).

0.500 1994

0.400

1996 1998 2000

0.300 0.200

2002

0.100

ar y, up Se pe co r nd ar Se y, co lo nd w ar U y ni ,u ve pp rs er i ty ,a U tte ni nd ve ed rs i ty ,d eg re e Vo ca tio na l Te ch ni ca l

ar y, lo w

Pr im

Pr im

N on e

0.000

14

Figure 4.7 Educational Attainment in Thailand 1994-2002, Central Region (Percent). 0.500 0.400

1994 1996 1998

0.300 0.200

2000 2002

0.100

ar y, up Se pe co r nd ar Se y, co lo nd w ar U y ni ,u ve pp rs er i ty ,a U tte ni nd ve ed rs i ty ,d eg re e Vo ca tio na l Te ch ni ca l

ar y, lo w

Pr im

Pr im

N on e

0.000

Figure 4.8 Educational Attainment in Thailand 1994-2002, Northern Region (Percent).

0.500 0.400

1994 1996 1998 2000 2002

0.300 0.200 0.100

Pr

N on e im ar y, Pr lo im w ar y, up Se pe co r nd ar Se y, co lo nd w a U ry ni ,u ve pp rs er ity ,a U t te ni nd ve ed rs ity ,d eg re e Vo ca tio na l Te ch ni ca l

0.000

15

Figure 4.9 Educational Attainment in Thailand 1994-2002, North-Eastern Region (Percent).

0.500 0.400

1994 1996 1998 2000 2002

0.300 0.200 0.100

Pr

N on e im ar y, Pr lo im w ar y, up Se pe co r nd ar Se y, co lo nd w a U ry ni ,u ve pp rs er ity ,a U t te ni nd ve ed rs ity ,d eg re e Vo ca tio na l Te ch ni ca l

0.000

Figure 4.10 Educational Attainment in Thailand 1994-2002, Southern Region (Percent). 0.400 0.300

1994

0.200

1996 1998

0.100

2000 2002

ar y, up Se pe co r nd ar Se y, co lo nd w ar U y ni ,u ve pp rs er i ty ,a U tte ni nd ve ed rs i ty ,d eg re e Vo ca tio na l Te ch ni ca l

ar y, lo w

Pr im

Pr im

N on e

0.000

16

5. Econometric Analysis In this section, the main findings of the empirical analysis are presented. Estimation is carried out at the level of individual workers and the community level data (municipal-non-municipal location and region of residence) are then matched on to these workers. Again, since the focus on this paper is on the returns to education, only the bare minimum of variables required to ensure valid inference on educational returns are included. The main model includes gender, age and age squared, educational attainment, municipal-non-municipal location and region of residence (the latter to capture spatial earnings differentials).6 This model is then subsequently estimated separately across gender, municipal-non-municipal location and region so as to examine the structural differences in education returns along these dimensions. The “preferred model” is estimated by maximum likelihood and includes household fixed effects. So as to provide a benchmark and since this model is the one typically employed in this type of analyses, results using standard OLS is included, as well. While the earnings equations analysis provides a first cut at the earnings differential related to gender and municpal-non-municipal location, this is then taken a bit further by means of Oaxaca-Blinder type earnings decompositions.

Results from specification tests Estimating the model with all five years pooled and including year dummies and allowing for the full set of interaction effects with the year dummies reveal that the pooled model may be rejected in favor of models estimated separately across the five survey years both when using OLS and when using fixed effects (Table C1, Appendix C). The next step, namely using the Hausman (1978) specification test for each individual year to test between the random and fixed effects specification, indicates that the explanatory variables are endogenous and that, therefore, the fixed 6

See Appendix A for more detail on the variables in the different regressions.

17

effects specification should be chosen over the random effects version (and OLS) for all models, see Appendices C through G. The fixed effects specification is therefore the preferred specification to be used for the analyses in this paper. I do, however, include ordinary least squares estimates, as well, so as to provide a benchmark model, as this is the most widely applied means of estimation in the earnings regressions literature, so as to allow comparison with the fixed effects results. Having chosen the fixed effects specification as “the” preferred specification, the last step is to test for the statistical significance of the household specific fixed effects. These are statistically significantly different from zero for all models, which implies that unobserved heterogeneity in terms of household specific fixed effects—capturing for example assortative mating and access to social and professional networks—are important for wage determination in Thailand. This accords with previous evidence, see Park and Smits (2003). In turn, this implies that estimates from standard OLS-regressions are biased. The importance of the household specific effects are also seen from the difference in the OLS and fixed effects estimates: OLS estimates are generally much higher than FE estimates, again indicating the importance of unobserved heterogeneity. Reliance on the OLS estimates therefore would overestimate education returns in Thailand substantially. Having settled on the fixed effects specification and the validity of this being estimated separately across years as discussed above, I perform likelihood ratio tests of additional sample splits across gender, municipal-non-municipal location and region of residence. The results from these tests support these additional sample splits and, hence, the validity of estimating the model separately across gender, municipal-non-municipal location and region of residence.

18

Education returns over time, full sample In order to convert the parameter estimates from Table C1 in Appendices C to rates of returns these estimates needs to be “de-exponentialized”. Again, the crude conversion factor given as exp[b] – 1, where b is the parameter estimate, typically employed in earnings equation studies may lead to distorted inference in so the alternative, more appropriate, conversion factor given as exp[b – Variance(b)/2] – 1 is used instead (see Kennedy, 1981, for further discussion).

Before examining the returns to education over time in more detail, however, a word of caution related to a few of the educational categories is in order. At first, the returns to teacher training appear very large. However, the sample of teachers is based on only few observations, and one therefore cannot attribute too much importance to the parameter estimates of returns for this group. Also, as a result of the small sample size, the parameter estimate for teacher training is very imprecisely measured, as is also seen for the 95 percent confidence interval (not shown, available upon request)—spanning a range of more than 100 percentage points in many cases. The same is true for the “other education” category, which also is quite small. This is not too much of a concern, however, since the focus of these analyses is on the relative returns of education in general, of which teacher training only amounts for a small fraction. The other types and levels of education account for much larger shares of the total pool of education and therefore both are of more policy relevance and at the same time their estimated education returns may be viewed as more accurate. From Figure 5.1, several points related to the developments in education returns over the period 1994-2002 are worth noting. First, education returns are substantial in Thailand over the period, the East-Asian Crisis notwithstanding. Second, however, at the same time returns are generally quite volatile over the period. Third, returns peak across all levels except university attendees in 2000 and thereafter declines. This hints at the impact from the East-Asian crisis 19

materializing after 2000, at least as far as returns to education are concerned. These major tendencies, however, should not hide the fact that the returns to primary education, especially lower primary education, is fairly constant over the period. One interpretation of this is that since this education is so basic—and, in Thailand, is also mandatory—there just is not that much of a response to supply and demand from this segment of the labor force. Returns are quite different for the various sub-samples. Returns are higher for females than for males across virtually all levels of education, partly reflecting the lower stocks of female human capital found in the descriptive analyses in the previous section and therefore higher associated marginal product from females (Figures 5.2 and 5.3). Similarly, there are substantial differences in education returns associated with geographical location, returns being higher in municipal areas relative to nom-municipal areas and in Bangkok relative to other regions, especially for higher levels of education. Again, this is primarily due to skills being in higher demand in urban areas (Figures 5.4-5.10). Figure 5.1 Education Returns in Thailand 1994-2002, Full Sample (Percent). 200 150

1994 1996 1998 2000 2002

100 50

y, up Se pe co r nd a ry Se ,l co ow nd ar y, U ni up ve pe rs r ity ,a tte U nd ni ve ed rs ity ,d eg re e Vo ca tio na l Te ch ni ca l

Pr im ar

Pr

im

ar y

,l

ow

0

20

Figure 5.2 Education Returns in Thailand 1994-2002, Females (Percent). 400 350 300 1994 1996 1998 2000 2002

250 200 150 100 50

Pr im

ar y, lo w

-50

Pr im ar y, up Se pe co r nd a ry Se ,l co ow nd ar U y, ni up ve pe rs r ity ,a tte U nd ni ve ed rs ity ,d eg re e Vo ca tio na l Te ch ni ca l

0

Figure 5.3 Education Returns in Thailand 1994-2002, Males (Percent). 200 150

1994 1996 1998 2000 2002

100 50

y, up Se pe co r nd a ry Se ,l co ow nd ar y, U ni up ve pe rs r ity ,a tte U nd ni ve ed rs ity ,d eg re e Vo ca tio na l Te ch ni ca l

Pr im ar

Pr

im

ar y

,l

ow

0

21

Figure 5.4 Education Returns in Thailand 1994-2002, Non-Municipal Areas (Percent). 200 150

1994 1996 1998 2000 2002

100 50

y, up Se pe co r nd a ry Se ,l co ow nd ar y, U ni up ve pe rs r ity ,a tte U nd ni ve ed rs ity ,d eg re e Vo ca tio na l Te ch ni ca l

Pr

Pr im ar

im

ar y

,l

ow

0

Figure 5.5 Education Returns in Thailand 1994-2002, Municipal Areas (Percent). 275 225 1994 1996 1998 2000 2002

175 125 75

y, up Se pe co r nd a ry Se ,l co ow nd ar y, U ni up ve pe rs r ity ,a tte U ni nd ve ed rs ity ,d eg re e Vo ca tio na l Te ch ni ca l

Pr im ar

Pr

im

ar y

,l

-25

ow

25

22

Figure 5.6 Education Returns in Thailand 1994-2002, Bangkok (Percent). 275 225 1994 1996 1998 2000 2002

175 125 75

y, up Se pe co r nd a ry Se ,l co ow nd ar y, U ni up ve pe rs r ity ,a tte U ni nd ve ed rs ity ,d eg re e Vo ca tio na l Te ch ni ca l

Pr

Pr im ar

im

ar y

,l

-25

ow

25

Figure 5.7 Education Returns in Thailand 1994-2002, Central Region (Percent). 200 180 160 140 120 100 80 60 40 20 0 y, up Se pe co r nd a ry Se ,l co ow nd ar y, U ni up ve pe rs r ity ,a tte U nd ni ve ed rs ity ,d eg re e Vo ca tio na l Te ch ni ca l

Pr im ar

Pr

im

ar y

,l

ow

1994 1996 1998 2000 2002

23

Figure 5.8 Education Returns in Thailand 1994-2002, Northern Region (Percent). 250 200

1994 1996 1998 2000 2002

150 100 50

y, up Se pe co r nd a ry Se ,l co ow nd ar y, U ni up ve pe rs r ity ,a tte U nd ni ve ed rs ity ,d eg re e Vo ca tio na l Te ch ni ca l

Pr im ar

Pr

im

ar y

,l

ow

0

Figure 5.9 Education Returns in Thailand 1994-2002, North-Eastern Region (Percent). 250 200 1994 1996 1998 2000 2002

150 100 50

Pr im ar y, lo Pr w im ar y, up Se pe co r nd a ry Se ,l co ow nd ar U y, ni up ve pe rs r ity ,a tte U ni nd ve ed rs ity ,d eg re e Vo ca tio na l Te ch ni ca l

0 -50

24

Figure 5.10 Education Returns in Thailand 1994-2002, Southern Region (Percent). 450 400 350 300

1994 1996 1998 2000 2002

250 200 150 100 50 y, up pe co r nd a ry Se ,l co ow nd ar U y, ni up ve pe rs r ity ,a tte U ni nd ve ed rs ity ,d eg re e Vo ca tio na l Te ch ni ca l Se

Pr im ar

Pr

im

ar y

,l

-50

ow

0

Initial look at the gender wage gap I will now examine the “returns to gender”, that is, provide and initial examination in the developments of male-female relative wages in Thailand in a similar fashion as that applied to the previous discussion of the developments in the returns to education in Thailand. Again using the formula from Kennedy (1981), the average returns for females relative to those of males may be calculated over the period as exp[coefficient-0.5*standard error]-1 using the estimation results from Appendices C and F (fixed effects specification). The resulting gender wage premium is graphed in Table 5.1. Note that for presentation purposes the absolute value of the gender wage premium has been graphed, so that a smaller value corresponds to a smaller gender earnings gap. From the figure, the gender wage gap has decreased over the period as a whole from about 33 percent in 1994 to about 24 percent in 2002 for the full sample. Against this overall decline the regional experiences have been quite different. Bangkok and Central region, although starting out 25

from the relatively highest levels in 1994 (about 42 and 34 percent, respectively) experienced the larges decline both in absolute and in relative terms, ending up at about 18 and 17 percent in 1994. For the three other regions, the earnings gap has even widened over the period as whole, although the development from 1996 and onwards is marked by and overall decline. For both the NorthEastern and Southern region, however, the last part of the period, from 2000 to 2002, saw a widening of the gender wage gap by about 3 percentage-points to about 31 and 24 percent, respectively.

Figure 5.11 Gender Wage Premium in Thailand 1994-2002, Full Sample and Across Regions (Absolute Value, Percent) 45 40 35 30

20

1994 1996 1998 2000

15

2002

25

10 5 0 Full sample

Bangkok

Central

North

North-east

South

Wage Decompositions across gender Going one step further, one may examine the observed gender earnings gap as given by the coefficient on the gender dummy discussed previously in a bit more detail. As indicated by Table 5.1 the overall female-male earnings gap has decreased over time: from a “raw” differential in the mean monthly earnings of 773.6 Baht in 1994, the period saw a consistent decline in the gap until 2000, followed by a slight increase in 2002. The observed gender gap in 2002 of about 736.9 Baht 26

in monthly earnings is still considerably lower than that in 1994, so that over the entire period the gender-wage gap has decreased by 4.8 percent in real terms. Decomposing the gender gap into observable and unobservable factors, one essentially compress the results on sample means from Appendix B so as to estimate the amount of the gap, which is attributable to differences in observable characteristics (endowments), namely experience and education and that which is due to other (unobservable) factors, often denoted “discrimination” in the literature. From the results in Table 5.1, three findings stand out. First, the difference in endowments of education and experience (age) favors males. Second, however, the amount attributable to differences in endowments is not large, accounting for only between about 2 and about 17 percent. Third, however, the fraction of the gap attributable to observable characteristics decreased over most of the period (although it picked up again from 2000 to 2002), interpretable as a decrease in “discrimination” over the period that is, a tendency toward wages reflecting individuals productive potential (as reflected by their levels of education and experience) rather than their gender. Going one step further, one may look at how the individual observable characteristics affect the gender wage gap, effectively decomposing the overall explained gender earnings gap into the contributions from endowments of different education levels/types and experience and the returns to these endowments variables (Oaxaca, 1973; Blinder, 1973; Reimers, 1983; Cotton, 1988; Neumark, 1988). Consider Table G1 in Appendix G, which present detailed female-male wage decomposition results for 1994. Intuitively, factors which have associated with them a positive value disfavor females, while factors which have associated with them a negative value disfavor females. For the case of secondary education, for example, the gender gap consistently disfavors females consistently across the four different decompositions (Oaxaca/Blinder, Reimers, Cotton and Neumark), although the impact here is greater for lower secondary than for higher secondary. 27

Indeed, the only factors that actually work for women are the quadratic potential general experience term, completed university, technical and teacher training. Overall, for all the detailed wage decompositions as a whole (Tables G1-G5 in Appendix G), females tend to be favored by the squared potential general experience, attending university, university completion, technical and teacher education. The favorable contribution (for females) from university attendance and technical and teacher education are quite small, however.

Table 5.1 Female-Male Earnings Decompositions 1994: 1996: 1998: 2000: 2002: Mean prediction, Males (M): 7.817 7.995 7.925 7.952 8.028 Mean prediction, Females (F): 7.498 7.688 7.690 7.783 7.844 Raw differential (R) {M-F}: 0.32 0.307 0.235 0.169 0.185 - due to endowments (E): -0.038 -0.028 -0.025 -0.020 0.001 - due to coefficients (C): 0.302 0.302 0.200 0.150 0.178 - due to interaction (CE): 0.055 0.033 0.061 0.039 0.006 Oaxaca Oaxaca/Blinder Reimers Cotton Neumark 1994: Unexplained (U){C+(1-D)CE}: 0.357 0.302 0.33 0.325 0.308 Explained (V) {E+D*CE}: -0.038 0.018 -0.01 -0.005 0.011 % unexplained {U/R}: 111.8 94.4 103.1 101.6 96.5 % explained (V/R): -11.8 5.6 -3.1 -1.6 3.5 1996: Unexplained (U){C+(1-D)CE}: 0.335 0.302 0.319 0.316 0.301 Explained (V) {E+D*CE}: -0.028 0.005 -0.011 -0.008 0.006 % unexplained {U/R}: 109.0 98.3 103.6 102.7 98.0 % explained (V/R): -9.0 1.7 -3.6 -2.7 2.0 1998: Unexplained (U){C+(1-D)CE}: 0.261 0.2 0.23 0.226 0.226 Explained (V) {E+D*CE}: -0.025 0.036 0.005 0.009 0.009 % unexplained {U/R}: 110.7 84.9 97.8 96.1 96.1 % explained (V/R): -10.7 15.1 2.2 3.9 3.9 2000: Unexplained (U){C+(1-D)CE}: 0.189 0.15 0.169 0.167 0.167 Explained (V) {E+D*CE}: -0.02 0.019 0 0.001 0.002 % unexplained {U/R}: 111.7 88.8 100.3 99.1 99.1 % explained (V/R): -11.7 11.2 -0.3 0.9 0.9 2002: Unexplained (U){C+(1-D)CE}: 0.184 0.178 0.181 0.181 0.185 Explained (V) {E+D*CE}: 0.001 0.007 0.004 0.004 0 % unexplained {U/R}: 99.7 96.4 98.1 97.9 100 % explained (V/R): 0.3 3.6 1.9 2.1 0

28

It should be noted that since the amount of the observed gender wage differential which can be explained is so small, effectively the “raw” differential as given by the coefficient on the gender dummy coincides with the “adjusted” differential, that is, where endowments of characteristics have been controlled for. I will therefore not get much mileage out of pursuing gender wage gap decompositions across regions, either, and therefore merely refer to the results for the “returns to gender” discussed in the previous section (that is, where the parameter on the gender dummy was estimated across regions).

