w_    s 1L f+q
POLICY RESEARCH WORKING PAPER   1949
Education and Earnings                                                   Mexicos government should
progressively shift more of the
Inequality in Mexico                                                     costs of higher education to
its direct beneficiaries,
Ulrich Ldchler                                                          facilitating the private
absorption of those costs
through student loan
programs designed to correct
market failures in the financial
sector.
The World Bank
Mexico Country Department                                                         H
July 1998 



| POLICY RESEARCH WORKING PAPER 1949
Summary findings
Education attainment levels increased dramatically for        * The social rates of return across levels ol schooling
Mexico's labor force in the 1980s and early 1990s. In       were more uniform in 1994 than in 1984, suggesting a
parallel, the country experienced a pronounced increase     more efficient assignment of education spending. At the
in earnings inequality from 1984-94, reflected in a         same time, the distribution of spending on education
higher dispersion of wages and an absolute decline in the    became more egalitarian, as per student spending in
real incomes of less educated, poorer Mexicans. This       higher education declined markedly compared with per
increased wage dispersion presents policymakers with a      student spending at the primary level. This surprising
tradeoff between efficiency considerations (favoring        coincidence in the pattern of spending on education was
increased spending on higher education) and equity          only possible because Mexico started out with a very
considerations (favoring a more equal distribution of per    distorted resource allocation in education that was both
student spending) in the allocation of fiscal resources to  highly inequitable and inefficient. As Mexico's
education.                                                  policymakers are on the way to correcting these
Lachler concludes that the best way to deal with this    distortions, the opportunities for avoiding the equity-
equity-efficiency tradeoff is to encourage greater private  efficiency tradeoff within Mexico's centralized education
participation in higher education. His main findings are   framework will become progressively exhausted.
that:                                                         * There is little reason to expect the pace of
* The accumulation of human capital during 1984-         technological change, which appears mainly responsible
94, as proxied by education attainment, was                 for raising wage dispersion and the relative returns on
accompanied by a more equal distribution of education       higher education, to abate. Efficiency considerations
attainment levels over that period and, thus, exerted an    dictate that Mexico should respond by devoting more
equalizing effect on the distribution of incomes. The      resources to higher education. However, the federal
increased income inequality observed over that period       budget, which traditionally has financed the lion's share
appears to be caused by an increased rate of skill-based    of higher education costs in Mexico, is unable to
technological change, whose transmission to Mexico and      accommodate additional spending on higher education,
other developing countries may have been facilitated by    while spending cuts elsewhere in the education sector are
the increased openness of their economies.                  bound to raise serious equity questions. Thus, to avoid
The greater dispersion of wages observed in Mexico     falling behind in terms of human capital accumulation,
during the past decade raised the rates of return on        greater private sector participation is necessary, at least
investing in higher education, reversing the traditional   in terms of cost recovery from the main beneficiaries of
pattern where primary education exhibits the highest        higher education.
rates of return.
This paper - a product of the Mexico Country Department - emerged from research in the preparation of the 1998
Country Economic Memorandum and is part of the department's larger effort to engage in a fruitful economic policy
dialogue with Mexican policymakers. Copies of the paper are available free from the World Bank, 1818 H Street NAWT,
Washington, DC 20433. Please contact Carmen Lazcano, room 14-135, telephone 202-473-7776, fax 202-522-2113,
Internet address clazcano@worldbank.org. The author may be contacted at ulachler(worldbank.org. July 1998. (27
pages)
The Policy Research W'orking Paper Series diysemiates the findings of work isn progress to encourage the exchaege of ideas about I
development issues. An objective of the series is to get the findings out q,uickly, even if the presqentationis are less thanl fulfly polisbed. Th2e
papers carry the names of the authors aiid shouild be cited accordiiigly. The fipidinzgs, interpretations, and coniclusionis expressed ;1V this
paper are entirely those of the authors. They do not necessarily represent the vieuw of the WGorld Bankz, its Executive Directors, or- the
countries they represent.
Produced by the Policy Researclh Dissemination Ceniter



EDUCATION AND EARNINGS INEQUALITY IN MEXICO
Ulrich Lachler*
* Principal Economist, Mexico Country Department, World Bank.
This paper was prepared with research assistance from Gladys Lopez Acevedo, who
processed the ENIGH data discussed therein. Many helpful comments were
received from Barry Bosworth, Nancy Gillespie, George Psacharopoulos, Rosa-
Maria Rubalcava, Fernando Cortes, Enrique Hernandez Laos, and Eduardo Velez.



EDUCATION AND EARNINGS INEQUALITY IN MEXICO
The education attainment of Mexico's labor force increased dramatically
during the 1980s and early 1990s, contrasting markedly with the uneven accumulation of
physical capital. At the same time, the rates of return on investments in different levels
of schooling show significantly less dispersion than they did a decade ago. This suggests
that investment in education has been taking place in a more socially efficient manner.
Another visible development over the last decade, however, has been a significant
increase in earnings inequality, accompanied by an absolute decline in the real incomes of
the less educated, poorer members of society. This comes as something of a surprise in
light of the great equalizing properties generally attributed to education, but is a
phenomenon that in recent years also has been observed in other developing as well as
developed countries.
The increased wage dispersion presents policymakers with two challenges:
the more immediate one is how to respond to the decline in real incomes facing a large
share of the country's population. The other related challenge, with implications for the
country's long-term growth prospects, concerns a tradeoff in the allocation of resources in
education and is especially acute in relatively centralized education system such as
Mexico's. Since the increased wage dispersion raises the rate of return from investing in
higher education, economic efficiency considerations would dictate a response that
devotes relatively more resources to higher education versus other levels. A resource
reallocation to that effect, however, means transferring resources toward segments of the
population that are already better off, thus conflicting with equity considerations.
Mexico's policymakers were able to avoid this policy dilemma over the
last decade: as shown below, the allocation of public spending on education has become
somewhat more egalitarian and, at the same time, the social rates of return associated
with different levels of education have become more uniform. This happy coincidence
was possible only because Mexico started out with a very distorted resource allocation in
education. As past resource misallocations are corrected, however, the opportunities for
further improvements in resource allocation within the existing, centralized education
framework are progressively exhausted. This is likely to result in increasing tensions in
the allocation of fiscal resources -- between efficiency considerations that argue for more
resources to higher education and egalitarian considerations that argue for a more equal
distribution of transfers within and outside the education sector. This paper argues that
the most promising way of dealing with these tensions is by seeking to clarify the roles
for the public and private sectors in education and by encouraging greater private
participation in higher education.