Wage Decompositions across municipal-non-municipal location Contrary to what was the case for the female-male earnings decompositions, the municipal-nonmunicipal decompositions reveal substantially larger fractions of the earnings gap being attributable to observable characteristics: while individuals from municipal areas are consistently rewarded in terms of higher earnings, a relatively larger fraction of this is due to observable characteristics than was the case for the gender earnings gap, namely between about 19 and 23 percent (Table 5.2). Similar to what was the case for the gender earnings gap, inequality between municipal and non-municipal earnings has declined over time. Moving again to the detailed wage decompositions, it is possible to decompose the municipalnon-municipal earnings gap into the contribution for the different observable characteristics and their returns. Appendix H reveals that potential general experience and lower and upper primary favor individuals from non-municipal areas, while all other factors (and their returns) favor individuals from municipal areas.

29

Table 5.2 Municipal-Non-municipal Earnings Decompositions 1994: 1996: 1998: 2000: 2002: Mean prediction, Municipal (M): 8.495 8.671 8.628 8.575 8.267 Mean prediction, Non-municipal (N): 7.264 7.468 7.424 7.415 7.412 Raw differential (R) {M-N}: 1.230 1.203 1.204 1.160 0.855 - due to endowments (E): 0.222 0.210 0.263 0.212 0.173 - due to coefficients (C): 0.967 0.924 0.954 0.917 0.654 - due to interaction (CE): 0.042 0.069 -0.013 0.031 0.028 Oaxaca Oaxaca/Blinder Reimers Cotton Neumark 1994: Unexplained (U){C+(1-D)CE}: 1.009 0.967 0.988 0.994 0.995 Explained (V) {E+D*CE}: 0.222 0.264 0.243 0.236 0.235 % unexplained {U/R}: 82 78.6 80.3 80.8 80.9 % explained (V/R): 18 21.4 19.7 19.2 19.1 1996: Unexplained (U){C+(1-D)CE}: 0.993 0.924 0.958 0.97 0.961 Explained (V) {E+D*CE}: 0.21 0.279 0.245 0.233 0.242 % unexplained {U/R}: 82.5 76.8 79.7 80.6 79.9 % explained (V/R): 17.5 23.2 20.3 19.4 20.1 1998: Unexplained (U){C+(1-D)CE}: 0.941 0.954 0.948 0.946 0.95 Explained (V) {E+D*CE}: 0.263 0.25 0.256 0.259 0.254 % unexplained {U/R}: 78.2 79.3 78.7 78.5 78.9 % explained (V/R): 21.8 20.7 21.3 21.5 21.1 2000: Unexplained (U){C+(1-D)CE}: 0.948 0.917 0.933 0.936 0.931 Explained (V) {E+D*CE}: 0.212 0.243 0.227 0.224 0.23 % unexplained {U/R}: 81.7 79.1 80.4 80.7 80.2 % explained (V/R): 18.3 20.9 19.6 19.3 19.8 2002: Unexplained (U){C+(1-D)CE}: 0.682 0.654 0.668 0.664 0.663 Explained (V) {E+D*CE}: 0.173 0.202 0.188 0.191 0.192 % unexplained {U/R}: 79.7 76.4 78.1 77.7 77.6 % explained (V/R): 20.3 23.6 21.9 22.3 22.4

6. Conclusion This paper examines education returns in Thailand over the period 1994-2002 addressing several weaknesses of previous studies related to the potential endogeneity of education, the correct interpretation of coefficients for dummy variables in semi-logarithmic regressions, testing for structural differences in education returns over time and inclusion of relevant explanatory variables in wage regressions are addressed. First, the developments in the stocks of human capital in terms of educational attainment of the workforce wage are examined and wage equations over this period are estimated. In so doing, 30

The differential impact of education on wages across gender, municipal-non-municipal location and region is examined. The results imply that the returns to education in Thailand over the period are substantial. Education returns peaked in 1999, indicating that the East Asian crisis did not have its full impact until after 2000. Additionally, the differential impact of education on wages across gender favors females, probably due to their lower human capital stock, especially for levels of education up to and including upper secondary education. Second, the paper examines the wage gap related to gender and municipal-non-municipal location by providing a wide range of alternative wage decompositions in the Blinder-Oaxaca tradition. The wage gaps related to gender and municipal-non-municipal location have decreased over time, while at the same time the share of the gap due to unobservable factors (though remaining large) has declined, as well, indicating that wage-discrimination has decreased in Thailand over the period. The main policy implication is that education should continue to play a major part on the political agenda in Thailand due to their high return. This is especially true for secondary and tertiary education. Also, the analyses revealed that only few individuals who attended university dropped out without receiving a degree, offering additional support to the “success conclusion” of tertiary education in Thailand. In expanding the education system, policy makers should particularly target females, due to their relative low levels of educational attainment, especially for education below and including upper secondary education. While this study found evidence of the concave age-earnings profile typically found in the earnings equations literature, future studies could try to examine the impact of the crises on different age segments of the work force in Thailand, including analyses of how the female-male and municipal-non-municipal earnings gaps and the amount attributable to observable characteristics have changed over the period.

31

References: Becker, Gary S. (1993) Human Capital, 3rd ed., Chicago: Chicago University Press. Blaug, Mark (1974) “An Economic Analysis of Personal Earnings in Thailand”, Economic Development and Cultural Change, 23 (1): 1-31. Blaug, Mark (1976) “The Rate of Return on Investment in Education in Thailand”, Journal of Development Studies, 12 (2): 270-83. Blinder, A.S. (1973) ”Wage Discrimination: Reduced Form and Structural Estimates”, The Journal of Human Resources, 8: 436-455. Chiswick, Carmel U. (1983) “Analysis of Earnings from Household Enterprises: Methodology and Application to Thailand”, Review of Economics and Statistics, 65 (4): 658-62. Cotton, J. (1988) “On the Decomposition of Wage Differentials”, The Review of Economics and Statistics, 70: 236-243. Daymont, T.N., Andrisani, P.J. (1984) “Job Preferences, College Major, and the Gender Gap in Earnings”, The Journal of Human Resources, 19: 408-428. Dolton, P.J., Makepeace, G.H. (1986) “Sample Selection and Male-Female Earnings Differentials in the Graduate Labour Market”, Oxford Economic Papers, 38: 317-341. Grootaert, Christian (1986) “The Role of Employment and Earnings in Analyzing Levels of Living: A General Methodology with Applications to Malaysia and Thailand”, Living Standards Measurement Study Working Paper No. 27, Washington D.C: World Bank. Hausman, Jerry A, (1978) “Specification Tests in Econometrics”, Econometrica, 46 (6): 1251-71. Hawley (2001) “Changing Returns to Education in Times of Prosperity and Crisis: Private Vocation Education in Thailand”, in: Thailand—Secondary Education for Employment— Volume II: Background Papers, Report No. 22660-TH, East Asia and Pacific Region, Washington, DC: World Bank. Jones, F.L., Kelley, J. (1984) ”Decomposing Differences Between Groups. A Cautionary Note on Measuring Discrimination”, Sociological Methods and Research, 12: 323-343. Kennedy, P. (1981), “Estimation with Correctly Interpreted Dummy Variables in Semilogarithmic Equations”, American Economic Review, 71: 801. Khoman, Sirilaksana (1993) "Education Policy" in Peter G. Warr (ed.) The Thai Economy in Transition, Cambridge: Cambridge University Press. Mincer, Jacob (1974) Schooling, Experience, and Earnings, New York: Columbia University Press. 32

Moenjak, Thammarak and Christopher Worswick (2003) “Vocational Education in Thailand: A Study of Choice and Returns”, Economics of Education Review, 22: 99-107. Neumark, D. (1988) “Employers' Discriminatory Behavior and the Estimation of Wage Discrimination”, The Journal of Human Resources, 23: 279-295. Oaxaca, R. (1973) “Male-Female Wage Differentials in Urban Labor Markets”, International Economic Review, 14: 693-709. Park, Hyunjoon and Jeroen Smits (2003) “Educational Assortative Mating in Ten East and SouthEast Asian Countries: Trends 1950s-1990s”, paper presented at the conference of the International Sociological Association RC28, March 1-3, 2003, Tokyo, Japan. Psacharopoulos, George (1985) “ Returns to Education: A Further International Update and Implications”, Journal of Human Resources, 20 (4), 583-597. Psacharopoulos, George (1994) “Returns to Education: A Global Update”, World Development, 22 (9), 1325-1343. Patrinos, Harry Anthony and George Psacharopoulos (2002) “Returns to Investment in Education: A Further Update”, Policy Research Working Paper No. 2881, Washington, DC: World Bank. Reimers, C.W. (1983) “Labor Market Discrimination Against Hispanic and Black Men”, The Review of Economics and Statistics, 65: 570-579. Winsborough, H.H., Dickenson, P. (1971) “Components of Negro-White Income Differences”, Proceedings of the American Statistical Association, Social Statistics Section: 6-8. Yamauchi, Futoshi (2003) “Why Do Schooling Returns Differ So Much? Observations and Puzzles from Thailand and the Philippines”, Mimeo, Washington, DC: International Food Policy Institute.

33

Appendix A. Definition of Variables Dependent variable: Log(Wages):

Log monthly wages and salaries of individual in real terms (deflated by CPI, 1995=100).

Explanatory variables: Age: Age squared: Female:

Age of individual Age of individual squared 1 if female, zero otherwise

Education: None (reference group): Primary, lower: Primary, upper: Secondary, lower: Secondary, upper: University, attended: University, degree: Vocational: Technical: Teacher: Other education:

1 if no education completed (incl. kindergarten), 0 otherwise 1 if 1 if primary grades 1-4, 0 otherwise 1 if primary grades 5-7 0 otherwise. 1 if secondary grades 1-3 0 otherwise. 1 if secondary grades 4-6, 0 otherwise. 1 if attended university, 0 otherwise. 1 if University, degree was obtained 1 if vocational, 0 otherwise. 1 if technical, 0 otherwise. 1 if teacher, 0 otherwise. 1 if other education, 0 otherwise.

Geographical Location: Municipal7 (reference group): Non-municipal:

1 if municipal area, 0 otherwise. 1 if Non-municipal area, 0 otherwise.

Bangkok (reference group): Central: North: North-east: South:

1 if Bangkok Metropolis, 0 otherwise. 1 if Central region, 0 otherwise. 1 if Northern region, 0 otherwise. 1 if North-eastern region, 0 otherwise. 1 if Southern region, 0 otherwise.

7

Prior to the SES 45 for 2002 NSO distinguished between municipal, area, sanitary district and village, wherefore I have combined sanitary district and village into non-municipal areas to ensure consistency with SES 45.

34

Appendix B. Descriptive Statistics for Estimation Samples Table B1 Descriptive Statistics: Full Sample, Across Gender and Across Municipal-Non-Municpal Areas 1994 Full Females Males Non-municipal Municipal Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Log(Wages) 7.448 1.405 7.242 1.495 7.594 1.319 7.102 1.350 8.587 0.886 Female 0.416 0.493 1.000 0.000 0.000 0.000 0.406 0.491 0.447 0.497 Age 34.0 11.3 32.7 11.1 34.9 11.3 34.4 11.4 32.9 10.6 Age squared 1284.4 846.5 1195.7 823.9 1347.4 856.7 1311.7 864.1 1194.5 779.2 None 0.038 0.192 0.054 0.225 0.028 0.164 0.046 0.209 0.014 0.118 Primary, low 0.438 0.496 0.410 0.492 0.458 0.498 0.502 0.500 0.227 0.419 Primary, upper 0.240 0.427 0.246 0.431 0.235 0.424 0.253 0.435 0.197 0.397 Secondary, low 0.091 0.288 0.075 0.263 0.103 0.304 0.075 0.263 0.147 0.354 Secondary, upper 0.044 0.204 0.037 0.189 0.048 0.214 0.033 0.179 0.079 0.270 University, attended 0.004 0.066 0.004 0.067 0.004 0.065 0.002 0.041 0.013 0.113 University, degree 0.075 0.264 0.099 0.299 0.059 0.235 0.045 0.206 0.177 0.382 Vocational 0.033 0.179 0.033 0.180 0.033 0.178 0.020 0.140 0.076 0.265 Technical 0.027 0.163 0.032 0.177 0.024 0.152 0.017 0.130 0.060 0.238 Teacher 0.008 0.090 0.009 0.094 0.008 0.087 0.008 0.089 0.009 0.092 Other education 0.000 0.013 0.000 0.013 0.000 0.013 0.000 0.010 0.000 0.019 Municipal 0.233 0.423 0.250 0.433 0.220 0.414 0.000 0.000 1.000 0.000 Non-municipal 0.767 0.423 0.750 0.433 0.780 0.414 1.000 0.000 0.000 0.000 Bangkok 0.219 0.414 0.226 0.419 0.214 0.410 0.264 0.441 0.071 0.258 Central 0.339 0.473 0.328 0.469 0.346 0.476 0.404 0.491 0.124 0.330 North 0.198 0.399 0.209 0.406 0.191 0.393 0.229 0.420 0.096 0.295 North-East 0.099 0.298 0.082 0.274 0.111 0.314 0.103 0.305 0.084 0.277 South 0.145 0.353 0.156 0.363 0.138 0.345 0.000 0.000 0.625 0.484 No. of observations: 24179 10023 14156 15913 8266 1996 Full Females Males Non-municipal Municipal Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Log(Wages) 7.646 1.330 7.430 1.431 7.798 1.232 7.301 1.263 8.742 0.862 Female 0.413 0.492 1.000 0.000 0.000 0.000 0.404 0.491 0.440 0.496 Age 34.9 11.3 33.7 11.0 35.7 11.4 35.4 11.4 33.1 10.6 Age squared 1344.7 861.3 1259.0 826.0 1404.8 880.4 1386.8 879.4 1210.7 786.7 None 0.039 0.193 0.059 0.236 0.024 0.155 0.047 0.212 0.012 0.111 Primary, low 0.412 0.492 0.398 0.489 0.423 0.494 0.481 0.500 0.196 0.397 Primary, upper 0.247 0.431 0.246 0.431 0.248 0.432 0.260 0.439 0.205 0.404 Secondary, low 0.102 0.302 0.078 0.269 0.118 0.323 0.084 0.277 0.158 0.365 Secondary, upper 0.043 0.204 0.037 0.190 0.047 0.213 0.036 0.187 0.065 0.247 University, attended 0.004 0.063 0.006 0.077 0.003 0.050 0.002 0.039 0.012 0.107 University, degree 0.085 0.278 0.108 0.311 0.068 0.252 0.046 0.209 0.207 0.406 Vocational 0.034 0.182 0.029 0.168 0.038 0.191 0.022 0.146 0.074 0.262 Technical 0.028 0.164 0.031 0.174 0.025 0.156 0.016 0.127 0.064 0.244 Teacher 0.006 0.078 0.006 0.079 0.006 0.077 0.006 0.077 0.006 0.080 Other education 0.000 0.011 0.000 0.011 0.000 0.011 0.000 0.010 0.000 0.013 Municipal 0.240 0.427 0.256 0.436 0.228 0.420 0.000 0.000 1.000 0.000 Non-municipal 0.760 0.427 0.744 0.436 0.772 0.420 1.000 0.000 0.000 0.000 Bangkok 0.151 0.358 0.160 0.367 0.144 0.351 0.000 0.000 0.630 0.483 Central 0.250 0.433 0.265 0.442 0.239 0.426 0.279 0.449 0.156 0.363 North 0.201 0.401 0.208 0.406 0.197 0.398 0.246 0.431 0.059 0.236 North-East 0.295 0.456 0.277 0.447 0.308 0.462 0.360 0.480 0.087 0.282 South 0.103 0.304 0.090 0.286 0.112 0.316 0.114 0.318 0.069 0.253 No. of observations: 24015 9979 14036 16041 7974

35

1998

Full Females Males Non-municipal Municipal Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Log(Wages) 7.625 1.388 7.479 1.471 7.736 1.312 7.257 1.331 8.755 0.843 Female 0.430 0.495 1.000 0.000 0.000 0.000 0.411 0.492 0.489 0.500 Age 35.1 10.9 34.1 10.6 35.9 11.1 35.3 11.1 34.5 10.4 Age squared 1352.1 827.6 1275.5 789.4 1409.8 850.7 1369.5 839.7 1298.8 787.1 None 0.033 0.180 0.047 0.213 0.023 0.150 0.041 0.199 0.010 0.099 Primary, low 0.355 0.478 0.341 0.474 0.365 0.481 0.418 0.493 0.162 0.368 Primary, upper 0.246 0.430 0.235 0.424 0.254 0.435 0.272 0.445 0.163 0.370 Secondary, low 0.119 0.323 0.102 0.303 0.131 0.338 0.107 0.310 0.153 0.360 Secondary, upper 0.057 0.231 0.049 0.216 0.062 0.241 0.051 0.219 0.075 0.263 University, attended 0.005 0.073 0.008 0.087 0.004 0.061 0.003 0.055 0.013 0.112 University, degree 0.106 0.308 0.138 0.344 0.082 0.274 0.057 0.231 0.256 0.436 Vocational 0.038 0.191 0.035 0.184 0.040 0.196 0.024 0.152 0.081 0.273 Technical 0.036 0.187 0.040 0.195 0.034 0.181 0.022 0.147 0.080 0.272 Teacher 0.006 0.075 0.006 0.077 0.005 0.073 0.005 0.072 0.007 0.083 Other education 0.000 0.010 0.000 0.009 0.000 0.011 0.000 0.009 0.000 0.012 Municipal 0.246 0.431 0.280 0.449 0.221 0.415 0.000 0.000 1.000 0.000 Non-municipal 0.754 0.431 0.720 0.449 0.779 0.415 1.000 0.000 0.000 0.000 Bangkok 0.155 0.362 0.178 0.382 0.137 0.344 0.000 0.000 0.628 0.483 Central 0.241 0.427 0.260 0.439 0.226 0.418 0.265 0.442 0.165 0.371 North 0.197 0.398 0.198 0.399 0.196 0.397 0.244 0.429 0.054 0.227 North-East 0.304 0.460 0.275 0.447 0.327 0.469 0.376 0.484 0.085 0.279 South 0.103 0.304 0.089 0.284 0.114 0.318 0.115 0.319 0.067 0.250 No. of observations: 22820 9919 12901 15261 7559 2000 Full Females Males Non-municipal Municipal Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Log(Wages) 7.583 1.443 7.473 1.506 7.670 1.385 7.186 1.392 8.724 0.863 Female 0.444 0.497 1.000 0.000 0.000 0.000 0.426 0.494 0.495 0.500 Age 35.7 10.9 34.9 10.6 36.3 11.0 36.1 11.0 34.4 10.4 Age squared 1390.8 830.6 1330.2 803.1 1439.2 848.7 1426.5 844.4 1288.2 780.5 None 0.031 0.172 0.042 0.201 0.021 0.144 0.038 0.191 0.009 0.094 Primary, low 0.333 0.471 0.327 0.469 0.338 0.473 0.396 0.489 0.151 0.358 Primary, upper 0.250 0.433 0.236 0.424 0.261 0.439 0.276 0.447 0.174 0.379 Secondary, low 0.122 0.327 0.104 0.305 0.136 0.343 0.113 0.316 0.146 0.353 Secondary, upper 0.068 0.252 0.062 0.242 0.073 0.259 0.057 0.232 0.099 0.299 University, attended 0.008 0.091 0.009 0.094 0.008 0.088 0.004 0.066 0.020 0.139 University, degree 0.111 0.314 0.142 0.349 0.086 0.281 0.062 0.241 0.251 0.434 Vocational 0.034 0.180 0.030 0.171 0.036 0.187 0.023 0.149 0.065 0.246 Technical 0.039 0.193 0.042 0.201 0.036 0.185 0.024 0.155 0.079 0.270 Teacher 0.005 0.070 0.005 0.071 0.005 0.069 0.005 0.068 0.005 0.074 Other education 0.001 0.028 0.001 0.034 0.001 0.023 0.001 0.032 0.000 0.009 Municipal 0.258 0.437 0.288 0.453 0.234 0.423 0.000 0.000 1.000 0.000 Non-municipal 0.742 0.437 0.712 0.453 0.766 0.423 1.000 0.000 0.000 0.000 Bangkok 0.162 0.368 0.182 0.386 0.145 0.352 0.000 0.000 0.627 0.484 Central 0.242 0.428 0.257 0.437 0.230 0.421 0.268 0.443 0.167 0.373 North 0.201 0.401 0.201 0.401 0.201 0.401 0.252 0.434 0.054 0.225 North-East 0.292 0.454 0.271 0.444 0.308 0.462 0.363 0.481 0.085 0.279 South 0.104 0.305 0.089 0.285 0.116 0.320 0.117 0.321 0.068 0.251 No. of observations: 23983 10820 13163 14453 9530

36

2002

Full Females Males Non-municipal Municipal Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Log(Wages) 7.729 1.342 7.612 1.410 7.825 1.275 7.259 1.282 8.548 1.010 female 0.452 0.498 1.000 0.000 0.000 0.000 0.434 0.496 0.483 0.500 Age 36.3 10.8 35.6 10.6 36.8 11.0 36.9 11.0 35.2 10.5 Age squared 1433.7 837.1 1379.7 811.8 1478.3 854.8 1483.6 856.5 1346.8 794.8 None 0.035 0.185 0.046 0.210 0.027 0.161 0.046 0.209 0.018 0.131 Primary, low 0.311 0.463 0.309 0.462 0.313 0.464 0.392 0.488 0.170 0.376 Primary, upper 0.234 0.423 0.223 0.416 0.243 0.429 0.261 0.439 0.185 0.389 Secondary, low 0.132 0.339 0.112 0.315 0.149 0.356 0.122 0.327 0.149 0.356 Secondary, upper 0.071 0.257 0.066 0.249 0.075 0.264 0.057 0.231 0.097 0.295 University, attended 0.006 0.074 0.006 0.079 0.005 0.069 0.002 0.050 0.011 0.103 University, degree 0.126 0.331 0.160 0.366 0.097 0.297 0.064 0.244 0.233 0.423 Vocational 0.038 0.190 0.032 0.176 0.042 0.201 0.024 0.152 0.061 0.240 Technical 0.045 0.207 0.043 0.202 0.047 0.211 0.029 0.168 0.072 0.258 Teacher 0.003 0.056 0.003 0.059 0.003 0.053 0.003 0.053 0.004 0.060 Other education 0.000 0.017 0.001 0.023 0.000 0.012 0.000 0.018 0.000 0.017 Municipal 0.365 0.481 0.390 0.488 0.344 0.475 0.000 0.000 1.000 0.000 Non-municipal 0.635 0.481 0.610 0.488 0.656 0.475 1.000 0.000 0.000 0.000 Bangkok 0.164 0.371 0.182 0.386 0.150 0.357 0.000 0.000 0.451 0.498 Central 0.248 0.432 0.259 0.438 0.240 0.427 0.251 0.434 0.244 0.429 North 0.182 0.386 0.182 0.386 0.182 0.386 0.231 0.422 0.096 0.295 North-East 0.291 0.454 0.274 0.446 0.305 0.460 0.381 0.486 0.134 0.340 South 0.114 0.318 0.103 0.305 0.123 0.329 0.137 0.344 0.075 0.264 No. of observations: 33320 15072 18248 12556 20764 Notes: Calculations apply sample weights.

37

Table B2 Descriptive Statistics: Full Sample and Across Regions Full Bangkok Central North North-East South Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Log(Wages) 7.448 1.405 6.944 1.332 7.003 1.491 7.681 1.123 7.805 1.096 8.685 0.814 Female 0.416 0.493 0.430 0.495 0.402 0.490 0.437 0.496 0.343 0.475 0.446 0.497 Age 34.0 11.3 35.8 11.5 33.8 11.3 34.1 11.3 33.7 11.0 32.0 10.4 Age squared 1284.4 846.5 1415.1 889.5 1270.8 843.6 1291.7 853.3 1255.2 825.9 1129.4 757.4 None 0.038 0.192 0.081 0.273 0.022 0.145 0.037 0.189 0.045 0.207 0.011 0.105 Primary, low 0.438 0.496 0.538 0.499 0.489 0.500 0.431 0.495 0.371 0.483 0.223 0.416 Primary, upper 0.240 0.427 0.205 0.404 0.263 0.440 0.248 0.432 0.242 0.428 0.226 0.418 Secondary, low 0.091 0.288 0.059 0.236 0.073 0.260 0.107 0.309 0.108 0.311 0.151 0.358 Secondary, upper 0.044 0.204 0.027 0.162 0.035 0.185 0.049 0.216 0.048 0.213 0.078 0.269 University, attended 0.004 0.066 0.001 0.036 0.003 0.057 0.001 0.038 0.005 0.070 0.015 0.121 University, degree 0.075 0.264 0.053 0.224 0.057 0.231 0.060 0.237 0.090 0.287 0.165 0.371 Vocational 0.033 0.179 0.015 0.121 0.023 0.149 0.036 0.188 0.040 0.196 0.075 0.264 Technical 0.027 0.163 0.015 0.123 0.022 0.148 0.024 0.154 0.041 0.199 0.052 0.221 Teacher 0.008 0.090 0.005 0.072 0.013 0.113 0.006 0.074 0.009 0.094 0.005 0.067 Other education 0.000 0.013 0.000 0.022 0.000 0.000 0.000 0.007 0.000 0.000 0.000 0.018 Municipal 0.233 0.423 0.076 0.265 0.086 0.280 0.113 0.316 0.197 0.398 1.000 0.000 Non-municipal 0.767 0.423 0.924 0.265 0.914 0.280 0.887 0.316 0.803 0.398 0.000 0.000 Bangkok 0.219 0.414 1.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Central 0.339 0.473 0.000 0.000 1.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 North 0.198 0.399 0.000 0.000 0.000 0.000 1.000 0.000 0.000 0.000 0.000 0.000 North-East 0.099 0.298 0.000 0.000 0.000 0.000 0.000 0.000 1.000 0.000 0.000 0.000 South 0.145 0.353 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 1.000 0.000 No. of observations: 24179 3566 7453 5441 3292 2373 1996 Full Bangkok Central North North-East South Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Log(Wages) 7.646 1.330 8.784 0.821 8.047 1.064 7.188 1.259 6.995 1.354 7.771 1.085 Female 0.413 0.492 0.437 0.496 0.438 0.496 0.426 0.495 0.387 0.487 0.361 0.480 Age 34.9 11.3 32.3 10.6 34.5 11.3 36.7 11.5 35.3 11.1 35.0 11.5 Age squared 1344.7 861.3 1154.0 779.8 1314.8 864.8 1479.2 905.7 1371.9 841.0 1354.9 881.8 None 0.039 0.193 0.012 0.110 0.034 0.182 0.078 0.268 0.024 0.152 0.056 0.230 Primary, low 0.412 0.492 0.198 0.398 0.378 0.485 0.509 0.500 0.501 0.500 0.370 0.483 Primary, upper 0.247 0.431 0.237 0.425 0.248 0.432 0.209 0.406 0.276 0.447 0.254 0.435 Secondary, low 0.102 0.302 0.160 0.366 0.127 0.332 0.069 0.253 0.068 0.252 0.117 0.322 Secondary, upper 0.043 0.204 0.059 0.236 0.052 0.222 0.034 0.180 0.032 0.177 0.049 0.217 University, attended 0.004 0.063 0.011 0.104 0.004 0.062 0.002 0.048 0.003 0.051 0.001 0.035 University, degree 0.085 0.278 0.191 0.393 0.078 0.268 0.055 0.228 0.060 0.238 0.073 0.260 Vocational 0.034 0.182 0.075 0.263 0.046 0.210 0.018 0.133 0.015 0.122 0.033 0.179 Technical 0.028 0.164 0.053 0.225 0.030 0.170 0.021 0.143 0.015 0.121 0.036 0.185 Teacher 0.006 0.078 0.004 0.064 0.004 0.062 0.006 0.077 0.007 0.084 0.011 0.106 Other education 0.000 0.011 0.000 0.016 0.000 0.011 0.000 0.016 0.000 0.000 0.000 0.000 Municipal 0.240 0.427 1.000 0.000 0.150 0.357 0.070 0.255 0.071 0.256 0.160 0.366 Non-municipal 0.760 0.427 0.000 0.000 0.850 0.357 0.930 0.255 0.929 0.256 0.840 0.366 Bangkok 0.151 0.358 1.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Central 0.250 0.433 0.000 0.000 1.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 North 0.201 0.401 0.000 0.000 0.000 0.000 1.000 0.000 0.000 0.000 0.000 0.000 North-East 0.295 0.456 0.000 0.000 0.000 0.000 0.000 0.000 1.000 0.000 0.000 0.000 South 0.103 0.304 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 1.000 0.000 No. of observations: 24015 1296 6736 5415 6386 3443 1994

38

1998

Full Bangkok Central North North-East South Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Log(Wages) 7.625 1.388 8.850 0.812 8.056 1.053 7.202 1.291 6.844 1.432 7.899 0.984 Female 0.430 0.495 0.494 0.500 0.465 0.499 0.433 0.496 0.389 0.488 0.370 0.483 Age 35.1 10.9 34.2 10.4 34.5 11.1 36.6 10.8 35.1 11.1 35.0 11.1 Age squared 1352.1 827.6 1275.9 780.2 1312.1 837.6 1452.4 836.7 1358.1 829.7 1349.7 831.1 None 0.033 0.180 0.008 0.090 0.026 0.160 0.073 0.259 0.021 0.142 0.051 0.220 Primary, low 0.355 0.478 0.165 0.371 0.322 0.467 0.455 0.498 0.428 0.495 0.307 0.461 Primary, upper 0.246 0.430 0.177 0.382 0.243 0.429 0.209 0.406 0.310 0.462 0.236 0.425 Secondary, low 0.119 0.323 0.156 0.363 0.149 0.356 0.089 0.285 0.093 0.290 0.125 0.331 Secondary, upper 0.057 0.231 0.066 0.249 0.070 0.255 0.041 0.198 0.039 0.193 0.094 0.292 University, attended 0.005 0.073 0.014 0.118 0.005 0.074 0.004 0.062 0.002 0.045 0.005 0.073 University, degree 0.106 0.308 0.257 0.437 0.089 0.284 0.075 0.263 0.067 0.250 0.092 0.289 Vocational 0.038 0.191 0.083 0.276 0.049 0.215 0.022 0.145 0.017 0.131 0.036 0.186 Technical 0.036 0.187 0.069 0.254 0.043 0.204 0.027 0.162 0.017 0.128 0.046 0.210 Teacher 0.006 0.075 0.004 0.062 0.004 0.063 0.007 0.084 0.007 0.081 0.007 0.082 Other education 0.000 0.010 0.000 0.015 0.000 0.006 0.000 0.017 0.000 0.000 0.000 0.000 Municipal 0.246 0.431 1.000 0.000 0.169 0.374 0.068 0.252 0.069 0.253 0.160 0.367 Non-municipal 0.754 0.431 0.000 0.000 0.831 0.374 0.932 0.252 0.931 0.253 0.840 0.367 Bangkok 0.155 0.362 1.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Central 0.241 0.427 0.000 0.000 1.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 North 0.197 0.398 0.000 0.000 0.000 0.000 1.000 0.000 0.000 0.000 0.000 0.000 North-East 0.304 0.460 0.000 0.000 0.000 0.000 0.000 0.000 1.000 0.000 0.000 0.000 South 0.103 0.304 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 1.000 0.000 No. of observations: 22820 1199 5991 5142 6377 3297 2000 Full Bangkok Central North North-East South Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Log(Wages) 7.583 1.443 8.820 0.832 8.147 1.000 6.992 1.319 6.728 1.497 7.885 1.025 Female 0.444 0.497 0.501 0.500 0.471 0.499 0.444 0.497 0.412 0.492 0.379 0.485 Age 35.7 10.9 33.8 10.2 34.7 10.6 37.5 11.0 36.4 11.1 35.2 10.7 Age squared 1390.8 830.6 1247.5 767.2 1318.9 808.9 1524.3 857.0 1450.2 852.6 1356.5 806.9 None 0.031 0.172 0.007 0.084 0.024 0.152 0.068 0.251 0.020 0.138 0.042 0.201 Primary, low 0.333 0.471 0.144 0.351 0.299 0.458 0.442 0.497 0.413 0.492 0.271 0.445 Primary, upper 0.250 0.433 0.189 0.391 0.230 0.421 0.234 0.424 0.313 0.464 0.244 0.430 Secondary, low 0.122 0.327 0.147 0.354 0.158 0.365 0.082 0.275 0.101 0.301 0.130 0.336 Secondary, upper 0.068 0.252 0.095 0.293 0.078 0.268 0.055 0.227 0.048 0.214 0.086 0.281 University, attended 0.008 0.091 0.025 0.156 0.005 0.067 0.005 0.068 0.005 0.071 0.008 0.089 University, degree 0.111 0.314 0.254 0.435 0.101 0.302 0.065 0.247 0.066 0.248 0.126 0.332 Vocational 0.034 0.180 0.063 0.243 0.050 0.218 0.017 0.130 0.014 0.117 0.037 0.188 Technical 0.039 0.193 0.073 0.260 0.052 0.222 0.027 0.162 0.016 0.125 0.040 0.195 Teacher 0.005 0.070 0.003 0.059 0.004 0.063 0.005 0.069 0.005 0.070 0.009 0.095 Other education 0.001 0.028 0.000 0.000 0.000 0.009 0.000 0.000 0.000 0.000 0.007 0.086 Municipal 0.258 0.437 1.000 0.000 0.178 0.383 0.069 0.253 0.075 0.264 0.168 0.374 Non-municipal 0.742 0.437 0.000 0.000 0.822 0.383 0.931 0.253 0.925 0.264 0.832 0.374 Bangkok 0.162 0.368 1.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Central 0.242 0.428 0.000 0.000 1.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 North 0.201 0.401 0.000 0.000 0.000 0.000 1.000 0.000 0.000 0.000 0.000 0.000 North-East 0.292 0.454 0.000 0.000 0.000 0.000 0.000 0.000 1.000 0.000 0.000 0.000 South 0.104 0.305 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 1.000 0.000 No. of observations: 23983 1276 6690 5391 6201 3507