2
A. The Growth of Education Attainment1
Education attainment levels increased rapidly in most developing
countries since the 1950s; Schultz (1988). While Mexico also partook in that
development, earlier studies had identified a significant lag in its education indicators.
Londonto (1996), for example, points to an "education deficit", according to which the
Latin American countries in general, and Mexico in particular, have approximately two
years less of education than would be expected for their level of development (as
measured by per-capita incomes). Elias (1992) finds that education was the most
important source of labor quality improvement in Latin America between 1950 and 1970,
but points out that such improvements did not take place to the same extent in Mexico.
This changed dramatically in the 1980s, as shown in the scatter diagram below.
Figure 1 describes the relation between income per capita and average
years of schooling of the population aged 15 years or more, using stacked cross-country
data for the years 1960, 1970, 1980, 1985 and 1990.2 Mexico's level of education
'The following discussion primarily focuses on the evolution of education attainment levels as measured
by years of schooling, on the assumption that the quality of education has remained more or less
constant. Although it is of great interest in this context, information on the quality of education in
Mexico is very scarce. The most common measures of "quality" are based on input measures, (e.g.,
real expenditures on education per student, student-teacher ratios, or class-density variables) or
secondary performance variables (e.g., evolution of repetition and desertion rates). On the basis of
these measures, Mexico also has performed comparatively well over the last few decades. These
measures are not very satisfactory, however, as they reflect many other influences in addition to the
quality of instruction as reflected in the acquisition of knowledge and skills. A comprehensive
education survey was conducted in Mexico in 1995 as part of the Third International Mathematics
and Science Study (TIMMS) that may have yielded important data to permit correlating the preceding
education sector characteristics with academic achievement levels to obtain a better understanding of
the determinants of education quality in Mexico. That data set has not been made publicly available.
2 The scatter diagram in Figure 1 is based on 317 observations from five different years. The observations
pertaining to Mexico, ordered by date, are as follows:
Ave. Schooling (years) Ln (GDP per capita; 1980 US$)
1960          2.76                   7.95
1970          3.68                   8.29
1980          4.77                   8.71
1985          5.20                   8.63
1990          6.72                   8.67
The trend line represents the least square regression line given by:
S = -13.17 + 2.28 Ln(GDPcap)         Adj. R2 = 0.68
(-18.7)   (26.0)                  t-values in parentheses
The application of Ramsey's RESET test to this regression equation failed to detect a specification error;
unlike with the alternative specification of type: S = a + bX + cX2. However, coefficient stability
tests indicated that the trend line is not constant across decades. This is reflected in the upward drift
of the coefficients associated with the dummies in the following equation:
S = -13.20 + 2.21 Ln(GDPcap) + 0.023DUM70 + 0.525DUM80 + 0.833DUM 85 + 0.997DUM90
(-19.0) (25.4)           (0.1)         (1.7)         (3.0)          (3.6)
Adj. R2 = 0.70, t-values in parentheses



Figure 1
Cross-Country Relation between Education Attainment and GDP
12 -
10                                                                         :*.+#   .
Mexico 1990
8 
6
4
o
0
5.00       S.S0        6.00        6.50       7.00        7.50        8.00       8.50        9.00        9.50        10.00
Ln (GDP per capita)
Source: Penn World Tables, mark 5.6, and Barro-Lee data set on international measures of schooling.
The chart is based on stacked cross-country data for 1960, 1970, 1980, 1985 and 1990. Country coverage varies according to data availability.
U.1-achler



4
attainment in 1960 was significantly below the world mean for countries at similar levels
of economic development. Although Mexico's education attainment increased steadily
over the following two decades, it continued to remain below the international trend line.
In the 1980s, however, the growth of education attainment in Mexico accelerated,
permitting it to catch up to international standards by 1990; where its placement in Figure
1 is slightly above the trend line.
The closure of Mexico's education gap vis-a-vis the rest of the world was
hastened in part by the country's economic stagnation. Mexico's real GDP per capita in
the mid-1990s was roughly the same as it had been in the first half of the 1980s. Had
Mexico continued to grow at the same pace as in the 1960s and 1970s (and assuming that
the gains in education attainment remain the same), its 1990 placement in Figure I would
have continued to remain below the cross-country trend line, though at a much reduced
distance compared to previous years.3
The preceding observation, however, should not detract from the
remarkable increase in schooling that occurred during the 1980s. While average
schooling level in Mexico increased by roughly one year per decade during 1960-1980
(from 2.76 to 4.77 years), it increased by two years in the decade between 1980-1990. As
described in Psacharopoulos et al (1996), this rapid improvement reflects the great efforts
made in Mexico to increase both the quality of and access to public education since 1950.
A consequence of these developments is that the share of workers with less than primary
education decreased from almost half of the labor force in 1984 to 36 percent in 1994,
while the share of workers with at least a completed secondary education increased from
26 to 39 percent; Table 1.
The rapid growth of education attainment in Mexico also stands out in
recent cross-country growth studies (e.g., Bosworth, 1997), which decompose per capita
growth into the contributions from several factors, including education attainment as a
proxy for human capital accumulation, in a growth accounting framework. These studies
reveal a major break in Mexico's growth performance after 1982. A similar break is
visible in the accumulation of physical capital and in total factor productivity growth, but
not in the accumulation of schooling, which performed well by world standards.
3 Mexico's education deficit -- i.e., the vertical difference from the world trend line in Figure 1 -- has
declined significantly since 1960, with the main catch-up occurring in 1980-90. This is clearly visible
from Figure 1, but also appears under less restrictive specifications of the estimated trend line. The
differences from the trendline are stated below, beginning with the most restrictive specification:
Equation Specification         1960  1970  1980  1985  1990   Std. Error of
(difference from world mean in years)  Regression
Stacked, without dummies      -2.22  -2.09  -1.95  -1.36 +0.08    1.52
Stacked with dummy variables  -1.62  -1.49  -1.82  -1.54  -0.26   1.48
Individual equations for each year  -1.55  -1.41  -1.84  -1.59  -0.26  na



5
B. Changes in Earnings Inequality
At the same time as this remarkable advance in education attainment was
taking place, the distribution of income in Mexico worsened notably; see e.g., De la Torre
(1997), Panuco-Laguette and Szekely (1996). For example, the Gini coefficient of
Mexico's total income distribution increased from 0.43 in 1984 to 0.48 in 1994. This
deterioration took place before the 1995 recession and, thus, cannot be attributed to
business cycle effects. Instead, the increase in overall income inequality appears to be
closely related to an increase in the dispersion of wages and salaries across different
schooling levels. Table 1 describes the change in real wages between 1984 and 1994,
using two concepts of remuneration (described in Section D.)
Table 1: Real Wage Rates in Mexico
(Unless stated otherwise, all figures refer to hourly rates expressed in constant 1994 Pesos)
Wages & Salaries         Monetary Income        Share of Workers
Schooling Level        1984  1994  difference  1984   1994  difference    1984  1994
0 - less than primary    3.17   3.08    -2.8%      4.83   4.12  -14.7%     48.1%  35.7%
1 - primary complete     5.23    4.42   -15.5%    7.90   5.39  -31.8%      26.3%  25.4%
2 - secondary complete    6.55    5.83   -11.0%    7.55    6.90   -8.6%     13.3%  21.3%
3 - preparatory complete   9.62   11.68   21.4%    10.64   12.84   20.7%    7.8%   11.1%
4 - university and above   14.93  21.96   47.1%    16.94   25.55   50.8%    4.5%    6.6%
weighted average    5.62   6.88    22.4%    7.00    7.57    8.1%
Source: Own calculations based on ENIGH84 and ENIGH94. All figures are weighted averages using
expansion factors to reflect national representation.
Even though the average wage increased during the decade spanned by
Table 1, most workers experienced a significant decline in their wages. This disparity in
the evolution of real wages is most visible in Figure 2, which shows that 83 percent of the
working population experienced a decline in real wages during the last decade.
Figure 2
MEXICO
Percentage Change in Real Wages by Percentile: 1984-1994
40.00%
30.00%0
20.00%    ~~based on                           .6 
a , 10.00%
0.00%
e    -10.00%   10 15 20 25 30 3 40 45 50 55 606570  85 90 95
CL  -20.00%
.-30.00%       ~-ory monetary incomes
-40.00%
Percentiles
Source: ENIGH84 and ENIGH94.