39

2002

Full Bangkok Central North North-East South Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Log(Wages) 7.729 1.342 8.817 0.852 8.187 1.000 7.241 1.227 6.981 1.416 7.847 1.025 Female 0.452 0.498 0.500 0.500 0.471 0.499 0.452 0.498 0.425 0.494 0.409 0.492 Age 36.3 10.8 34.4 10.3 34.9 10.8 38.4 10.9 37.5 10.7 35.5 11.1 Age squared 1433.7 837.1 1290.0 774.9 1335.2 816.2 1594.6 860.3 1519.4 841.2 1380.5 852.3 None 0.035 0.185 0.014 0.117 0.026 0.159 0.087 0.282 0.021 0.142 0.042 0.201 Primary, low 0.311 0.463 0.134 0.341 0.260 0.438 0.405 0.491 0.410 0.492 0.275 0.446 Primary, upper 0.234 0.423 0.193 0.395 0.217 0.412 0.211 0.408 0.277 0.448 0.253 0.435 Secondary, low 0.132 0.339 0.148 0.356 0.177 0.381 0.093 0.290 0.109 0.311 0.133 0.340 Secondary, upper 0.071 0.257 0.098 0.298 0.089 0.285 0.053 0.225 0.048 0.214 0.079 0.269 University, attended 0.006 0.074 0.014 0.119 0.004 0.064 0.005 0.071 0.002 0.046 0.005 0.072 University, degree 0.126 0.331 0.261 0.439 0.117 0.321 0.088 0.283 0.083 0.275 0.118 0.323 Vocational 0.038 0.190 0.068 0.252 0.052 0.223 0.021 0.143 0.020 0.139 0.034 0.180 Technical 0.045 0.207 0.067 0.250 0.055 0.228 0.033 0.179 0.026 0.161 0.056 0.231 Teacher 0.003 0.056 0.001 0.033 0.003 0.055 0.003 0.056 0.004 0.063 0.004 0.062 Other education 0.000 0.017 0.000 0.016 0.000 0.003 0.001 0.026 0.000 0.021 0.000 0.005 Municipal 0.365 0.481 1.000 0.000 0.358 0.479 0.193 0.395 0.168 0.374 0.240 0.427 Non-municipal 0.635 0.481 0.000 0.000 0.642 0.479 0.807 0.395 0.832 0.374 0.760 0.427 Bangkok 0.164 0.371 1.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Central 0.248 0.432 0.000 0.000 1.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 North 0.182 0.386 0.000 0.000 0.000 0.000 1.000 0.000 0.000 0.000 0.000 0.000 North-East 0.291 0.454 0.000 0.000 0.000 0.000 0.000 0.000 1.000 0.000 0.000 0.000 South 0.114 0.318 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 1.000 0.000 No. of observations: 33320 1485 10830 6941 7958 5013 Notes: Calculations apply sample weights.

40

Appendix C. Earnings Equations Results for Full Sample Table C1. Results for Full Sample OLS: 1994 Female -0.394*** [0.020] Age 0.090*** [0.006] Age squared -0.001*** [0.000] Primary, low 0.370*** [0.054] Primary, upper 0.873*** [0.058] Secondary, low 1.307*** [0.060] Secondary, upper 1.578*** [0.066] University, attended 1.448*** [0.112] University, degree 2.394*** [0.056] Vocational 1.766*** [0.065] Technical 2.065*** [0.065] Teacher 2.399*** [0.077] Other education -0.189 [0.690] Non-municipal -0.653*** [0.019] Central -0.047* [0.028] North 0.574*** [0.025] North-east 0.507*** [0.030] South 0.629*** [0.029] Constant 5.254*** [0.116] Observations 24179 R-squared 0.46 Specification tests:8 LR tests for sample split across years Hausman test –stat., -χ2(12) for RE vs. FE: F-test,for all ui = 0:

8

--

1996 -0.402*** [0.018] 0.081*** [0.005] -0.001*** [0.000] 0.476*** [0.052] 0.885*** [0.056] 1.276*** [0.058] 1.468*** [0.064] 1.829*** [0.112] 2.351*** [0.057] 1.702*** [0.064] 1.949*** [0.063] 2.358*** [0.103] 0.500 [0.734] -0.633*** [0.018] 0.101*** [0.027] -0.555*** [0.030] -0.801*** [0.029] -0.192*** [0.031] 6.083*** [0.108] 24015 0.51

---

1998 -0.376*** [0.018] 0.086*** [0.005] -0.001*** [0.000] 0.288*** [0.051] 0.739*** [0.055] 1.195*** [0.060] 1.381*** [0.061] 1.580*** [0.089] 2.208*** [0.054] 1.496*** [0.066] 1.730*** [0.061] 2.151*** [0.124] 0.964*** [0.371] -0.552*** [0.019] 0.055** [0.025] -0.611*** [0.029] -0.977*** [0.028] -0.142*** [0.029] 5.982*** [0.118] 22820 0.52

2000 -0.307*** [0.019] 0.083*** [0.006] -0.001*** [0.000] 0.283*** [0.060] 0.755*** [0.063] 1.211*** [0.065] 1.374*** [0.070] 1.638*** [0.138] 2.250*** [0.061] 1.529*** [0.070] 1.736*** [0.069] 2.209*** [0.129] 1.606*** [0.190] -0.605*** [0.018] 0.166*** [0.024] -0.730*** [0.029] -1.028*** [0.029] -0.151*** [0.028] 5.890*** [0.130] 23983 0.53

χ2(72) = 7392.87*** ----

--

2002 -0.304*** [0.016] 0.097*** [0.005] -0.001*** [0.000] 0.206*** [0.046] 0.686*** [0.049] 1.030*** [0.050] 1.204*** [0.053] 1.329*** [0.103] 2.083*** [0.047] 1.359*** [0.060] 1.653*** [0.053] 2.124*** [0.109] 1.598*** [0.131] -0.504*** [0.014] -0.027 [0.019] -0.725*** [0.022] -0.999*** [0.024] -0.286*** [0.023] 5.756*** [0.105] 33320 0.53

---

See Appendix I for details on the specification tests.

41

Fixed effects specification: 1994 1996 1998 -0.398*** -0.385*** -0.343*** [0.012] [0.012] [0.012] 0.067*** 0.062*** 0.068*** [0.005] [0.004] [0.005] -0.001*** -0.001*** -0.001*** [0.000] [0.000] [0.000] 0.180*** 0.162*** 0.092** [0.043] [0.041] [0.043] 0.463*** 0.446*** 0.445*** [0.049] [0.045] [0.048] 0.564*** 0.492*** 0.567*** [0.053] [0.050] [0.051] 0.588*** 0.554*** 0.615*** [0.060] [0.057] [0.057] 0.450*** 0.666*** 0.457*** [0.137] [0.121] [0.121] 1.037*** 0.988*** 0.990*** [0.063] [0.058] [0.059] 0.675*** 0.663*** 0.610*** [0.064] [0.060] [0.063] 0.729*** 0.797*** 0.687*** [0.067] [0.062] [0.063] 1.035*** 0.888*** 0.939*** [0.100] [0.097] [0.103] -0.385 0.259 0.63 [0.598] [0.895] [0.907] ----

2000 -0.268*** [0.011] 0.067*** [0.004] -0.001*** [0.000] 0.168*** [0.046] 0.538*** [0.049] 0.642*** [0.052] 0.704*** [0.056] 0.557*** [0.093] 1.098*** [0.058] 0.709*** [0.061] 0.817*** [0.060] 0.993*** [0.104] 0.762 [0.628] --

2002 -0.278*** [0.010] 0.077*** [0.004] -0.001*** [0.000] 0.142*** [0.039] 0.478*** [0.041] 0.597*** [0.043] 0.626*** [0.047] 0.620*** [0.092] 1.043*** [0.049] 0.648*** [0.052] 0.792*** [0.051] 1.033*** [0.107] 0.774* [0.409] --

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

6.098*** [0.099] 24181 0.33

6.344*** [0.096] 24015 0.33

6.104*** [0.100] 22820 0.33

5.948*** [0.098] 23983 0.33

5.879*** [0.084] 33320 0.33

χ2(52) = 582.99*** 794.02*** 803.03*** 769.05*** 872.14*** 1052.11*** F(15478, 8689) = 2.86***

F(15454, F(14478, 8547) = 8328) = 2.66*** 2.74***

F(15001, F(21071, 8968) = 12235) = 3.04 *** 2.58***

Notes: Numbers in brackets are standard deviations. The OLS-specification applies sample weights. For the 1994 OLSspecification, two observations drop out due to missing weights. *: statistically significant at 10 percent; **: statistically significant at 5 percent; ***: statistically significant at 1 percent. – indicates that result is not available due to locational variables drop out of the fixed effects specification, since they do not vary within households.

Table C2. Additional Tests for Sample Split, Fixed Effects Specification 1994 LR test for sample split across: Gender

χ2(12) = 23900.83***

Municipal-non-municipal location χ2(13) = 1250.973*** χ2(49) = 1288.067***

1996

1998

2000

2002

χ2(10) = 17770.37*** χ2(11) = 19784.97*** χ2(10) = 18804.56*** χ2(11) = 27954.99*** χ2(12) = 1404.914*** χ2(12) = 1239.311*** χ2(13) = 786.4292*** χ2(13) = 488.142*** χ2(48) = 1270.349 *** χ2(48) = 1343.154*** χ2(49) = 891.4716*** χ2(50) = 967.4602***

Region

Notes: ***: statistically significant at 1 percent.

42

Appendix D. Earnings Equations Results Across Female-Male Sub-samples Table D1. Results for Female Sub-sample OLS: 1994 1996 1998 2000 2002 Age 0.082*** 0.067*** 0.086*** 0.070*** 0.095*** [0.009] [0.008] [0.009] [0.009] [0.008] Age squared -0.001*** -0.001*** -0.001*** -0.001*** -0.001*** [0.000] [0.000] [0.000] [0.000] [0.000] Primary, low 0.264*** 0.321*** 0.110* 0.195** 0.106* [0.071] [0.072] [0.065] [0.079] [0.064] Primary, upper 0.894*** 0.789*** 0.603*** 0.718*** 0.621*** [0.081] [0.082] [0.074] [0.086] [0.069] Secondary, low 1.362*** 1.202*** 1.113*** 1.280*** 1.075*** [0.087] [0.086] [0.081] [0.089] [0.075] Secondary, upper 1.719*** 1.495*** 1.330*** 1.422*** 1.238*** [0.091] [0.097] [0.086] [0.097] [0.079] University, attended 1.662*** 1.595*** 1.501*** 1.628*** 1.203*** [0.157] [0.161] [0.108] [0.209] [0.158] University, degree 2.498*** 2.319*** 2.136*** 2.227*** 2.047*** [0.073] [0.077] [0.068] [0.081] [0.065] Vocational 1.911*** 1.666*** 1.321*** 1.607*** 1.437*** [0.090] [0.094] [0.094] [0.099] [0.079] Technical 2.159*** 1.818*** 1.564*** 1.757*** 1.646*** [0.091] [0.089] [0.079] [0.092] [0.075] Teacher 2.459*** 2.509*** 2.204*** 2.073*** 2.028*** [0.130] [0.178] [0.158] [0.191] [0.167] Other education 0.719** -0.799* 1.373*** 1.653*** 1.481*** [0.362] [0.484] [0.070] [0.282] [0.160] Non-municipal -0.588*** -0.656*** -0.574*** -0.601*** -0.521*** [0.031] [0.028] [0.029] [0.027] [0.021] Central -0.042 0.166*** 0.120*** 0.207*** -0.002 [0.045] [0.039] [0.038] [0.034] [0.026] North 0.654*** -0.598*** -0.618*** -0.712*** -0.749*** [0.039] [0.045] [0.044] [0.042] [0.033] North-east 0.516*** -0.908*** -1.027*** -1.059*** -1.088*** [0.052] [0.045] [0.043] [0.044] [0.036] South 0.683*** -0.227*** -0.080* -0.172*** -0.309*** [0.045] [0.047] [0.042] [0.041] [0.034] Constant 5.070*** 6.148*** 5.871*** 5.953*** 5.666*** [0.179] [0.170] [0.177] [0.188] [0.157] Observations 10023 9979 9919 10820 15072 R-squared 0.51 0.54 0.55 0.57 0.57 Specification tests:9 Hausman test –stat., χ2(10) for RE vs. FE:

--

--

--

--

--

F-test,for all ui = 0:

--

--

--

--

--

Fixed effects specification:

1994 1996 0.061*** 0.069*** [0.010] [0.010] -0.001*** -0.001*** [0.000] [0.000] 0.261*** 0.031 [0.100] [0.103] 0.651*** 0.654*** [0.115] [0.114] 0.949*** 0.678*** [0.126] [0.122] 0.941*** 0.588*** [0.142] [0.138] 0.893*** 0.841*** [0.277] [0.224] 1.582*** 1.244*** [0.133] [0.128] 0.958*** 0.676*** [0.142] [0.157] 1.071*** 1.056*** [0.149] [0.146] 1.854*** 1.362*** [0.237] [0.245] -0.61 -[0.826] -----------

-----------

1998 2000 2002 0.081*** 0.085*** 0.080*** [0.010] [0.009] [0.008] -0.001*** -0.001*** -0.001*** [0.000] [0.000] [0.000] -0.193* 0.344*** 0.160* [0.105] [0.104] [0.091] 0.410*** 0.914*** 0.785*** [0.114] [0.111] [0.094] 0.590*** 1.044*** 0.965*** [0.122] [0.117] [0.099] 0.738*** 0.995*** 1.051*** [0.133] [0.123] [0.107] 0.231 0.750*** 0.702*** [0.241] [0.183] [0.197] 1.056*** 1.614*** 1.342*** [0.128] [0.119] [0.102] 0.437*** 1.173*** 1.074*** [0.159] [0.139] [0.120] 0.588*** 1.312*** 1.150*** [0.138] [0.128] [0.116] 1.261*** 1.588*** 1.411*** [0.281] [0.251] [0.276] -------------

-------------

-------------

5.641*** 5.794*** 5.724*** 5.094*** 5.302*** [0.217] [0.214] [0.219] [0.199] [0.180] 10024 9979 9919 10820 15072 0.42 0.39 0.39 0.38 0.37

152.99*** F(8673, 1338) = 2.50***

180.83*** F(8737, 1230) = 2.39***

189.98*** F(8672, 1235) = 2.30***

196.66*** F(9431, 1377) = 2.74***

250.08*** F(13282, 1778) = 2.10***

Notes: Numbers in brackets are standard deviations. The OLS-specification applies sample weights. *: statistically significant at 10 percent; **: statistically significant at 5 percent; ***: statistically significant at 1 percent.

9

See Appendix I for details on the specification tests.