6
The deterioration of real wages was partly compensated for through
increases in hours worked, suggesting backward bending labor supply curves at low
income levels. This was not enough, however, to prevent an erosion of overall income
for the majority of workers; Table 2.
Table 2: Average Hours Worked and Income Received
Schooling Level                   Wage Earners           Monetary Income
Recipients
Avg. Hours per Week        1984  1994 change         1984  1994  change
0 - less than primary          44.8  50.7   13.3%      43.0  46.8    8.8%
1 - primary complete           45.1  49.0    8.6%      44.7  48.2    7.9%
2 - secondary complete         44.4  46.8    5.6%       44.5  46.8    5.2%
3 - preparatory complete       38.7  43.7   12.8%       38.9  44.1   13.4%
4 - university and above       41.6  44.7    7.5%       42.2  44.7    5.9%
weighted average         44.1  47.8    8.5%       43.3  46.7      7.8%
Avg. Annual Income    1984   1994  change         1984   1994  change
(in 1994 Pesos)
0 - less than primary          7,530  7,674    1.9%     9,514  8,308  -12.7%
1 - primary complete          12,705 11,316  -10.9%    15,888 13,222  -16.8%
2 - secondary complete        15,343 14,097   -8.1%    16,995 15,978   -6.0%
3 - preparatory complete      18,643 25,709   37.9%    19,483 27,898    43.2%
4 - university and above      33,899 55,428   63.5%    38,741 62,769    62.0%
weighted average     13,057 16,658   27.6%    14,268 16,931    18.7%
Source: Own calculations based on ENIGH84 and ENIGH94. All figures are weighted
averages using expansion factors to achieve representation at the national level.
One explanation for a deteriorating income distribution that is closely
connected to the advance in education attainment has been made by Ram (1 990).
Education is generally considered to have an equalizing effect on incomes, which would
indicate a positive relation between measures of income inequality and of education
inequality. However, the dispersion of schooling attainment and the mean level of
schooling need not be positively related in a monotonic manner. Upon exploring this
relationship by way of a cross-country analysis, Ram finds a curvilinear (Kuznets-type
inverted-U) relation between the mean level of schooling and schooling inequality in the
labor force.4 That is, as schooling expands, education inequality first increases, but then
starts declining after reaching a peak. That turning point occurs when mean schooling is
4The theoretical rationale underlying this fmding is reminiscent of the "Laffer curve" in public finance
theory. It stated that tax revenues are 0 when tax rates are 0 and will again become 0 when the tax rate
approaches 100% (as that would eliminate all incentive to work.) Since revenues are positive at tax
rates between 0% and 100%, the revenue-rate relationship must broadly approximate an inverted-U.
In the case of education, the "theory" states that when no one in society is educated, the distribution
of education is perfectly equal, as it would be when everyone in society has a Ph.D. (considered for
simplicity to be the maximal level of education attainment). In the process of going from a Zero-
education society to the Land of Ph.D.s, the distribution of education must perforce become more
unequal before it ultimately improves again.



7
about 6.8 years for the full sample of countries, and 6.3 years for the sub-sample of less
developed countries. So, if Mexican schooling levels had been below the turning point, it
is conceivable that the rapid increase of education attainment in the 1980s may have
contributed to greater earnings inequality via an increased inequality of schooling
attainment.
The evidence does not support this explanation for the increased earnings
inequality in Mexico: using Ram's (1990) method of calculation based on 5 schooling
levels, the mean level of schooling in Mexico was 6.45 years in 1984 and 7.65 years in
1994. This means that Mexico started in 1984 with a mean level of schooling that was
slightly above the "turning point" according to the LDC subsample and slightly below the
turning point for the full sample of countries, while it ended up in 1994 with a mean
schooling level substantially above the turning point. If the curvilinear relation is evenly
distributed around the turning point, we should clearly expect the inequality of schooling
in Mexico to have declined over that decade.
This conclusion is confirmed by decomposing the change of the Gini
Index between 1984 and 1994 into the proportion that is attributable to changes in the
wage rate, in hours worked and in education attainment. Table 3 presents the results of
such an exercise, both for the distribution of Wage and Salary incomes, as well as the
broader Monetary income concept. For each income concept, several "synthetic" Gini
indexes are calculated:5 the first two, Gini-1984 and Gini-1994, are based on the
distribution of wage rates, hours worked and education attainment that pertain in each
year. The next three indexes are derived by recalculating the 1984 Gini index after
successively replacing the 1984 distribution of wage rates, hours worked and schooling
attainment by its 1994 counterpart. The second column under each income concept
reports the value of each Gini index calculated in this manner, while the third column
reports the percentage difference of these values relative to the original 1984 Gini index.
Table 3: Sources of Increased Earnings Inequality
Wages & Salaries                  Monetary Income
Level %change over                Level %change over
Gini-84                           Gini-84
Gini 1984     0.2281    0.0       Gini 1984      0.1712    0.0
Gini 1994     0.3034   33.0       Gini 1994      0.2812    64.3
Gini-Awage    0.2962   29.9       Gini-Awage     0.2543    48.5
Gini-Ahours    0.2208   -3.2      Gini-Ahours     0.1696   -0.9
Gini-Aeducation  0.2222   -2.6     Gini-Aeducation  0.1710    -0.1
Source: Own calculations based on data from Tables I and 2.
Note: the %-changes in the lower half of each column do not add up to the %-difference
between Gini-94 and Gini-84 due to large rounding errors.
Table 3 indicates that the "synthetic" Gini index increased by 33 percent
between 1984 and 1994 (under the Wage and Salary income concept) and that over 90
5 These "synthetic" Gini indexes are constructed under the assumption that all the individuals with a given
schooling level earn the same wage rate and work the same number of hours as the averages shown
for each schooling level in Table 2.



8
percent (29.9 . 33.0) of that increase in earnings inequality is attributable to the change in
the distribution of wage rates over that decade. (Using the broader Monetary income
concept, around 75 percent of the increase in earnings inequality is explained by the
change in wage rates.) On the other hand, changes in the distribution of education
attainment are shown in both cases to have contributed to a reduction in earnings
inequality, though by modest amounts; -2.6 and -0.1 percent. Changes in hours worked
also contributed modestly toward reducing income inequality. In light of this evidence,
we can safely rule out increased schooling inequality as an explanation of the increased
income inequality that took place in Mexico over the last decade.
Three broad hypotheses that do not hinge on changes in the distribution of
education attainment are frequently advanced to explain the similar increases in earnings
inequality experienced in Mexico and other countries.6 These link the increase of
earnings inequality to (i) the increased openness of the economy, (ii) institutional changes
in the labor market, and (iii) skill-biased technological change. The first of these
hypotheses argues that as trade barriers are reduced, an economy is placed under
increased competitive pressures to specialize along its lines of comparative advantage. A
developed country that is relatively high skill-abundant, like the United States, will be
induced to specialize in high skill- or education-intensive activities as its low-skill
industries come under increased competitive pressure from low skill-abundant, low-wage
countries.
This explanation has several problems when applied to the United States,
and becomes even less persuasive when applied to Mexico. Mexico greatly liberalized its
trade regime since 1984. However, the reduction of its trade barriers has mostly been vis-
a-vis imports from the developed countries, notably the United States and Canada, whose
share in total Mexican merchandise imports increased from 68 percent in 1985 to 73
percent in 1993 (and to 77.5 percent in 1996). Since Mexico is a low skill-abundant
country compared to its two northern neighbors, it would be expected that the
liberalization of trade Would have induced a specialization pattern that would raise the
relative demand (and hence wages) for the lesser educated members of the labor force.
This did not happen. Instead, the increase in earnings inequality observed in Mexico is
identical to that observed in the United States: less educated workers experienced real
wage declines, while highly educated workers experienced real wage improvements. 7
6 See, for example, the Symposium on Wage Inequality published in Journal of Economic Perspectives
(1997) and the Symposium on "How International Exchange, Technology and Institutions Affect
Workers" published in The World Bank Economic Review (1997).
7 The trade-based explanation may still be relevant to the extent that greater openness facilitates the transfer
of ideas and technology, which is identified below as the more persuasive explanation of the increase
in earnings inequality. A variant on the globalization/technology nexus explanation, advanced by
Feenstra and Hanson (1994), involves outsourcing behavior where multinational enterprises in the
developed country relocate their lower skill-intensive activities to the less skill-abundant developed
countries. However, what is a low skill activity in the United States may be a high-skill activity in
Mexico, which could explain the identical evolution of earnings inequality in both countries.