43

Table D2. Results for Male Sub-sample 1994 0.106*** [0.007] -0.001*** [0.000] 0.370*** [0.081] 0.802*** [0.084] 1.210*** [0.086] 1.435*** [0.094] 1.238*** [0.152] 2.218*** [0.085] 1.596*** [0.094] 1.904*** [0.092] 2.320*** [0.102] -1.062** [0.517] -0.695*** [0.024] -0.055 [0.035] 0.501*** [0.033] 0.477*** [0.036] 0.576*** [0.037] 4.986*** [0.151]

1996 0.097*** [0.006] -0.001*** [0.000] 0.605*** [0.074] 0.984*** [0.077] 1.353*** [0.079] 1.493*** [0.087] 2.071*** [0.147] 2.356*** [0.084] 1.772*** [0.087] 2.067*** [0.090] 2.286*** [0.118] 1.498*** [0.278] -0.614*** [0.022] 0.043 [0.036] -0.521*** [0.039] -0.732*** [0.038] -0.171*** [0.040] 5.592*** [0.135]

OLS: 1998 0.093*** [0.007] -0.001*** [0.000] 0.422*** [0.084] 0.854*** [0.087] 1.258*** [0.092] 1.434*** [0.093] 1.561*** [0.144] 2.233*** [0.088] 1.628*** [0.099] 1.850*** [0.096] 2.115*** [0.193] 0.813*** [0.315] -0.530*** [0.024] -0.001 [0.034] -0.597*** [0.040] -0.944*** [0.037] -0.182*** [0.040] 5.628*** [0.156]

14156 0.42

14036 0.47

12901 0.49

13163 0.5

18248 0.49

χ2(12) for RE vs. FE:

--

--

--

--

--

F-test,for all ui = 0:

--

--

--

--

--

Age Age squared Primary, low Primary, upper Secondary, low Secondary, upper University, attended University, degree Vocational Technical Teacher Other education Non-municipal Central North North-east South Constant Observations R-squared

Fixed effects specification:

2000 0.100*** [0.008] -0.001*** [0.000] 0.321*** [0.090] 0.757*** [0.093] 1.143*** [0.095] 1.306*** [0.101] 1.603*** [0.180] 2.233*** [0.093] 1.458*** [0.100] 1.685*** [0.102] 2.291*** [0.163] 1.359*** [0.150] -0.605*** [0.025] 0.131*** [0.035] -0.729*** [0.041] -0.995*** [0.040] -0.132*** [0.039] 5.494*** [0.177]

2002 0.103*** [0.006] -0.001*** [0.000] 0.281*** [0.067] 0.737*** [0.068] 1.007*** [0.070] 1.179*** [0.072] 1.429*** [0.132] 2.088*** [0.069] 1.320*** [0.086] 1.651*** [0.073] 2.187*** [0.135] 1.941*** [0.299] -0.491*** [0.018] -0.053** [0.026] -0.694*** [0.030] -0.925*** [0.031] -0.260*** [0.031] 5.492*** [0.139]

1994 0.087*** [0.008] -0.001*** [0.000] 0.262*** [0.100] 0.601*** [0.109] 0.506*** [0.117] 0.480*** [0.134] 0.303 [0.298] 0.784*** [0.157] 0.404*** [0.145] 0.468*** [0.153] 0.729** [0.334] -0.75 [0.960]

1996 0.076*** [0.007] -0.001*** [0.000] 0.288*** [0.094] 0.605*** [0.103] 0.581*** [0.110] 0.603*** [0.126] 0.553* [0.310] 0.991*** [0.149] 0.699*** [0.134] 0.609*** [0.145] 1.028*** [0.289]

2000 0.081*** [0.008] -0.001*** [0.000] 0.068 [0.118] 0.615*** [0.122] 0.673*** [0.128] 0.741*** [0.136] 0.587*** [0.218] 0.932*** [0.153] 0.710*** [0.147] 0.740*** [0.151] 0.843*** [0.310]

--

1998 0.062*** [0.008] -0.001*** [0.000] 0.426*** [0.121] 0.859*** [0.132] 0.881*** [0.139] 0.782*** [0.150] 0.495 [0.357] 1.000*** [0.177] 0.733*** [0.167] 0.673*** [0.177] 0.838** [0.330] 1.05 [0.988]

--

2002 0.090*** [0.007] -0.001*** [0.000] 0.201** [0.100] 0.590*** [0.105] 0.652*** [0.110] 0.544*** [0.117] 0.542** [0.236] 0.882*** [0.133] 0.603*** [0.131] 0.686*** [0.131] 1.110*** [0.340] 0.282 [0.783]

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

--

14157 0.21

--

--

5.660*** 5.943*** 5.795*** 5.550*** 5.549*** [0.187] [0.183] [0.214] [0.197] [0.170] 14036 0.30

12901 0.23

13163 0.29

18248 0.27

Specification tests:10 Hausman test –stat.,

10

See Appendix I for details on the specification tests.

44

287.45*** 211.49*** 199.97*** 179.57*** F(12291, F(12216, F(11232, F(11335, 1853) = 1808) = 1656) = 1816) = 2.16*** 2.08*** 1.81*** 2.16***

263.20*** F(15886, 2349) = 1.80***

Notes: Numbers in brackets are standard deviations. The OLS-specification applies sample weights. *: statistically significant at 10 percent; **: statistically significant at 5 percent; ***: statistically significant at 1 percent.

45

Appendix E. Results Across Municipal and Non-Municipal Sub-samples Table E1. Results for Non-Municipal Sub-sample 1994 Female -0.430*** [0.024] Age 0.086*** [0.007] Age squared -0.001*** [0.000] Primary, low 0.337*** [0.057] Primary, upper 0.809*** [0.064] Secondary, low 1.338*** [0.070] Secondary, upper 1.696*** [0.081] University, attended 1.955*** [0.169] University, degree 2.781*** [0.061] Vocational 2.023*** [0.085] Technical 2.416*** [0.076] Teacher 2.635*** [0.089] Other education -0.865 [0.611] Central -0.079*** [0.030] North 0.591*** [0.028] North-east 0.543*** [0.034] South --Constant 4.748*** [0.138] Observations 15913 R-squared 0.35 Specification tests:11 Hausman test –stat., χ2(12) for RE vs. FE: -F-test,for all ui = 0:

--

1996 -0.456*** [0.021] 0.075*** [0.005] -0.001*** [0.000] 0.460*** [0.055] 0.814*** [0.061] 1.293*** [0.065] 1.505*** [0.077] 2.401*** [0.135] 2.685*** [0.061] 1.867*** [0.079] 2.129*** [0.084] 2.644*** [0.106] -0.146 [0.605] 0.296*** [0.029] -0.384*** [0.031] -0.648*** [0.032] --5.471*** [0.121] 16041 0.38

OLS: 1998 -0.424*** [0.022] 0.083*** [0.006] -0.001*** [0.000] 0.254*** [0.054] 0.645*** [0.061] 1.185*** [0.070] 1.368*** [0.071] 1.702*** [0.141] 2.539*** [0.059] 1.667*** [0.089] 1.920*** [0.075] 2.418*** [0.138] 0.617** [0.288] 0.707*** [0.028] ---0.387*** [0.030] 0.501*** [0.033] 4.984*** [0.138] 15261 0.41

2000 -0.347*** [0.025] 0.076*** [0.007] -0.001*** [0.000] 0.255*** [0.064] 0.653*** [0.069] 1.181*** [0.074] 1.345*** [0.083] 2.376*** [0.162] 2.563*** [0.068] 1.625*** [0.089] 1.875*** [0.089] 2.464*** [0.161] 1.537*** [0.197] 0.939*** [0.028] ---0.322*** [0.034] 0.614*** [0.033] 4.808*** [0.154] 14453 0.43

2002 -0.376*** [0.024] 0.093*** [0.007] -0.001*** [0.000] 0.227*** [0.054] 0.629*** [0.059] 0.991*** [0.065] 1.209*** [0.073] 1.698*** [0.166] 2.334*** [0.062] 1.409*** [0.100] 1.841*** [0.073] 2.285*** [0.169] 1.787*** [0.113] 0.748*** [0.027] ---0.318*** [0.032] 0.469*** [0.029] 4.747*** [0.145] 12556 0.39

--

--

--

--

--

--

--

--

1994 -0.446*** [0.016] 0.062*** [0.006] -0.001*** [0.000] 0.111** [0.049] 0.318*** [0.057] 0.415*** [0.066] 0.364*** [0.079] 0.366 [0.257] 0.962*** [0.094] 0.530*** [0.091] 0.603*** [0.098] 0.943*** [0.157] -0.726 [0.982] --------6.093*** [0.121] 15915 0.25

Fixed effects specification: 1996 1998 2000 -0.443*** -0.392*** -0.312*** [0.015] [0.015] [0.015] 0.045*** 0.063*** 0.064*** [0.005] [0.006] [0.006] -0.001*** -0.001*** -0.001*** [0.000] [0.000] [0.000] 0.089* 0.041 0.109** [0.046] [0.048] [0.052] 0.262*** 0.291*** 0.371*** [0.052] [0.055] [0.058] 0.300*** 0.443*** 0.466*** [0.060] [0.061] [0.064] 0.317*** 0.475*** 0.529*** [0.072] [0.070] [0.070] 0.699*** 0.620*** 0.483*** [0.212] [0.194] [0.156] 0.830*** 0.951*** 0.936*** [0.085] [0.083] [0.081] 0.458*** 0.455*** 0.503*** [0.083] [0.087] [0.086] 0.542*** 0.645*** 0.669*** [0.091] [0.090] [0.083] 0.812*** 0.960*** 0.929*** [0.139] [0.145] [0.156] 0.250 0.460 0.682 [0.929] [0.949] [0.918] ------------------------6.622*** 6.137*** 5.947*** [0.116] [0.122] [0.125] 16041 15261 14453 0.24 0.27 0.26

659.65*** 688.89*** 564.30*** 688.56*** 386.38*** F(10079, F(10163, F(9622, F(8995, F(7841, 5822) = 5864) = 5625) = 5444) = 4701) = 2.96*** 3.18*** 2.96*** 2.97*** 3.43***

Notes: Numbers in brackets are standard deviations. The OLS-specification applies sample weights. *: statistically significant at 10 percent; **: statistically significant at 5 percent; ***: statistically significant at 1 percent.

11

See Appendix I for details on the specification tests.

46

2002 -0.325*** [0.016] 0.062*** [0.006] -0.001*** [0.000] 0.078 [0.054] 0.288*** [0.060] 0.364*** [0.066] 0.396*** [0.073] 0.365 [0.227] 0.890*** [0.090] 0.445*** [0.089] 0.544*** [0.089] 0.954*** [0.239] 1.581** [0.798] --------6.095*** [0.133] 12556 0.22

Table E2. Results for Municipal Sub-sample 1994 Female -0.255*** [0.021] Age 0.079*** [0.007] Age squared -0.001*** [0.000] Primary, low 0.487*** [0.117] Primary, upper 0.855*** [0.118] Secondary, low 0.995*** [0.118] Secondary, upper 1.168*** [0.122] University, attended 1.061*** [0.157] University, degree 1.792*** [0.115] Vocational 1.314*** [0.120] Technical 1.481*** [0.120] Teacher 1.403*** [0.138] Other education 0.242 [0.534] Central 0.141*** [0.036] North 0.174*** [0.036] North-east 0.106*** [0.037] South 0.494*** [0.032] Constant 5.515*** [0.178] Observations 8266 R-squared 0.44 Specification tests:12 Hausman test – stat., χ2(12) for RE vs. FE: -F-test,for all ui = 0:

--

1996 -0.224*** [0.027] 0.077*** [0.009] -0.001*** [0.000] 0.337*** [0.128] 0.731*** [0.129] 0.854*** [0.129] 1.012*** [0.131]

OLS: 1998 -0.229*** [0.022] 0.082*** [0.007] -0.001*** [0.000] 0.294*** [0.097] 0.677*** [0.096] 0.802*** [0.097] 0.986*** [0.098]

2000 -0.185*** [0.021] 0.085*** [0.008] -0.001*** [0.000] 0.304** [0.120] 0.723*** [0.122] 0.875*** [0.123] 1.063*** [0.123]

2002 -0.182*** [0.015] 0.094*** [0.005] -0.001*** [0.000] 0.055 [0.072] 0.575*** [0.072] 0.847*** [0.073] 0.981*** [0.073]

1994 -0.282*** [0.019] 0.079*** [0.007] -0.001*** [0.000] 0.376*** [0.093] 0.649*** [0.097] 0.817*** [0.100] 0.943*** [0.103]

Fixed effects specification: 1996 1998 2000 -0.248*** -0.225*** -0.190*** [0.018] [0.018] [0.017] 0.097*** 0.080*** 0.065*** [0.007] [0.008] [0.007] -0.001*** -0.001*** -0.000*** [0.000] [0.000] [0.000] 0.282*** 0.158 0.189* [0.094] [0.110] [0.101] 0.651*** 0.530*** 0.583*** [0.097] [0.114] [0.104] 0.738*** 0.636*** 0.720*** [0.099] [0.114] [0.104] 0.871*** 0.749*** 0.786*** [0.105] [0.118] [0.107]

2002 -0.247*** [0.012] 0.085*** [0.005] -0.001*** [0.000] 0.164*** [0.056] 0.547*** [0.058] 0.687*** [0.059] 0.725*** [0.062]

1.252*** [0.160] 1.680*** [0.129] 1.157*** [0.131] 1.410*** [0.130] 1.097*** [0.156] 0.814*** [0.155] -0.071** [0.031] -0.398*** [0.032] -0.428*** [0.028] -0.333*** [0.031] 6.213*** [0.216] 7974 0.46

1.158*** [0.122] 1.566*** [0.095] 0.960*** [0.101] 1.186*** [0.098] 1.113*** [0.108] 1.138*** [0.094] -0.179*** [0.026] -0.461*** [0.029] -0.414*** [0.025] -0.337*** [0.028] 6.103*** [0.169] 7559 0.46

0.916*** [0.169] 1.656*** [0.120] 1.073*** [0.124] 1.264*** [0.122] 1.195*** [0.151] 1.565*** [0.124] -0.117*** [0.025] -0.510*** [0.026] -0.477*** [0.026] -0.394*** [0.028] 5.894*** [0.201] 9530 0.45

0.988*** [0.123] 1.696*** [0.071] 1.080*** [0.077] 1.290*** [0.074] 1.536*** [0.104] 1.060*** [0.132] -0.160*** [0.018] -0.746*** [0.022] -0.746*** [0.021] -0.435*** [0.022] 5.815*** [0.125] 20764 0.47

0.769*** [0.161] 1.293*** [0.104] 1.004*** [0.105] 1.063*** [0.107] 1.262*** [0.136] 0.547 [0.672] --------5.952*** [0.167] 8266 0.39

0.970*** [0.150] 1.238*** [0.101] 0.969*** [0.105] 1.127*** [0.105] 0.985*** [0.139] ----------5.787*** [0.166] 7974 0.42

0.740*** [0.103] 1.119*** [0.063] 0.736*** [0.067] 0.902*** [0.066] 1.052*** [0.121] 0.379 [0.469] --------5.777*** [0.109] 20764 0.35

--

--

--

--

--

--

--

--

0.507*** [0.164] 1.061*** [0.117] 0.741*** [0.119] 0.790*** [0.119] 0.897*** [0.157] ----------6.099*** [0.181] 7559 0.40

0.691*** [0.133] 1.204*** [0.105] 0.830*** [0.109] 0.924*** [0.108] 0.962*** [0.150] 0.798 [0.815] --------6.205*** [0.162] 9530 0.39

114.73*** 101.81*** 149.84*** 160.56*** 641.39*** F(5398, F(5290, F(4855, F(6005, F(13229, 2691) = 3511) = 7521) = 2854) = 2671) = 1.86*** 1.79*** 1.78*** 1.88*** 2.29***

Notes: Numbers in brackets are standard deviations. The OLS-specification applies sample weights. *: statistically significant at 10 percent; **: statistically significant at 5 percent; ***: statistically significant at 1 percent.

12

See Appendix I for details on the specification tests.