9
The second explanation revolves around institutional changes such as
reductions in the minimum wage, the decreasing strength of trade unions and the
declining share of state-owned enterprises. The existence of a binding minimum wage,
for example, truncates the lower end of the wage distribution. As the minimum wage is
allowed to erode away, say through inflation, it becomes less binding by moving further
down the low end of the wage distribution, with the result that, ceteris paribus, a higher
share of wages will lie below the previous minimum wage level. That translates into an
increased dispersion in wages and earnings. Similarly, strong trade unions have often
been found to exert an egalitarian effect on the wage distribution, while at the same time
commanding a wage premium for union members. Any waning of union strength, as
happened in the United States over the last two decades, therefore would contribute to an
increase in wage dispersion. A review of institutional developments in the Mexican labor
market by Hemrndez et al (1998) showed that these factors may have contributed to a
higher wage dispersion. Most important in this regard was the apparent decline in union
strength, in part due to the privatization of public enterprises during the early 1990s. That
review also showed that the real decline of minimum wages since the early 1980s may
have had an impact as well, but mainly confined to wages in the primary sector.
Perhaps the most persuasive explanation, both for the United States and
Mexico, is that which links earnings inequality to skill-biased technological changes that
raise the relative demand for higher-skilled labor. According to the typology used by
Johnson (1997), the type of technological change that drives up the wages of more highly
skilled workers and drives down those of less skilled workers -- as occurred in both the
United States and Mexico -- is extensive skill-biased technological change. Under this
type of technological change, skilled workers become more efficient in jobs that were
traditionally performed by unskilled workers. A demand/supply analysis in Hernandez et
al (1998) provides support for this explanation in the case of Mexico: it showed that
secular changes in demand for workers differentiated by education level were the
dominant force behind recent adjustments in the Mexican labor market, with the market
for more educated workers dominated by intra-sectoral increases in demand, and for less
educated workers by intra-sectoral decreases in demand, independent of gender and
experience. That outcome suggests that the rise in wage dispersion originates from
changes in the intra-sectoral production structure, which favors educated versus
uneducated workers, rather than from changing trade patterns, which would affect the
demand for labor differently across sectors.8
C. Some Background Facts on Education Spending
Mexico's overall spending on education, as a share of GDP, is slightly
above the OECD average and compares favorably with education spending levels in other
developing countries. Total spending per student in Mexico increased steeply in the
second half of the 1980s and early 1990s, even though the total student population
increased, from 24 million in 1985 to 27 mnillion in 1995. Part of this increase reflects a
sThese findings correlate with earlier findings for Mexico (Cragg and Epelbaum, 1994) as well as for
Chile, following its trade liberalization program (Robbins, 1994).



10
rebound from the expenditure cuts that were made in response to the debt crisis in 1982.
By 1994, however, total spending on education had risen to 5.4 percent of GDP, or
almost a full percentage point above its previous peak of 4.5 percent reached in 1982.
Figure 3
MEXICO
Education Spending per Student Capita
3500.00 -
8 3000.00                       Prvate
o 2500.00
Z2000.00-
1500.00-                                     Muipa
aC
1000.00
O 500.00                 Federal
Source: Informe de Gobiemo, 1989 and 1997.
Figure 3 describes the evolution of funding for education in Mexico since
the early 1980s. Private spending represenited less than 10 percent of total expenditures
in education in 1987, and has fallen to below 5 percent in recent years. (This figure
somewhat underestimates total private costs, which, with some adjustments noted in the
Annex could be around 27% of total education spending.) Similarly, the share of state
and municipal spending peaked at 17% of total spending in 1990, but since then also
declined to less than 10 percent of total spending. The federal government, on the other
hand, currently accounts for 87 percent of total sector spending. These figures are not far
out of line with those from an OECD review of education expenditures, which indicated
that in 1994, about 19 percent of spending on education in Mexico came from private
sources, compared to a country average of 13 percent for the whole of the OECD.9
Mexico does not stand out particularly in the preceding comparison. As in
most countries, the public sector directly or indirectly finances the lion's share of
education expenditures. Mexico does stand out in contrast to its OECD partners,
however, when these expenditures are disaggregated by level of instruction. As shown in
OECD (1998, pg. 97) the ratio of spending per student at the tertiary level compared to
spending at the primary level exceeded 500 percent in 1994 and was the highest among
OECD countries. (The ratio in Korea, for example, is close to the OECD average of 264
percent, or less than half the Mexican ratio.)
9These figures comprise total expenditures on education, including privately purchased materials, whereas
the earlier figures shown in Table 3 are identical to the what the OECD expenditure review refers to
as direct spending on educational institutions.



11
Table 4: Federal Education Spending per Student by Level
Federal Spending per Student
(constant 1994 Pesos)         (as share spent on Primary)
primary secondary preparatory university  prim. second. prep. univ.
1980   1166.1   2571.8   4823.7    8667.3        1     2.2   4.1  7.4
1981    1447.6   2790.6   6156.9   11544.0       1     1.9   4.3  8.0
1982    935.8   2027.3   4586.4   10061.0        1     2.2   4.9  10.8
1983    710.5   1562.0   2714.7    6451.4        1     2.2   3.8  9.1
1984    691.1   1644.3  3108.4    6544.2         1     2.4   4.5  9.5
1985    715.0   1473.3  3582.2    7174.8         1     2.1   5.0  10.0
1986    961.5   2094.6  3329.3    11686.2        '     2.2   3.5  12.2
1987    683.7   1308.3  3364.0    6334.3         1     1.9   4.9  9.3
1988    665.1   1303.0  3423.0    6520.2         1     2.0   5.1  9.8
1989    734.0   1438.9  3756.0    6176.3         1     2.0   5.1  8.4
1990    808.4   1685.8  3502.8    6473.5         1     2.1   4.3   8.0
1991    1011.4   1942.6  3772.9    6898.3        1     1.9   3.7  6.8
1992   1285.9   2318.7  3384.3       9313.2      1     1.8   2.6  7.2
1993   1584.6   2568.4  4304.6    11637.2        1     1.6   2.7  7.3
1994   1731.3   3139.4  5215.1    13253.0        1     1.8   3.1  7.7
1995    1679.0   2511.0  4508.1    11202.3       1     1.5   2.7  6.7
1996    1647.7   2464.2  4442.9    11328.5       1     1.5   2.7  6.9
1997    1811.9   2709.7  4882.1    12692.2       1     1.5   2.7  7.0
Source: Presidencia de la Republica, Primer Informe de Gobiemo, 1989, and Tercer
Informe de Gobiemo, 1997.
The great disparity in per-student spending at different education levels in
Mexico is also reflected in Table 4, which only refers to federal expenditures. This Table
also shows, however, that the evolution of public spending on education in Mexico has
become more egalitarian across different schooling categories over time. In the early
1980s, the amount of federal spending per university student averaged 10 times the
amount spent per primary student. This ratio fell to around 7 times in the early 1 990s.
Federal spending on the other levels relative to the primary level indicate a similar
decline, even though the absolute amounts increased at all levels. In other words, even
though the recent pattern of spending on education in Mexico exhibited a highly uneven
bias in favor of tertiary education relative to other categories, the pattern of spending in
earlier periods exhibited and even greater bias.
D. The Rates of Return on Investments in Education
One way to assess the economic merits of the preceding spending patterns,
considered as investments in education, is by comparing their relative rates of return. This
section compares the rates of return to education as derived from several human capital
earnings functions. These were estimated with data from two income-expenditure
surveys (ENIGH) carried out by the National Statistical Institute (INEGI) in the third
quarters of 1984 and 1994. These two survey years provide an excellent setting for
comparisons: first, both years can be considered macroeconomically comparable in that
the economy is neither in recession nor booming. Secondly, both years are almost equi-
distant in time from the major stabilization and structural reform measures implemented



12
by the Mexican Government in the latter half of the 1980s. They are especially suitable,
therefore, for making 'before and after' comparisons.
The ENIGH surveys identify the educational attainment, income received
and number of hours worked per week by each family member.'0 Income is differentiated
into about 25 items, which were aggregated into three broad categories: (i) wage and
salary income, (ii) current monetary income, which includes wages and salaries, income
from self-employment, property income and rents, monetary transfers and income from
financial assets, and (iii) total current income, which includes all of the above items, plus
non-monetized income such as imputed rent, in-kind transfers and stock dividends. In
contrast to this disaggregation of income sources, the ENIGH surveys do not differentiate
the number of hours worked per week by activity. The rates of return calculated below
are based on two measures of earnings per hour: the wage rate, which refers to wage and
salary earnings per week, divided by the total number of hours worked per week, and the
monetary earnings rate, which refers to current monetary income divided by the same
number of hours worked per week.
Tables 5 and 6 present the private and the public rates of return on
investments in education in Mexico. The method used to estimate these rates of return is
explained in detail in the Annex. The distinction between public and private rates of
return in this context centers around who bears the cost of financing the investment in
education: in both cases, the benefits of an additional year of schooling are reflected in
the stream of additional earnings that accrue to an individual as a result of that additional
education. The rate of return to education is then the discount rate that equates the
present value of this stream of additional earnings to the present value of the costs
incurred. When only private costs are considered in this calculation, it is tenned the
private rate of return, and when all public and private costs are considered, it is termed
the social rate of retum.
Table 5: Private Rates of Return to Schooling in Mexico
(Private cost of education is 100% of foregone full year's earnings; KP = 1.)
Wage Earners      Monetary Income Recipients
1984        1994         1984       1994
Average "Mincerian"        15.2%        16.7%        14.6%       17.2%
by schooling level
1 - primary complete       16.9%        13.8%       14.7%        14.3%
2 - secondary complete     13.9%        16.7%        14.5%       16.0%
3 - preparatory complete   15.5%        18.2%        14.4%       19.2%
4 - university and above   10.3%        19.9%        12.8%       20.6%
Source: Tables Aland A2, using the method described in the Annex.
°Several sources of potential bias in the data base require attention: first, casual evidence indicates that the
sample of families included in the ENIGH surveys is truncated in that the very rich and the very poor
appear to be underrepresented. Indigenous groups, which are estimated at around 1o percent of
Mexico's population, also appear underrepresented. Finally, there appear to have been some
sampling problems involving the 1984 survey (the first year the survey was conducted by INEGI),
which may have affected its comparability with the 1994 survey.



13
Two phenomena about the private rates of return draw attention in Table
5: the first is that the overall rate of return to education has increased between 1984 and
1994, as seen from the "Mincerian" rates of return. This is a direct consequence of the
increased dispersion in wages that took place over that period. Secondly, the rates of
return for different levels of schooling have changed in ranking. In 1984, the lower levels
of education uniformly yielded greater rates of return than investments in higher levels of
education. This ranking is completely reversed in 1994, with the private rates of return to
primary education declining significantly below the rates of return to higher education,
which almost doubled."1
Table 6: Social Rates of Return to Schooling in Mexico
Wage Earners     Monetary Income Recipients
Schooling Level           1984        1994         1984       1994
l-priinarycomplete        15.5%      11.3%        13.7%      11.8%
2 - secondary complete    12.4%       13.0%       13.1%       12.7%
3 - preparatory complete   12.9%      13.2%       12.2%       14.4%
44-university and above    7.6%       13.2%        9.6%       13.9%
Source: Tables Aland A2, using the method described in the Annex.
As in the case of the private rates of return, the social rates in Table 6 also
indicate a decline in the rate of return for primary education and increases in the rates of
returns for higher education during the decade between 1984 and 1994. A noteworthy
difference to the private rates of return, however, is that the social rates of return exhibit a
much lower dispersion across schooling levels in 1994 than in 1984. That is, the social
rates of return associated with different investments are much closer together, which, in a
standard capital market context, suggests a decline in market segmentation and, thus, a
more efficient allocation of investment resources in education.
E. A Policy Dilemma
A general efficiency criterion for investing is that resources should be first
invested in those activities that yield the highest rate of return until the point where the
marginal rate of return drops to the level of the activity with the next highest rate of
return. Barring any corner solutions and assuming that investments in education exhibit
decreasing marginal rates of return, this rule will have the effect of equating rates of
return across different activities. Suppose that the government were following this
efficiency rule in its allocation of public resources in education in the face of increased
wage dispersion. As shown in the Annex, that would mean that the pattern of public
spending in education would have to adjust in the same direction as the private rates of
" Theprivate rates of return in Table 5 pertaining to 1984 are very similar to the private rates reported for
Mexico in Psacharopoulos (1994a). The main difference arises with regard to primary education,
where his paper reports a significantly higher return. The reason is because Psacharopoulos assumes
that the foregone earnings from attending primary school are roughly half the figure used in our
calculations, which are based on Annex Table A3. Psacharopoulos' lower assumption is meant to
adjust for the very low opportunity cost of young children attending primary school.