47

Appendix F. Earnings Equations Results Across Regional Sub-samples Table F1. Results for Bangkok Region Sub-sample 1994 Female -0.523*** [0.039] Age 0.079*** [0.011] Age squared -0.001*** [0.000] Primary, low 0.426*** [0.086] Primary, upper 0.821*** [0.095] Secondary, low 1.277*** [0.114] Secondary, upper 1.588*** [0.144] University, attended 1.173*** [0.259] University, degree 2.842*** [0.094] Vocational 2.034*** [0.132] Technical 2.509*** [0.127] Teacher 2.649*** [0.132] Other education -1.064*** [0.250] Non-municipal -0.690*** [0.043] Constant 5.546*** [0.230] Observations 5620 R-squared 0.37 Specification tests:13 Hausman test –stat., χ2(12) for RE vs. FE: -F-test,for all ui = 0:

--

1996 -0.226*** [0.040] 0.072*** [0.013] -0.001*** [0.000] 0.215 [0.190] 0.567*** [0.190] 0.663*** [0.191] 0.813*** [0.195] 1.087*** [0.236] 1.447*** [0.190] 0.997*** [0.193] 1.249*** [0.194] 0.626** [0.243] 0.618*** [0.237] --6.486*** [0.320] 2035 0.43

OLS: 1998 -0.261*** [0.031] 0.081*** [0.010] -0.001*** [0.000] 0.306* [0.157] 0.621*** [0.155] 0.681*** [0.156] 0.910*** [0.158] 1.056*** [0.184] 1.483*** [0.155] 0.855*** [0.161] 1.049*** [0.161] 0.985*** [0.175] 1.079*** [0.153] --6.236*** [0.248] 2013 0.44

2000 -0.197*** [0.031] 0.090*** [0.012] -0.001*** [0.000] 0.456*** [0.138] 0.839*** [0.140] 0.914*** [0.141] 1.143*** [0.142] 0.926*** [0.202] 1.761*** [0.138] 1.113*** [0.146] 1.351*** [0.140] 1.107*** [0.184] ----5.754*** [0.266] 2194 0.45

2002 -0.164*** [0.027] 0.084*** [0.009] -0.001*** [0.000] -0.057 [0.157] 0.337** [0.153] 0.523*** [0.154] 0.706*** [0.154] 0.564*** [0.214] 1.334*** [0.153] 0.732*** [0.161] 0.853*** [0.158] 1.192*** [0.158] 0.659*** [0.150] --6.230*** [0.236] 2578 0.45

--

--

--

--

--

--

--

--

1994 -0.551*** [0.027] 0.075*** [0.010] -0.001*** [0.000] 0.188** [0.078] 0.495*** [0.094] 0.691*** [0.111] 0.598*** [0.130] 0.61 [0.448] 1.257*** [0.141] 0.880*** [0.149] 0.794*** [0.161] 1.214*** [0.248] -0.624 [1.035] --5.709*** [0.220] 5620 0.37

Fixed effects specification: 1996 1998 2000 -0.219*** -0.297*** -0.218*** [0.033] [0.033] [0.030] 0.075*** 0.071*** 0.066*** [0.013] [0.013] [0.012] -0.001*** -0.001*** -0.000*** [0.000] [0.000] [0.000] 0.378** 0.403 0.153 [0.181] [0.277] [0.227] 0.593*** 0.713*** 0.495** [0.184] [0.275] [0.231] 0.733*** 0.772*** 0.647*** [0.186] [0.277] [0.230] 0.906*** 0.832*** 0.778*** [0.200] [0.282] [0.237] 1.030*** 0.778** 0.4 [0.251] [0.324] [0.261] 1.178*** 1.172*** 1.332*** [0.189] [0.282] [0.232] 0.967*** 0.777*** 0.769*** [0.193] [0.284] [0.234] 1.053*** 0.821*** 0.977*** [0.199] [0.282] [0.237] 0.651* 0.186 1.395*** [0.390] [0.480] [0.452] ------------6.422*** 6.423*** 6.526*** [0.295] [0.344] [0.323] 2035 2013 2194 0.41 0.38 0.44

138.25*** 34.34*** F(3565, F(1295, 2041) = 727) = 2.42*** 1.64***

38.67*** F(1198, 802) = 1.68***

14.52 F(1275, 906) = 1.93***

2002 -0.210*** [0.029] 0.093*** [0.012] -0.001*** [0.000] -0.017 [0.143] 0.285** [0.140] 0.553*** [0.141] 0.600*** [0.147] 0.585*** [0.197] 1.017*** [0.140] 0.608*** [0.148] 0.741*** [0.150] 0.742 [0.477] 0.321 [0.757] --6.183*** [0.249] 2578 0.43

31.24*** F(1484, 1080) = 1.91***

Notes: Numbers in brackets are standard deviations. The OLS-specification applies sample weights. *: statistically significant at 10 percent; **: statistically significant at 5 percent; ***: statistically significant at 1 percent.

13

See Appendix I for details on the specification tests.

48

Table F2. Results for Central Region Sub-sample 1994 Female -0.445*** [0.042] Age 0.106*** [0.012] Age squared -0.001*** [0.000] Primary, low 0.142 [0.122] Primary, upper 0.848*** [0.134] Secondary, low 1.412*** [0.141] Secondary, upper 1.834*** [0.152] University, attended 1.981*** [0.219] University, degree 2.706*** [0.124] Vocational 2.152*** [0.146] Technical 2.441*** [0.137] Teacher 2.586*** [0.150] Other education Non-municipal Constant Observations R-squared Specification tests:14 Hausman test –stat., χ2(12) for RE vs. FE: F-test,for all ui = 0:

1996 -0.230*** [0.027] 0.089*** [0.008] -0.001*** [0.000] 0.419*** [0.091] 0.823*** [0.094] 1.210*** [0.094] 1.480*** [0.097] 1.704*** [0.147] 2.180*** [0.095] 1.628*** [0.102] 1.853*** [0.099] 1.789*** [0.149] 0.815*** [0.216] -0.761*** -0.433*** [0.039] [0.028] 5.123*** 5.876*** [0.254] [0.172] 7453 6736 0.37 0.38

OLS: 1998 -0.246*** [0.030] 0.096*** [0.009] -0.001*** [0.000] 0.419*** [0.139] 0.924*** [0.141] 1.331*** [0.141] 1.488*** [0.142] 1.776*** [0.157] 2.091*** [0.139] 1.579*** [0.156] 1.772*** [0.144] 1.865*** [0.194] 2.183*** [0.141] -0.317*** [0.028] 5.474*** [0.218] 5991 0.34

2000 -0.181*** [0.028] 0.085*** [0.008] -0.001*** [0.000] 0.361*** [0.102] 0.849*** [0.106] 1.258*** [0.107] 1.454*** [0.110] 1.478*** [0.195] 2.032*** [0.105] 1.594*** [0.112] 1.614*** [0.115] 1.609*** [0.231] 1.782*** [0.108] -0.337*** [0.027] 5.728*** [0.186] 6690 0.36

2002 -0.152*** [0.023] 0.109*** [0.008] -0.001*** [0.000] 0.518*** [0.094] 0.952*** [0.093] 1.290*** [0.095] 1.381*** [0.096] 1.264*** [0.172] 2.099*** [0.093] 1.599*** [0.102] 1.782*** [0.096] 1.521*** [0.198] 0.504*** [0.094] -0.349*** [0.020] 5.152*** [0.172] 10830 0.37

--

--

--

--

--

--

--

--

--

--

1994 -0.427*** [0.024] 0.054*** [0.009] -0.001*** [0.000] 0.295*** [0.109] 0.572*** [0.116] 0.616*** [0.125] 0.529*** [0.135] 0.722** [0.283] 1.050*** [0.141] 0.602*** [0.146] 0.665*** [0.148] 1.110*** [0.195] ----5.994*** [0.201] 7453 0.29

Fixed effects specification: 1996 1998 2000 -0.279*** -0.243*** -0.194*** [0.018] [0.020] [0.019] 0.051*** 0.055*** 0.060*** [0.007] [0.008] [0.007] -0.001*** -0.001*** -0.001*** [0.000] [0.000] [0.000] 0.029 0.037 0.160** [0.066] [0.077] [0.080] 0.216*** 0.299*** 0.394*** [0.073] [0.085] [0.086] 0.285*** 0.400*** 0.559*** [0.077] [0.089] [0.090] 0.357*** 0.541*** 0.579*** [0.089] [0.098] [0.095] 0.375* 0.651*** 0.441** [0.208] [0.213] [0.179] 0.815*** 0.705*** 0.892*** [0.094] [0.109] [0.101] 0.486*** 0.494*** 0.701*** [0.091] [0.105] [0.101] 0.595*** 0.638*** 0.625*** [0.099] [0.112] [0.101] 0.725*** 0.863*** 0.982*** [0.176] [0.229] [0.209] --0.606 --[0.793] ------6.999*** 6.754*** 6.569*** [0.147] [0.171] [0.166] 6736 5991 6690 0.32 0.31 0.32

454.41*** 263.09*** 215.69*** 206.96*** 229.70*** F(4729, F(4143, F(3740, F(4061, F(6577, 2711) = 2580) = 2238) = 2615) = 4240) = 3.42*** 3.18*** 2.83*** 2.50*** 2.33***

Notes: Numbers in brackets are standard deviations. The OLS-specification applies sample weights. *: statistically significant at 10 percent; **: statistically significant at 5 percent; ***: statistically significant at 1 percent.

14

See Appendix I for details on the specification tests.

49

2002 -0.190*** [0.016] 0.072*** [0.006] -0.001*** [0.000] 0.317*** [0.068] 0.607*** [0.072] 0.734*** [0.075] 0.774*** [0.079] 0.805*** [0.166] 1.096*** [0.084] 0.831*** [0.085] 0.851*** [0.086] 0.644*** [0.224] ----6.104*** [0.134] 10830 0.30

Table F3. Results for Northern Region Sub-sample 1994 Female -0.278*** [0.033] Age 0.080*** [0.009] Age squared -0.001*** [0.000] Primary, low 0.494*** [0.110] Primary, upper 0.801*** [0.113] Secondary, low 1.369*** [0.118] Secondary, upper 1.552*** [0.124] University, attended 1.824*** [0.253] University, degree 2.290*** [0.111] Vocational 1.776*** [0.130] Technical 2.073*** [0.119] Teacher 2.244*** [0.138] Other education 3.016*** [0.131] Non-municipal -0.411*** [0.030] Constant 5.762*** [0.206] Observations 5441 R-squared 0.33 Specification tests:15 Hausman test –stat., χ2(12) for RE vs. -FE: F-test,for all ui = 0:

--

1996 -0.464*** [0.037] 0.091*** [0.010] -0.001*** [0.000] 0.630*** [0.077] 1.030*** [0.090] 1.612*** [0.105] 1.960*** [0.116] 2.055*** [0.306] 2.889*** [0.085] 2.099*** [0.128] 2.399*** [0.109] 2.803*** [0.169] -0.515 [0.479] -0.639*** [0.038] 5.199*** [0.207] 5415 0.41

OLS: 1998 -0.455*** [0.039] 0.093*** [0.012] -0.001*** [0.000] 0.258*** [0.078] 0.567*** [0.097] 1.228*** [0.126] 1.457*** [0.122] 1.687*** [0.350] 2.587*** [0.087] 1.598*** [0.163] 2.188*** [0.106] 2.501*** [0.185] 0.369*** [0.086] -0.546*** [0.042] 5.360*** [0.272] 5142 0.41

2000 -0.355*** [0.041] 0.094*** [0.012] -0.001*** [0.000] 0.214*** [0.080] 0.692*** [0.095] 1.226*** [0.112] 1.400*** [0.135] 2.299*** [0.125] 2.701*** [0.085] 1.716*** [0.149] 2.211*** [0.118] 2.605*** [0.248] ---0.676*** [0.037] 5.080*** [0.259] 5391 0.38

2002 -0.363*** [0.032] 0.087*** [0.010] -0.001*** [0.000] 0.170*** [0.062] 0.538*** [0.071] 0.976*** [0.080] 1.168*** [0.091] 2.010*** [0.198] 2.374*** [0.072] 1.533*** [0.106] 1.999*** [0.089] 2.306*** [0.175] 1.727*** [0.218] -0.465*** [0.027] 5.211*** [0.206] 6941 0.41

--

--

--

--

--

--

--

--

1994 -0.266*** [0.022] 0.063*** [0.008] -0.001*** [0.000] 0.023 [0.074] 0.226*** [0.084] 0.351*** [0.092] 0.444*** [0.107] 0.234 [0.481] 0.904*** [0.120] 0.478*** [0.114] 0.688*** [0.126] 1.060*** [0.189] ----6.462*** [0.175] 5442 0.32

Fixed effects specification: 1996 1998 2000 -0.491*** -0.430*** -0.341*** [0.025] [0.025] [0.024] 0.076*** 0.072*** 0.081*** [0.009] [0.010] [0.009] -0.001*** -0.001*** -0.001*** [0.000] [0.000] [0.000] 0.227*** 0.052 0.140* [0.071] [0.074] [0.075] 0.463*** 0.389*** 0.567*** [0.084] [0.087] [0.085] 0.619*** 0.596*** 0.664*** [0.101] [0.097] [0.096] 0.689*** 0.648*** 0.644*** [0.115] [0.113] [0.105] 0.602* 0.844*** 0.897*** [0.357] [0.299] [0.202] 1.120*** 1.059*** 1.150*** [0.124] [0.120] [0.112] 0.739*** 0.641*** 0.625*** [0.148] [0.143] [0.129] 0.841*** 0.585*** 0.947*** [0.133] [0.140] [0.120] 1.130*** 0.772*** 1.233*** [0.202] [0.215] [0.216] 0.328 0.564 -[0.902] [0.938] -------5.907*** 5.929*** 5.404*** [0.204] [0.221] [0.203] 5415 5142 5391 0.38 0.40 0.40

158.42*** 166.45*** 195.11*** 239.52*** 201.12*** F(3418, F(3560, F(3241, F(3382, F(4525, 2011) = 1841) = 1887) = 1996) = 2402) = 3.12*** 2.61*** 2.68*** 3.13*** 2.63***

Notes: Numbers in brackets are standard deviations. The OLS-specification applies sample weights. *: statistically significant at 10 percent; **: statistically significant at 5 percent; ***: statistically significant at 1 percent.

15

See Appendix I for details on the specification tests.

50

2002 -0.342*** [0.022] 0.091*** [0.009] -0.001*** [0.000] 0.106 [0.067] 0.426*** [0.076] 0.603*** [0.086] 0.637*** [0.098] 0.784*** [0.239] 1.179*** [0.107] 0.617*** [0.114] 0.915*** [0.110] 1.499*** [0.261] 1.318** [0.602] --5.362*** [0.188] 6941 0.42

Table F4. Results for North-eastern Region Sub-sample 1994 Female -0.288*** [0.051] Age 0.103*** [0.015] Age squared -0.001*** [0.000] Primary, low 0.430*** [0.121] Primary, upper 0.738*** [0.129] Secondary, low 1.131*** [0.132] Secondary, upper 1.386*** [0.137] University, attended 1.572*** [0.225] University, degree 2.166*** [0.121] Vocational 1.433*** [0.162] Technical 1.686*** [0.134] Teacher 1.759*** [0.177] Other education --Non-municipal -0.362*** [0.037] Constant 5.281*** [0.288] Observations 3292 R-squared 0.38 Specification tests:16 Hausman test –stat., χ2(12) for RE vs. FE: -F-test,for all ui = 0:

--

1996 -0.601*** [0.040] 0.057*** [0.011] -0.001*** [0.000] 0.124 [0.174] 0.457** [0.185] 1.045*** [0.191] 1.087*** [0.212] 2.075*** [0.249] 2.575*** [0.177] 1.762*** [0.219] 2.131*** [0.202] 2.693*** [0.207] ---0.781*** [0.037] 6.314*** [0.285] 6386 0.38

OLS: 1998 -0.522*** [0.041] 0.076*** [0.012] -0.001*** [0.000] 0.173 [0.129] 0.629*** [0.139] 1.298*** [0.151] 1.510*** [0.163] 1.794*** [0.239] 2.804*** [0.131] 2.076*** [0.166] 2.080*** [0.173] 2.477*** [0.230] ---0.825*** [0.037] 5.573*** [0.263] 6377 0.39

2000 -0.431*** [0.049] 0.065*** [0.014] -0.001*** [0.000] 0.025 [0.204] 0.372* [0.210] 1.039*** [0.217] 1.218*** [0.237] 2.727*** [0.252] 2.684*** [0.205] 1.760*** [0.253] 1.896*** [0.259] 2.600*** [0.242] ---0.943*** [0.044] 5.887*** [0.344] 6201 0.35

2002 -0.495*** [0.043] 0.099*** [0.013] -0.001*** [0.000] 0.005 [0.165] 0.491*** [0.173] 0.869*** [0.180] 1.223*** [0.196] 1.736*** [0.211] 2.470*** [0.167] 1.319*** [0.244] 1.892*** [0.184] 2.537*** [0.182] 1.670*** [0.168] -0.660*** [0.033] 5.171*** [0.301] 7958 0.39

--

--

--

--

--

--

--

--

1994 -0.355*** [0.035] 0.089*** [0.012] -0.001*** [0.000] 0.076 [0.104] 0.424*** [0.118] 0.507*** [0.135] 0.593*** [0.154] 0.607** [0.291] 0.886*** [0.156] 0.633*** [0.159] 0.721*** [0.157] 0.663*** [0.230] ----5.943*** [0.266] 3293 0.36

Fixed effects specification: 1996 1998 2000 -0.468*** -0.399*** -0.316*** [0.027] [0.026] [0.026] 0.049*** 0.071*** 0.071*** [0.011] [0.010] [0.010] -0.000*** -0.001*** -0.001*** [0.000] [0.000] [0.000] -0.032 0.101 -0.024 [0.115] [0.114] [0.125] 0.384*** 0.574*** 0.449*** [0.122] [0.122] [0.131] 0.451*** 0.797*** 0.521*** [0.135] [0.130] [0.138] 0.362** 0.703*** 0.637*** [0.151] [0.145] [0.146] 0.691*** 0.251 0.598** [0.255] [0.293] [0.248] 1.006*** 1.240*** 1.037*** [0.153] [0.145] [0.151] 0.617*** 0.676*** 0.547*** [0.169] [0.162] [0.166] 0.879*** 0.889*** 0.688*** [0.166] [0.153] [0.158] 0.816*** 1.366*** 0.698*** [0.220] [0.205] [0.240] ------------6.252*** 5.483*** 5.484*** [0.236] [0.225] [0.234] 6386 6377 6201 0.41 0.42 0.38

75.36*** 232.72*** 265.73*** 313.99*** 319.53*** F(2291, F(4169, F(4064, F(3936, F(5215, 989) = 2204) = 2300) = 2252) = 2730) = 2.09*** 2.54*** 2.81*** 3.30*** 2.53***

Notes: Numbers in brackets are standard deviations. The OLS-specification applies sample weights. *: statistically significant at 10 percent; **: statistically significant at 5 percent; ***: statistically significant at 1 percent.