14
return to education. In other words, as the dispersion of wages increases, the private rates
of return on investments in higher education increase. Other things equal, that also raises
the social rates of return for higher education levels in the same proportion. Therefore, to
prevent a misalignment of the social rates of return in education, the government would
have to shift resources away from primary education (which raises its rate of return) and
toward higher education (thereby causing a compensating reduction in its marginal rate of
return).
The preceding response, based on efficiency considerations, would
conflict with equity considerations in the short run, though not necessarily in terms of
long term consequences. Insofar as the primary level encompasses a higher proportion of
individuals from poor households than do higher education levels, the efficient response
would change the pattern of direct spending toward more affluent individuals. Moreover,
that change in spending would come about after the already existing income disparities
across households had been aggravated by the movement in relative wages. Over time,
the increased spending on higher education leads to a higher relative supply of more
educated workers, which depresses their relative wage and, thereby, has an equalizing
effect on total earnings. That equalizing consequence occurs in the long run, however,
and would have little impact on the earnings of the currently affected generation of
workers.
A surprising thing about Mexico over the last decade is that even though
public spending has become more evenly distributed, the social rates of return in Table 6
also have become more uniform in 1994 compared to 1984. The answer to this puzzle
lies in the very distorted pattern of education spending that existed at the beginning of the
period. Public education spending in the early 1980s was very skewed toward higher
education, which exhibited the lowest rates of return. That is, it was both inefficient and
inequitable.'2 By moving toward a more even distribution of per capita spending across
different levels, equity was improved. At the same time, the external environment
changed in a manner that raised the relative return to higher education, thereby tending to
make more efficient what had initially been an inefficient allocation of resources.
As the social rates of return to education become equalized, however, the
opportunities for achieving gains by undoing past mistakes will be exhausted. That
means that policymakers will no longer be able to evade this equity-efficiency tradeoff,
assuming that the rate of skill-biased technological change continues as before. So long
as education spending remains concentrated at the federal government level, that same
government will have to decide whether to respond to exogenous increases in the rates of
return to higher education by shifting more resources toward higher education at the
expense of increased inequity in spending, or by maintaining a more equitable spending
distribution at the expense of increased inefficiency.
12 This observation is consistent with earlier findings by Elias (1992), which indicated that growth in the
"quality" of the labor force contributed little to Mexico's overall economic growth in the 1960s and
1970s, compared with other Latin American countries.



15
There is no easy response to this dilemma. The politically least
controversial response of assigning more resources to higher education through a higher
overall budget to education runs into fiscal constraints."3 Barring more public resources,
the only option for expanding of investment in higher education (desired on efficiency
grounds) without cutting public funding for primary education is by attracting greater
private sector participation.
To facilitate greater private participation, it is useful to distinguish more
clearly the roles of the private and public sectors. The classic assignment of
responsibilities limits the public sector's involvement to the provision of public goods, as
well as correction for market failures and externalities. In that context, it is often argued
that the positive externalities from education are highest at the primary level and
progressively decline at the higher levels; Schultz (1988). That is, at higher education
levels, the individuals receiving the education capture a progressively larger share of the
total social benefits yielded from the education in the form of higher wages. That being
the case, there is less of a risk that private individuals left to their own devices would
underinvest from a social viewpoint in higher education than that they would underinvest
in primary education. The externalities from different levels of education are difficult to
measure and compare empirically. However, if the rationale for state intervention in the
provision of higher education services is based on the argument that rates of return were
too low to induce the right amount of investment from the private sector, that rationale
has certainly become less compelling with the significant increases in private rates of
return observed in recent years.
The options for greater private sector participation in higher education are
many. They range from greater participation in the funding of education services to
greater participation in the direct provision of services.14 At the very least, the rise in
private rates of return for higher education makes a compelling case for increasing the
level of cost-recovery in higher education, independently of whether it is the State or the
private sector that provides those services. The lack of long-term financing for private
investments in higher education represents an important obstacle to the implementation of
this solution. That obstacle is rooted in a systemic market failure (due, among other, to
bans on indentured servitude) that poses problems in most countries, but is especially
pronounced in Mexico on account of its very weak financial system. The proper role of
government is precisely to correct or compensate for such market failures. A promising
way to correct these market failures in Mexico is through student loan programs, or
means-tested financial aid and scholarship programs. These programs are currently
almost non-existent in Mexico, as the public sector has relied mainly on the direct, cost-
free provision of higher education services. The scholarship programs suggested here as
an alternative promise to yield a more efficient resource allocation than a cost-free
3 A sense of the potential magnitude of these fiscal costs can be obtained from OECD (1997, Chart B 1.3).
It estimates that raising Mexico's enrollment ratios in tertiary education to the OECD average level,
would involve an increase of expenditures on educational institutions by 1.4 percent of GDP. This is
almost twice the size of the fiscal revenue shock (estimated at 0.8 percent of GDP) caused by the
decline of oil prices in late 1997 and early 1998.
14 See OECD (1997, pg. 54) for a list of various options for the funding of tertiary institutions.



16
provision of services. Although such programs have rarely been devoid of subsidy
components wherever they have been implemented worldwide, that subsidy is more
closely targeted to the source of the market failure."5
Looking toward the future, there are two developments that bode well for
the distribution of income in Mexico in future years. One is that Mexico has gotten past
"Ram's turning point" (section B), so that further advances in the average schooling level
should result in a more even distribution of education across the labor force. Other things
being equal, that should translate into a more equal distribution of earnings. The other
development is that the supply of workers with higher education is increasing, which is
the desired market response to the increased wage premium on higher education. The
increasing supply of more educated workers should eventually reduce the wage premium
received by such workers and, thus, also tend to equalize the earnings distribution. In
view of the long gestation periods associated with investments in education, however, the
income-equalizing dynamics in both cases operate with a long-term horizon. Any
reallocations in education expenditures, therefore, should be made with that horizon in
mind and not be seen as substitutes for income transfers designed to compensate for
current income inequalities.
F. Summary
This chapter arrived at six principal conclusions and one recommendation:
*  The accumulation of human capital, as proxied by education attainment, does not
appear to be among the factors responsible for Mexico's disappointing growth
performance since the early 1980s, but rather stands out positively in historical and
international comparisons.
*  Mexico experienced a pronounced increase in earnings inequality during the period
1984-94, in spite of a rapid expansion of education attainment levels that exerted an
equalizing impact on earnings. Of the various hypotheses that have been advanced to
explain the increased earnings inequality in Mexico as well as other developing and
developed countries, the most persuasive appears to be that it is caused by an
increased rate of skill-biased technological change, whose transmission to developing
countries may have been facilitated by the increased openness of those economies.
The preceding analysis contains an interesting parallel to Beristain's (1991) analysis of options to meet
the rapidly increasing demand for higher education foreseen in Mexico for the 1990s. In his analysis,
the demand for higher education emerges from exogenous demographic pressures. In this paper, the
increased demand results from exogenous increases in the dispersion of wages observed in Mexico
and other countries. Among the response options outlined by Beristain, the first is to maintain the
status quo, which mainly involves the provision of services by public universities on a cost-free basis.
The second option is to charge an increasing share of education costs to beneficiaries, while keeping
the provision of service in public hands. The third option goes one step further by transferring an
increasing portion of the provision of services to private hands. He concludes that Mexico's fiscal
constraints and political considerations render the first option non-viable, so that increased demand
for university education is best accomodated through a response that involves a combination of the
second and third option. The conclusions of this paper also point in the same direction.



17
* The increased earnings inequality is associated closely with a higher dispersion of the
average wages received by workers with different schooling attainment. This had the
consequence, in turn, of raising the private rates of return to higher levels of
education; in effect reversing the traditional pattern of rates of return, where the
highest rates are reported for the primary level. The social rates of return also exhibit
this reversal in the relative magnitude of the rates of return. In contrast to the
increased dispersion of the private rates, however, the social rates of return across
different schooling levels have become more uniform in 1994, compared with 1984.
This suggests that the assignment of education resources has become more efficient.
* Government spending on education rose significantly in the early 1990s. It also
became more equitably distributed as per-student spending in higher education
declined markedly compared to per-student spending at the primary level. Spending
on education continues to be very concentrated in the public sector, especially at the
federal level.
- That a more equitable distribution of public spending on education was achieved
simultaneously with a more efficient spending pattern -- as reflected in the reduced
dispersion of social rates of return -- is very surprising. Normally one expects a
tradeoff between equity and efficiency in this context. What explains this result is
that the spending pattern in the early 1980s was both highly inefficient and
inequitable. So reduced spending at higher levels made for a more equitable
distribution of resources, while simultaneous changes in the external environment that
raised the returns on higher schooling levels, caused the allocation of resources
(which continues to be biased toward higher education) to become more efficient.
- As the social rates of return are equalized, the possibilities of exploiting initial
inequities and inefficiencies will become exhausted. Furthermore, there is little
reason to expect the pace of technological change to abate, so that wage dispersion
may well continue to increase. This will continue to raise the rates of return for
higher education, warranting the shift of more resources toward higher education on
the basis of efficiency considerations, which ultimately translate into productivity
considerations. To the extent that resource shifts in that direction are thwarted by
equity considerations, output and productivity growth will suffer.
The recommended solution to the preceding dilemma is for the government to
progressively pass on a greater share of the costs of higher education to its direct
beneficiaries, while facilitating the private absorption of those costs through student
loan programs designed to correct market failures in the financial sector.



18
Annex
Technical Note on Measuring the Rates of Return in Education
The rates of return on investmnents in education presented earlier are
derived according to a method outlined by Chiswick (1997). It assumes that the annual
earnings of a worker with schooling level s, denoted EB, are identically equal to the annual
earnings that he would have received with a year less of schooling, plus the cost of
investing in one extra year of schooling, C5, multiplied times the rate of return on that
investment, r8, or,
Es =Es5+ r8C, =E.,( 1 + r,C /E.,) =E.-1(+ r K+     )
=Eo BI (1 + rt K),
t-I
where K, represents the cost of investing in schooling level t relative to a full year's
potential earnings if investments were not made in that level of schooling. Taking the
natural logarithm of both sides of this equation, the relation between earnings and
schooling can then be stated as:
A.1)  Ln(E,) = Ln(E. ) + Z5Ln(l + rK).
The assumptions that r, and K, do not vary with years of schooling, together with the
approximation rule for logarithms, give rise to the following "Mincerian" specification
of the earnings function that can be estimated:
A.2)  Ln(E5) = Ln(E0) +  (rK)S.
The estimated equations based on this "Mincerian" specification, including a separate
variable representing years of labor market experience, are shown in Table Al. 16 The
corresponding rates of return derived from these estimated coefficients were presented in
Table 5 under the assumption that K is equal to 1. Also, since the rate of return is
assumed to be the same for all levels of schooling in this case, the marginal rate of return
from an additional year of schooling is the same as the average return.
16The inclusion of labor market experience, using a quadratic specification, has become a standard
procedure in the estimation of human capital earnings function ever since Jacob Mincer showed in
1974 that its exclusion yields biased estimates of the schooling variable coefficients.



19
Table Al: Estimated Mincerian Earnings Functions
Dependent variable = Ln(hourly income)
Wage Earners            Monetary Income Recipients
Independent     1984                1994               1984           1994
Variables
coeff  std.error   coeff  std.error  coeff  std.    coeff std.error
Constant      -0.437    0.042   -0.676   0.026    Error           -0.233   0.027
Experience    0.070    0.003    0.062   0.002    -0.307   0.041    0.060   0.001
ExperienceA2  -0.001    0.000   -0.001   0.000   0.065   0.002   -0.001   0.000
Schooling     0.152    0.003    0.167   0.002    -0.001   0.000    0.172   0.002
0.146   0.003
R-square            0.35             0.38              0.27             0.33
Adj R-sq.           0.35             0.38              0.27             0.33
No. observ.        4,864            12,991            7,555            22,269
Source: Estimated with data from ENIGH84 and ENIGH94
The "extended" version of the earnings function is derived by assuming
that rt and Kt vary across the four education categories stated in Table A2:
*  Category 1 refers to having completed the primary education level (which comprises
6 years) and may include some attendance at the secondary level,
*  Category 2 refers to having completed the secondary level (comprising an additional
3 years), and may include some attendance at the preparatory level,
*  Category 3 refers to having completed the preparatory level (3 years), and may
include some university attendance, and
*  Category  refers to having completed the university level (4 years) and any
additional post-graduate education.
Table A2: Estimated Extended Earnings Functions
(Dependent variable = Ln(hourly income)
Wage Earners         Monetary Income Recipients
Independent variables     1984             1994           1984            1994
coeff std.error  coeff std. error  coeff std.error  coeff std.error
Constant             -0.108   0.040  -0.214  0.026   0.038  0.039   -0.233   0.027
Experience            0.070   0.003   0.062  0.002   0.065   0.002   0.060   0.002
Experience^2         -0.001    0.000  -0.001  0.000  -0.001  0.000  -0.001   0.000
Dummy Variables
Primary complete      0.674   0.030   0.553   0.021   0.586   0.028  0.606   0.021
Secondary complete    1.088   0.036   1.043   0.023   1.016   0.036   1.104   0.024
Preparatory complete    1.644   0.046   1.704   0.027   1.546   0.047   1.779   0.030
University complete    1.979   0.057   2.362   0.032   1.946   0.057  2.453   0.035
R-square                  0.34             0.37           0.26            0.32
Adj. R-sq.                0.34             0.37           0.26            0.32
No. Observations          4864            13,991          7,555          22,269
Source: Estimated with data from ENIGH84 and ENIGH94



20
Using these schooling category distinctions, the extended earnings
function is derived from expression (A.2) as:
A.3) Ln(E8) = Ln(E0) + (r1 K,)S,
+ (r2 K2)(S, + S2)
+ (r3 K3)(S + S2 + S3)
+ (r4K4)(SI+S2+S3+ S4),
= Ln(E0) + b,D, + b2D2 + b3D3 + b4D4
where the b1 's refer to the coefficient estimates in Table A2 associated with each Dummy
variable, Di , and the Si 's refer to the number of years needed for moving from schooling
level i-I to level i. The estimation of this earnings functions using Dummy variables
(such that each individual is only associated with a single Dummy =1)", yields the
coefficients presented in Table A2. To derive the average rates of return from these.
coefficients, it is first necessary to deflate them by the corresponding S's and K's shown
in the equation above. [For example, r, = bl/ (KISI).]
Table A3 presents the cumulative number of years in school used for each
schooling category to derive the r1's according to the preceding formula. These figures do
not take into account repetition rates, which break the one-to-one correspondence
between years spent in school versus schooling level attained. Instead, Table A3 assumes
that persons who reported having completed a particular schooling level spent the
minimum time needed to attain that level, while persons who did not complete a certain
level were assigned a schooling time that is half-way between having completed the
previous level without repetition and completing the next level."8
Table A3: Assumed Average Years of Schooling
Wage Earners   Monet. Inc. Recip.
School Attainment Levels           1984    1994       1984     1994
0 - Less than Primary complete     2.21     2.07     2.21      1.93
1 - Primary and some Secondary     6.21     6.20     6.21      6.18
2 - Secondary and some Preparatory    9.18  9.14     9.18      9.29
3 - Preparatory and some University   12.76    12.78    12.85  12.80
4 - University complete and above  16.03    16.08    16.03    16.08
Note: It is assumed that completion of primary, secondary, preparatory and
university levels takes 6, 9, 12 and 16 years.
17 An altemative procedure would have been to assign dummy variables for each level of education
completed, such that some individuals will have a positive Dummy assigned to them more than once.
A person whose maximum attainment is the preparatory level, for example, would count as positive
in three Dummy variables (primary, secondary, preparatory) under that procedure, while under the
present procedure he only counts positive once, in D3.
Is The procedure was chosen for lack of conveniently available data for making the requisite adjustments,
in full recognition that it may result in important biases that overestimate the true rates of return (see
Behrman and Deolalikar, 1991). One extenuating circumstance in favor of this procedure is that the
focus of this report is mainly on the change in the rates of return of education over time, rather than
on their absolute value. Although this too will yield biases so long as repetition and drop-out rates do
not remain constant over time, the impact will not be as strong.