16

See Appendix I for details on the specification tests.

51

2002 -0.363*** [0.023] 0.068*** [0.010] -0.001*** [0.000] -0.046 [0.124] 0.408*** [0.129] 0.469*** [0.134] 0.581*** [0.144] 0.544** [0.238] 1.053*** [0.146] 0.597*** [0.162] 0.781*** [0.156] 0.989*** [0.237] ----5.704*** [0.232] 7958 0.43

Table F5. Results for Southern Region Sub-sample 1994 Female -0.240*** [0.029] Age 0.056*** [0.010] Age squared -0.000*** [0.000] Primary, low 0.518** [0.226] Primary, upper 0.829*** [0.227] Secondary, low 0.901*** [0.228] Secondary, upper 1.086*** [0.232] University, attended 0.907*** [0.260] University, degree 1.747*** [0.224] Vocational 1.254*** [0.230] Technical 1.379*** [0.230] Teacher 1.132*** [0.282] Other education 0.875*** [0.239] Non-municipal --Constant 6.460*** [0.314] Observations 2373 R-squared 0.42 Specification tests:17 Hausman test –stat., χ2(12) for RE vs. FE: -F-test,for all ui = 0:

--

1996 -0.382*** [0.045] 0.091*** [0.012] -0.001*** [0.000] 0.672*** [0.086] 0.955*** [0.097] 1.151*** [0.105] 1.288*** [0.128] 1.926*** [0.207] 2.230*** [0.091] 1.688*** [0.141] 1.574*** [0.141] 2.122*** [0.134] ---0.564*** [0.040] 5.567*** [0.257] 3443 0.36

OLS: 1998 -0.229*** [0.042] 0.062*** [0.012] -0.001*** [0.000] 0.392*** [0.085] 0.643*** [0.094] 0.852*** [0.133] 1.053*** [0.129] 1.395*** [0.136] 1.897*** [0.087] 1.258*** [0.127] 1.428*** [0.107] 1.692*** [0.236] ---0.405*** [0.039] 6.180*** [0.255] 3297 0.34

2000 -0.283*** [0.042] 0.080*** [0.013] -0.001*** [0.000] 0.653*** [0.127] 0.996*** [0.132] 1.258*** [0.141] 1.301*** [0.145] 1.926*** [0.172] 2.249*** [0.127] 1.431*** [0.153] 1.793*** [0.149] 2.225*** [0.187] 1.718*** [0.221] -0.397*** [0.037] 5.488*** [0.270] 3507 0.4

2002 -0.273*** [0.034] 0.086*** [0.010] -0.001*** [0.000] 0.401*** [0.079] 0.799*** [0.083] 1.045*** [0.090] 1.176*** [0.096] 1.388*** [0.145] 1.992*** [0.083] 1.319*** [0.146] 1.619*** [0.102] 2.034*** [0.168] 0.655*** [0.081] -0.375*** [0.031] 5.527*** [0.197] 5013 0.38

--

--

--

--

--

--

--

--

1994 -0.244*** [0.032] 0.050*** [0.013] 0.000 [0.000] 0.782*** [0.190] 0.970*** [0.199] 1.102*** [0.200] 1.301*** [0.205] 0.831*** [0.261] 1.669*** [0.205] 1.410*** [0.206] 1.397*** [0.209] 1.596*** [0.323] 0.765 [0.654] --6.375*** [0.308] 2373 0.40

27.29*** F(1471, 888) = 1.63***

Fixed effects specification: 1996 1998 2000 -0.371*** -0.314*** -0.242*** [0.032] [0.031] [0.031] 0.087*** 0.085*** 0.053*** [0.012] [0.011] [0.012] -0.001*** -0.001*** -0.001*** [0.000] [0.000] [0.000] 0.351*** 0.07 0.342*** [0.098] [0.095] [0.125] 0.586*** 0.224** 0.625*** [0.110] [0.110] [0.133] 0.477*** 0.198* 0.645*** [0.120] [0.117] [0.140] 0.600*** 0.297** 0.742*** [0.148] [0.125] [0.151] 1.084** -0.212 0.642*** [0.530] [0.335] [0.237] 0.925*** 0.766*** 0.970*** [0.152] [0.141] [0.156] 0.700*** 0.510*** 0.704*** [0.156] [0.141] [0.166] 0.745*** 0.433*** 0.844*** [0.153] [0.141] [0.164] 0.957*** 0.591*** 0.959*** [0.213] [0.225] [0.217] --0.977 --[0.863] ------5.920*** 6.234*** 6.412*** [0.249] [0.231] [0.261] 3443 3297 3507 0.30 0.29 0.37

2002 -0.282*** [0.025] 0.073*** [0.009] -0.001*** [0.000] 0.095 [0.083] 0.312*** [0.090] 0.329*** [0.095] 0.290*** [0.100] 0.275 [0.211] 0.703*** [0.109] 0.224* [0.123] 0.483*** [0.112] 0.915*** [0.198] 0.113 [0.688] --6.378*** [0.195] 5013 0.31

94.17*** 101.57*** 111.41*** 175.48*** F(2283, F(2231, F(2343, F(3266, 1147) = 1053) = 1150) = 1733) = 2.08*** 2.08*** 1.93*** 2.14***

Notes: Numbers in brackets are standard deviations. The OLS-specification applies sample weights. *: statistically significant at 10 percent; **: statistically significant at 5 percent; ***: statistically significant at 1 percent.

17

See Appendix I for details on the specification tests.

52

Appendix G. Detailed Female-Male Wage Decompositions Results

Table G1. Detailed Female-Male Wage Decomposition Results, 1994 Explained: D= Variables From From Interaction Oaxaca/Blinder Reimers Cotton Neumark Endowment Coefficients (D=0) Age 0.146 0.867 0.063 0.209 0.178 0.183 0.182 Age squared -0.124 -0.329 -0.047 -0.172 -0.148 -0.152 -0.138 Primary, low 0.012 0 0 0.012 0.012 0.012 0.015 Primary, upper 0 -0.011 0 0 0 0 0 Secondary, low 0.039 -0.036 -0.018 0.021 0.03 0.028 0.035 Secondary, upper 0.014 -0.021 -0.007 0.007 0.01 0.01 0.013 University, attended 0 -0.003 0 0 0 0 0 University, degree -0.104 -0.119 0.053 -0.052 -0.078 -0.073 -0.08 Vocational 0.003 -0.023 -0.002 0.001 0.002 0.002 0.003 Technical -0.013 -0.027 0.007 -0.006 -0.009 -0.009 -0.012 Teacher -0.01 -0.015 0.006 -0.004 -0.007 -0.006 -0.006 Other education 0 0 0 0 0 0 0 Constant 0 0.019 0 0 0 0 0 Total -0.038 0.302 0.055 0.018 -0.01 -0.005 0.011 Coefficients, Means and Predictions: Variables Age Age squared Primary, low Primary, upper Secondary, low Secondary, upper University, attended University, degree Vocational Technical Teacher Other education Constant Total

Males: Females: Pooled model: Mean Pred. Coef. Mean Pred. Coef. 0.087 35.454 3.093 0.061 33.053 2.017 0.076 -0.0011 382.75 -1.366 -0.0011 209.006 -0.865 -0.001 0.262 0.403 0.106 0.261 0.359 0.094 0.33 0.601 0.212 0.128 0.651 0.212 0.138 0.705 0.506 0.122 0.062 0.949 0.081 0.077 0.849 0.48 0.061 0.029 0.941 0.046 0.043 0.882 0.303 0.005 0.001 0.893 0.005 0.005 0.737 0.784 0.084 0.065 1.582 0.149 0.236 1.221 0.404 0.044 0.018 0.958 0.042 0.04 0.948 0.468 0.033 0.015 1.071 0.045 0.048 0.996 0.729 0.008 0.006 1.854 0.013 0.025 1.137 -0.75 0 0 0.61 0 0 -0.226 5.66 1 5.66 5.641 1 5.641 5.468 7.817 7.498

Coef.

53

Table G2. Detailed Female-Male Wage Decomposition Results, 1996 Explained: D= Variables From From Interaction Oaxaca/Blinder Reimers Endowment Coefficients (D=0) Age 0.138 0.224 0.013 0.151 Age squared -0.116 -0.074 -0.009 -0.124 Primary, low 0.001 0.088 0.008 0.009 Primary, upper 0.011 -0.01 -0.001 0.01 Secondary, low 0.033 -0.008 -0.005 0.028 Secondary, upper 0.006 0.001 0 0.006 University, attended -0.003 -0.002 0.001 -0.002 University, degree -0.082 -0.04 0.017 -0.065 Vocational 0.005 0.001 0 0.005 Technical -0.017 -0.022 0.007 -0.01 Teacher -0.003 -0.003 0.001 -0.002 Other education 0 0 0 0 Constant 0 0.149 0 0 Total -0.028 0.302 0.033 0.005

Cotton

0.144 -0.12 0.005 0.01 0.03 0.006 -0.002 -0.074 0.005 -0.013 -0.003 0 0 -0.011

0.145 -0.121 0.006 0.01 0.03 0.006 -0.002 -0.072 0.005 -0.013 -0.002 0 0 -0.008

Neumark

0.139 -0.106 0.01 0.011 0.037 0.008 -0.003 -0.078 0.006 -0.016 -0.002 0 0 0.006

Coefficients, Means and Predictions: Variables Age Age squared Primary, low Primary, upper Secondary, low Secondary, upper University, attended University, degree Vocational Technical Teacher Other education Constant Total

Males: Coef. Mean Pred. 0.076 36.093 -0.0011 429.031 0.288 0.375 0.605 0.223 0.581 0.135 0.603 0.055 0.553 0.004 0.991 0.094 0.699 0.047 0.609 0.034 1.028 0.008 0 0 5.943 1

Females: Pooled model: Coef. Mean Pred. Coef. 2.741 0.069 34.108 2.366 0.07 -1.201 -0.0011 280.956 -1.002 -0.001 0.108 0.031 0.343 0.011 0.302 0.135 0.654 0.206 0.135 0.665 0.078 0.678 0.086 0.059 0.769 0.033 0.588 0.045 0.027 0.814 0.002 0.841 0.008 0.007 0.87 0.093 1.244 0.16 0.199 1.18 0.033 0.676 0.04 0.027 0.909 0.021 1.056 0.05 0.053 1.006 0.008 1.362 0.01 0.014 1.032 0 0 0 0 -0.085 5.943 5.794 1 5.794 5.756 7.995 7.688

54

Table G3. Detailed Female-Male Wage Decomposition Results, 1998 Explained: D= Variables From From Interaction Oaxaca/Blinder Reimers Cotton Neumark Endowment Coefficients (D=0) Age 0.152 -0.656 -0.035 0.116 0.134 0.132 0.137 Age squared -0.132 0.471 0.051 -0.081 -0.106 -0.103 -0.102 Primary, low -0.005 0.187 0.018 0.012 0.003 0.004 0.006 Primary, upper 0.01 0.091 0.011 0.02 0.015 0.016 0.015 Secondary, low 0.024 0.029 0.012 0.035 0.03 0.03 0.033 Secondary, upper 0.013 0.002 0.001 0.013 0.013 0.013 0.015 University, attended -0.001 0.002 -0.001 -0.002 -0.001 -0.001 -0.002 University, degree -0.077 -0.01 0.004 -0.073 -0.075 -0.075 -0.083 Vocational 0.002 0.012 0.002 0.004 0.003 0.003 0.005 Technical -0.008 0.005 -0.001 -0.009 -0.008 -0.008 -0.011 Teacher -0.002 -0.004 0.001 -0.002 -0.002 -0.002 -0.002 Other education 0 0 0 0 0 0 0 Constant 0 0.071 0 0 0 0 0 Total -0.025 0.2 0.061 0.036 0.005 0.009 0.009 Coefficients, Means and Predictions: Variables Age Age squared Primary, low Primary, upper Secondary, low Secondary, upper University, attended University, degree Vocational Technical Teacher Other education Constant Total

Males: Females: Pooled model: Coef. Mean Pred. Coef. Mean Pred. Coef. 0.062 36.345 2.264 0.081 34.479 2.803 0.073 -0.0011 442.258 -0.827 -0.0011 301.126 -1.217 -0.001 0.426 0.331 0.141 -0.193 0.303 -0.058 0.218 0.859 0.226 0.194 0.41 0.203 0.083 0.648 0.881 0.138 0.122 0.59 0.098 0.058 0.814 0.782 0.073 0.057 0.738 0.056 0.042 0.866 0.495 0.005 0.002 0.231 0.008 0.002 0.658 1 0.11 0.11 1.056 0.183 0.193 1.14 0.733 0.044 0.033 0.437 0.039 0.017 0.839 0.673 0.041 0.028 0.588 0.054 0.032 0.893 0.838 0.007 0.006 1.261 0.009 0.011 1.075 1.05 0 0 0 0 0 0.843 5.795 1 5.795 5.724 1 5.724 5.604 7.925 7.69

55

Table G4. Detailed Female-Male Wage Decomposition Results, 2000 Explained: D= Variables From From Interaction Oaxaca/Blinder Reimers Cotton Neumark Endowment Coefficients (D=0) Age 0.136 -0.147 -0.007 0.129 0.133 0.132 0.112 Age squared -0.113 0.229 0.021 -0.091 -0.102 -0.101 -0.081 Primary, low 0.006 -0.075 -0.005 0.001 0.004 0.003 0.005 Primary, upper 0.024 -0.058 -0.008 0.016 0.02 0.02 0.018 Secondary, low 0.043 -0.038 -0.015 0.028 0.036 0.035 0.034 Secondary, upper 0.016 -0.017 -0.004 0.012 0.014 0.014 0.014 University, attended -0.001 -0.002 0 -0.001 -0.001 -0.001 -0.001 University, degree -0.131 -0.14 0.055 -0.076 -0.103 -0.101 -0.1 Vocational 0.011 -0.017 -0.004 0.007 0.009 0.009 0.008 Technical -0.01 -0.034 0.004 -0.005 -0.008 -0.007 -0.007 Teacher -0.002 -0.006 0.001 -0.001 -0.002 -0.002 -0.001 Other education 0 0 0 0 0 0 0 Constant 0 0.456 0 0 0 0 0 Total -0.02 0.15 0.039 0.019 0 0.001 0.002 Coefficients, Means and Predictions: Variables Age Age squared Primary, low Primary, upper Secondary, low Secondary, upper University, attended University, degree Vocational Technical Teacher Other education Constant Total

Males: Females: Pooled model: Coef. Mean Pred. Coef. Mean Pred. Coef. 0.081 36.824 2.965 0.085 35.216 2.982 0.07 -0.0011 478.234 -1.077 -0.0011 352.783 -1.214 -0.001 0.068 0.291 0.02 0.344 0.273 0.094 0.257 0.615 0.221 0.136 0.914 0.194 0.178 0.687 0.673 0.145 0.098 1.044 0.104 0.108 0.823 0.741 0.084 0.063 0.995 0.069 0.068 0.883 0.587 0.009 0.005 0.75 0.01 0.007 0.752 0.932 0.124 0.116 1.614 0.205 0.331 1.226 0.71 0.047 0.033 1.173 0.037 0.044 0.903 0.74 0.052 0.039 1.312 0.06 0.078 0.975 0.843 0.007 0.006 1.588 0.008 0.013 1.102 0 0 0 0 0 0 1.17 5.55 1 5.55 5.094 1 5.094 5.579 7.952 7.783

56

Table G5. Detailed Female-Male Wage Decomposition Results, 2002 Explained: D= Variables From From Interaction Oaxaca/Blinder Reimers Cotton Neumark Endowment Coefficients (D=0) Age 0.114 0.35 0.014 0.127 0.12 0.121 0.112 Age squared -0.092 -0.178 -0.015 -0.106 -0.099 -0.1 -0.087 Primary, low 0.001 0.011 0 0.002 0.001 0.002 0.002 Primary, upper 0.022 -0.039 -0.005 0.016 0.019 0.019 0.017 Secondary, low 0.037 -0.034 -0.012 0.025 0.031 0.03 0.029 Secondary, upper 0.012 -0.036 -0.006 0.006 0.009 0.009 0.009 University, attended 0 -0.001 0 0 0 0 0 University, degree -0.112 -0.097 0.038 -0.074 -0.093 -0.091 -0.098 Vocational 0.014 -0.017 -0.006 0.008 0.011 0.01 0.011 Technical 0.008 -0.025 -0.003 0.005 0.007 0.007 0.007 Teacher -0.002 -0.002 0.001 -0.002 -0.002 -0.002 -0.002 Other education 0 0 0 0 0 0 0 Constant 0 0.246 0 0 0 0 0 Total 0.001 0.178 0.006 0.007 0.004 0.004 0 Coefficients, Means and Predictions: Variables Age Age squared Primary, low Primary, upper Secondary, low Secondary, upper University, attended University, degree Vocational Technical Teacher Other education Constant Total

Males: Coef. Mean Pred. Coef. 0.09 37.163 3.354 -0.0011 503.048 -1.394 0.201 0.275 0.055 0.59 0.225 0.133 0.652 0.148 0.096 0.544 0.083 0.045 0.542 0.006 0.003 0.882 0.126 0.112 0.603 0.048 0.029 0.686 0.061 0.042 1.11 0.004 0.005 0.282 0 0 5.549 1 5.549 8.028

57

Females: Pooled model: Mean Pred. Coef. 0.08 35.752 2.876 0.079 -0.0011 388.414 -1.109 -0.001 0.16 0.267 0.043 0.215 0.785 0.197 0.155 0.62 0.965 0.11 0.106 0.771 1.051 0.072 0.075 0.801 0.702 0.007 0.005 0.777 1.342 0.21 0.282 1.168 1.074 0.036 0.038 0.841 1.15 0.054 0.062 0.978 1.411 0.006 0.009 1.134 0 0 0 0.877 5.302 1 5.302 5.506 7.844

Appendix H. Detailed Municipal-Non-Municipal Wage Decompositions Results Table H1. Detailed Municipal-Non-Municipal Wage Decomposition Results, 1994 Explained: D= Variables From Endowment From Coefficients Interaction Oaxaca/Blinder Reimers Cotton Neumark (D=0) Age -0.016 0.379 -0.002 -0.019 -0.018 -0.017 -0.017 Age squared 0.024 0.254 -0.005 0.019 0.021 0.022 0.022 Primary, low -0.069 0.082 -0.043 -0.111 -0.09 -0.083 -0.08 Primary, upper -0.051 0.045 -0.016 -0.066 -0.059 -0.056 -0.06 Secondary, low 0.043 0.021 0.014 0.057 0.05 0.048 0.049 Secondary, upper 0.029 0.018 0.019 0.049 0.039 0.036 0.038 University, attended 0.005 0.001 0.002 0.008 0.006 0.006 0.006 University, degree 0.168 0.015 0.036 0.204 0.186 0.181 0.178 Vocational 0.039 0.01 0.018 0.058 0.049 0.046 0.046 Technical 0.042 0.007 0.015 0.057 0.049 0.047 0.046 Teacher 0.006 0.003 0.002 0.009 0.008 0.007 0.007 Other education 0 0 0.001 0 0 0 0 Constant 0 0.131 0 0 0 0 0 Total 0.222 0.967 0.042 0.264 0.243 0.236 0.235 Coefficients, Means and Predictions: Municipal: Variables Coef. Mean Age 0.084 Age squared -0.0011 Primary, low 0.46 Primary, upper 0.786 Secondary, low 0.988 Secondary, upper 1.137 University, attended 0.955 University, degree 1.406 Vocational 1.186 Technical 1.231 Teacher 1.347 Other education 0.683 Constant 5.523 Total

Non-Municipal: Pooled model: Coef. Mean Pred. Coef. 2.889 0.073 34.536 2.528 0.076 -0.872 -0.0011 320.24 -1.145 -0.001 0.104 0.284 0.467 0.133 0.33 0.123 0.599 0.241 0.144 0.705 0.141 0.746 0.085 0.063 0.849 0.094 0.68 0.04 0.027 0.882 0.01 0.668 0.002 0.002 0.737 0.29 1.157 0.061 0.071 1.221 0.089 0.81 0.026 0.021 0.948 0.084 0.909 0.022 0.02 0.996 0.02 0.978 0.008 0.008 1.137 0 -0.622 0 0 -0.226 5.523 5.392 1 5.392 5.468 8.495 7.264

Pred. 34.311 292.408 0.226 0.157 0.143 0.083 0.01 0.207 0.075 0.068 0.015 0.001 1

58

Table H2. Detailed Municipal-Non-Municipal Wage Decomposition Results, 1996 Explained: D= Variables From Endowment From Coefficients Interaction Oaxaca/Blinder Reimers Cotton Neumark (D=0) Age -0.033 1.712 -0.029 -0.062 -0.047 -0.042 -0.042 Age squared 0.034 -0.418 0.017 0.051 0.043 0.04 0.04 Primary, low -0.063 0.047 -0.026 -0.089 -0.076 -0.071 -0.075 Primary, upper -0.048 0.064 -0.024 -0.072 -0.06 -0.056 -0.061 Secondary, low 0.033 0.028 0.016 0.048 0.04 0.038 0.041 Secondary, upper 0.019 0.017 0.012 0.031 0.025 0.023 0.025 University, attended 0.009 0 0.002 0.011 0.01 0.01 0.009 University, degree 0.172 0.022 0.058 0.23 0.201 0.192 0.2 Vocational 0.037 0.01 0.016 0.053 0.045 0.042 0.043 Technical 0.044 0.011 0.027 0.071 0.057 0.053 0.057 Teacher 0.005 0.001 0.001 0.006 0.006 0.006 0.006 Other education 0 0 0 0 0 0 0 Constant 0 -0.57 0 0 0 0 0 Total 0.21 0.924 0.069 0.279 0.245 0.233 0.242 Coefficients, Means and Predictions: Municipal: Variables Coef. Mean Age 0.102 Age squared -0.0011 Primary, low 0.358 Primary, upper 0.776 Secondary, low 0.906 Secondary, upper 1.024 University, attended 1.095 University, degree 1.358 Vocational 1.106 Technical 1.253 Teacher 1.066 Other education 0 Constant 5.397 Total

Non-Municipal: Pooled model: Pred. Coef. Mean Pred. Coef. 34.865 3.57 0.054 35.469 1.919 0.07 330.205 -1.223 -0.0011 386.041 -0.856 -0.001 0.196 0.07 0.253 0.444 0.112 0.302 0.154 0.12 0.515 0.247 0.127 0.665 0.15 0.136 0.614 0.097 0.06 0.769 0.071 0.073 0.621 0.041 0.026 0.814 0.013 0.014 0.915 0.002 0.002 0.87 0.235 0.319 1.017 0.065 0.066 1.18 0.076 0.084 0.767 0.028 0.022 0.909 0.078 0.098 0.771 0.022 0.017 1.006 0.013 0.014 0.966 0.007 0.007 1.032 0 0 -0.148 0 0 -0.085 1 5.397 5.967 1 5.967 5.756 8.671 7.468

59

Table H3. Detailed Municipal-Non-Municipal Wage Decomposition Results, 1998 Explained: D= Variables From Endowment From Coefficients Interaction Oaxaca/Blinder Reimers Cotton Neumark (D=0) Age -0.013 0.601 -0.003 -0.016 -0.015 -0.014 -0.014 Age squared 0.021 0.144 -0.003 0.018 0.019 0.02 0.02 Primary, low -0.042 0.018 -0.01 -0.053 -0.048 -0.046 -0.05 Primary, upper -0.06 0.032 -0.014 -0.074 -0.067 -0.064 -0.074 Secondary, low 0.022 0.006 0.002 0.024 0.023 0.023 0.025 Secondary, upper 0.02 0.006 0.003 0.023 0.021 0.021 0.022 University, attended 0.006 -0.001 -0.002 0.004 0.005 0.005 0.004 University, degree 0.209 0.004 0.01 0.219 0.214 0.213 0.217 Vocational 0.039 0.003 0.007 0.045 0.042 0.041 0.044 Technical 0.055 0 0 0.055 0.055 0.055 0.054 Teacher 0.005 -0.001 -0.001 0.005 0.005 0.005 0.005 Other education 0 0 0 0 0 0 0 Constant 0 0.143 0 0 0 0 0 Total 0.263 0.954 -0.013 0.25 0.256 0.259 0.254 Coefficients, Means and Predictions: Municipal: Variables Coef. Mean Age 0.085 Age squared -0.0011 Primary, low 0.232 Primary, upper 0.65 Secondary, low 0.783 Secondary, upper 0.889 University, attended 0.606 University, degree 1.15 Vocational 0.867 Technical 0.901 Teacher 0.962 Other education 0 Constant 5.737 Total

Non-Municipal: Pooled model: Pred. Coef. Mean Pred. Coef. 35.405 3.004 0.068 35.598 2.42 0.073 362.805 -0.901 -0.0011 389.883 -1.062 -0.001 0.167 0.039 0.187 0.394 0.074 0.218 0.14 0.091 0.523 0.254 0.133 0.648 0.141 0.111 0.726 0.111 0.08 0.814 0.083 0.074 0.777 0.057 0.045 0.866 0.011 0.007 0.903 0.004 0.004 0.658 0.269 0.309 1.1 0.078 0.086 1.14 0.077 0.067 0.74 0.025 0.018 0.839 0.088 0.079 0.909 0.027 0.024 0.893 0.011 0.011 1.12 0.006 0.007 1.075 0 0 0.702 0 0 0.843 1 5.737 5.595 1 5.595 5.604 8.628 7.424

60

Table H4. Detailed Municipal-Non-Municipal Wage Decomposition Results, 2000 Explained: D= Variables From Endowment From Coefficients Interaction Oaxaca/Blinder Reimers Cotton Neumark (D=0) Age -0.035 -0.043 0.001 -0.035 -0.035 -0.035 -0.037 Age squared 0.033 0.437 -0.014 0.019 0.026 0.028 0.029 Primary, low -0.043 0.015 -0.008 -0.051 -0.047 -0.046 -0.052 Primary, upper -0.059 0.037 -0.016 -0.075 -0.067 -0.065 -0.075 Secondary, low 0.015 0.019 0.004 0.019 0.017 0.016 0.018 Secondary, upper 0.019 0.012 0.004 0.023 0.021 0.021 0.023 University, attended 0.007 0.001 0.001 0.008 0.008 0.007 0.007 University, degree 0.195 0.019 0.038 0.233 0.214 0.21 0.22 Vocational 0.029 0.006 0.009 0.038 0.033 0.032 0.035 Technical 0.049 0.007 0.012 0.061 0.055 0.053 0.057 Teacher 0.003 0 0 0.003 0.003 0.003 0.004 Other education 0 0 0 0 0 0 0 Constant 0 0.408 0 0 0 0 0 Total 0.212 0.917 0.031 0.243 0.227 0.224 0.23 Coefficients, Means and Predictions: Municipal: Variables Coef. Mean Age 0.066 Age squared -0.0001 Primary, low 0.252 Primary, upper 0.689 Secondary, low 0.845 Secondary, upper 0.91 University, attended 0.831 University, degree 1.295 Vocational 0.967 Technical 1.041 Teacher 1.038 Other education 1.095 Constant 5.937 Total

Non-Municipal: Pooled model: Pred. Coef. Mean Pred. Coef. 35.781 2.367 0.067 36.308 2.445 0.07 394.441 -0.599 -0.0011 439.568 -1.055 -0.001 0.162 0.041 0.212 0.363 0.077 0.257 0.143 0.099 0.542 0.252 0.137 0.687 0.14 0.118 0.68 0.118 0.08 0.823 0.093 0.084 0.738 0.067 0.05 0.883 0.015 0.013 0.703 0.005 0.004 0.752 0.269 0.348 1.084 0.089 0.097 1.226 0.066 0.064 0.73 0.027 0.02 0.903 0.091 0.095 0.834 0.033 0.027 0.975 0.009 0.009 1.052 0.006 0.006 1.102 0 0 1.144 0 0 1.17 1 5.937 5.529 1 5.529 5.579 8.575 7.415

61

Table H5. Detailed Municipal-Non-Municipal Wage Decomposition Results, 2002 Explained: D= Variables From Endowment From Coefficients Interaction Oaxaca/Blinder Reimers Cotton Neumark (D=0) Age -0.018 0.81 -0.006 -0.023 -0.021 -0.021 -0.021 Age squared 0.02 -0.104 0.002 0.022 0.021 0.021 0.021 Primary, low -0.024 0.034 -0.015 -0.038 -0.031 -0.033 -0.034 Primary, upper -0.036 0.066 -0.021 -0.057 -0.047 -0.049 -0.052 Secondary, low 0.006 0.034 0.003 0.009 0.008 0.008 0.009 Secondary, upper 0.014 0.019 0.007 0.022 0.018 0.019 0.02 University, attended 0.003 0.001 0.002 0.005 0.004 0.004 0.004 University, degree 0.158 0.015 0.035 0.193 0.176 0.18 0.181 Vocational 0.018 0.007 0.007 0.025 0.021 0.022 0.023 Technical 0.028 0.012 0.013 0.042 0.035 0.037 0.038 Teacher 0.004 0 0 0.004 0.004 0.004 0.004 Other education 0 0 0 0 0 0 0 Constant 0 -0.24 0 0 0 0 0 Total 0.173 0.654 0.028 0.202 0.188 0.191 0.192 Coefficients, Means and Predictions: Municipal: Variables Coef. Mean Age 0.087 Age squared -0.0011 Primary, low 0.243 Primary, upper 0.689 Secondary, low 0.846 Secondary, upper 0.892 University, attended 0.899 University, degree 1.242 Vocational 0.919 Technical 1.081 Teacher 1.138 Other education 0.366 Constant 5.433 Total

Non-Municipal: Pooled model: Pred. Coef. Mean Pred. Coef. 36.423 3.175 0.065 36.693 2.388 0.079 440.73 -1.126 -0.0011 468.499 -1.044 -0.001 0.212 0.051 0.15 0.369 0.055 0.215 0.181 0.125 0.439 0.264 0.116 0.62 0.135 0.114 0.573 0.123 0.071 0.771 0.087 0.078 0.589 0.063 0.037 0.801 0.008 0.008 0.476 0.003 0.001 0.777 0.223 0.277 1.018 0.067 0.069 1.168 0.053 0.049 0.663 0.026 0.017 0.841 0.072 0.078 0.738 0.034 0.025 0.978 0.006 0.007 1.149 0.003 0.004 1.134 0 0 2.082 0 0 0.877 1 5.433 5.673 1 5.673 5.506 8.267 7.412

62

Appendix I. Specification Tests (1) Hausman (1978) specification test This test is a general specification test for discriminating between two estimators, where one is efficient under H0 but inconsistent under HA and the other estimator is consistent both under H0 and HA but inefficient under H0. The way I set it up here amounts to testing between the random (RE) and fixed effects (FE) versions of the earnings regression with household specific effects. The test hypotheses are: H0: The difference in the estimated parameters from the two models is not statistically significantly different from zero, so that the efficient but possibly inconsistent model (that is, the random effects specification) should be chosen. HA: The difference in the estimated parameters from the two models is statistically significantly different from zero, so that the consistent model (that is, the fixed effects specification) should be chosen. The test results all support the fixed effects, except for the case of the year 2000 for Bangkok (Table F1, Appendix F). (2) LR-test for Pooling of Models/Exclusion of Variables (Chow-Type Test): This test is useful when one wants to compare – in order to choose between – two rivaling (and nested) specifications. The test builds on the log-likelihoods of the estimated models in question and the test statistic is: 2{abs[log-likelihoodfull model – loglikelihoodreduced model]} and is distributed χ2 with degrees of freedom given as the difference between the number of parameters estimated in the full model (i.e., under HA) and the number of parameters estimated in the reduced model (i.e., under H0). Hence, the hypotheses are: H0: The additional variables should not be included, i.e. the sample split is not permissible and so the pooled model is valid HA: The additional variables should be included, i.e. the sample split is permissible and the pooled model is not valid and the full model should be chosen instead. I use this test to determine whether the earnings equations should be estimated for all five survey years individually or whether it is permissible to pool all five survey years into one regression model (with common slopes, i.e. identical returns to education across all five survey years). In all cases I reject pooling, whereby it is permissible to estimate the earnings equations separately across the five survey years. Additionally, I test whether the samples may further be split across gender, nonmunicipal-municipal areas and regions. Again the test results support the sample splits. 63

Intuitively, rejection of H0 can be interpreted as evidence of structural differences in the wage determinants of the sub-samples, in other words, that the returns structure (related to education and potential general experience and gender (the latter only relevant for the splits across municipal-non-municipal location and region)) is not identical across subsamples. (3) F-test for joint significance of the fixed effects This test amounts to an additional examination of whether the fixed effects specification is relevant by testing whether ui, the fixed effect, is statistically different from zero. The test hypotheses are: H0: All ui =0. HA: All ui ≠0. The results from these tests all support rejection of H0 which, in turn indicating additional support for the fixed effects model.

64