21
The rates of return (the ri 's) in equation (A.3) represent average rates of
return to education over the entire period that it takes to reach a certain schooling level.
The marginal rate of return (denoted as m.) obtained by moving from one level of
schooling to the next, however, can be easily derived by expressing the average rate as
the weighted sum of marginal rates. For example, the dummy coefficient for having at
least completed the secondary level, b2 = (r2 K2)(SI + S2) = (r1 K,)S, + (m2 K2 )S2 = b,
+ (m2 K2 )S2, so that the unadjusted marginal return for investing in secondary education
is m2 K2 = (b2 - bY)1S2. The private and social rates of return to education are derived
from these unadjusted marginal rates by dividing through with the appropriate Ki. The
appropriate K for each level i can be expressed as:
A.4)  Private K    = KP = (xE1., + C,P)/Ei-l
A.5)  Social K     = K,s = (xEi1 + CiP + Cg )/ E1. = KiP + (C1g / EiI).
In the above notation, the private cost of education is composed of C1P, which denotes an
individual's private out-of-pocket expenses associated with one year of schooling to
reach schooling level i, and of xEi1l, which denotes the level of earnings foregone by a
student with schooling level i-1 that is studying to reach schooling level i. The variable x
ranges between 0 and 1, and measures the extent to which a student is able to work part-
time while attending school (in inverse fashion). A standard convention in previous work
is that if a student has full-time employment, he does not forego any earnings and x = 0.
This convention, however, fails to recognize the private cost in terms of foregone leisure.
Finally, the social cost of education consists of the private costs plus Cig, which denotes
the government's expenses per year to advance one student to level i.
Private Rates of Return. Table 5 earlier presented the private marginal
rates of return derived under the assumption that the private K's (equation A.4) are equal
to 1 for all education levels. This assumption is based on two considerations: first, it
assumes that students on average value their leisure at the same rate as the wage foregone
by studying. That means that if a student works while going to school he is giving up an
amount of leisure that is worth the same as the amount of earnings given up by a student
who decides not to work while going to school. The second assumption is that out-of-
pocket private expenses of going to school are negligible, as suggested by Figure 3,
which shows that the overwhelming share of direct expenditures on education is
accounted for by the federal government. (This assumption is reviewed further below.)
Social Rates of Return. The social rates of return to education are
obtained by taking into account the total cost of the investment associated with moving
an individual from one education category to the next, independent of whether these costs
are borne by the individual or the government. As derived here, they do not contemplate
any externalities that might be associated with different levels of education. Table 6
reported the social marginal rates of return to education. They were derived by dividing
the private rates in Table 5 by the social K's shown in equation (A.5), again under the
assumption that the private K's are equal to one. (The public cost ratio, C,S/Ei.1, was
obtained for different schooling levels from the information contained in Tables 2 and 3.)



22
Note that as long as the the private individual does not bear all the costs involved with
education (i.e., Cig / Ei-l > 0), the social rates of return will by construction always be
lower than the private rates of return, which is borne out in a comparison of Tables 5 and
Table 6.
Estimated Private Costs of Schooling in 1994
The information in Figure 3 was taken from official reports that may have
underestimated the total private out-of-pocket costs associated with investments in
education. This Annex provides a separate estimate of private costs of schooling using
the income-expenditure survey of 1994. ENIGH94 provides information on direct
outlays associated with schooling by level (pre-primary, primary, secondary, preparatory
and university) in each family, together with spending on schooling articles that is not
broken down by level. In Annex Table A4, these last costs are divided up across different
schooling levels in proportion to the direct outlays. The resulting total of private
expenditures on education is then divided by an estimate of the number of students
attending at each level. According to these figures, private spending accounted for about
27 percent of total spending on education in 1994. These private costs of education are
still fairly modest in terms of average earnings, but turn out to be significant when
compared to the earnings of workers without experience, which may be a more
appropriate comparison for a student deciding whether to continue his studies to a certain
level or begin working without completing his degree.
Annex Table A4: Private Spending on Education in Mexico, 1994
Private Expenditures  As share of annual earnings of workers with prior level of
Per Student    schooling attainment:
(1994 Pesos)     Ave. 1994 Earnings    Workers w/o experience
__________  ______________ W&  S  Monet. Inc.     W&S        Monet. Inc.
Prinary              617.2          8.0%      7.4%         26.8%       29.8%
Secondary           1081.6          9.6%     8.2%         28.0%        27.6%
Preparatory         1625.6         11.5%    10.2%          27.1%       26.0%
University          3,369.4        13.1%     12.1%         31.0%       29.1%
average          1047.0          6.3%     6.2%
Total Students
in sample     Non-working (%/6)  Working (%/)  Ave. hrs/week
Primary              8131             97.6%            2.4%            43.0
Secondary            3374            91.3%             8.7%            31.3
Preparatory          1940             82.8%           17.2%            34.6
University           970              70.8%           29.2%            36.7
Total           14415           92.3%              7.7%           36.5
Source: Own calculations from ENIGH94. To calculate the spending shares, average 1994 Earnings are
from Table 2 and the earnings of workers without experience are calculated from the earnings function
coefficients in Table A2.
The lower half of Annex Table A4, shows the proportion of students that
work, at least part-time, while going to school. The overall low figure of 7.7 percent is



23
strongly influenced by the primary level average, which comprises more than half of all
students included in the survey. As we move up the schooling levels, part -time
employment becomes much more significant. At the highest level, an estimated 29.2
percent of all students attending university are working an average of 36.7 hours per
week. Although these figures are often used to calculate the total private opportunity cost
of going to school, this paper ignores that approach on the assumption that the earnings
from part-time work are offset by the loss of leisure of equal value.
Annex Table A5 presents the private and social rates of return for wage
earners after taking into account the private costs of education suggested above. Instead
of being equated to one, the 'private K' from equation (A.4) is now equal to 1 plus the
ratio of the private outlays per student from Table A5, divided by the average income of
persons with schooling levels one category below the one being aspired to by the student.
(Furthermore, it is assumed that the resulting ratio applies also in 1984.) The private rate
of return is then obtained by dividing the rates in Table 5 by this new private K, and
similarly, the social rates of return are derived by revising the values for the 'social K'
with these revised private K's. The results in Table A5 indicate that the basic conclusions
derived earlier about the tendency toward a reduced dispersion of social rates of return in
1994 do not change substantially with these revisions.
Annex Table A5: Revised Private and Public Rates of Return to Education
Based on average foregone   Based on foregone
Schooling level     earnings; Table 2         earnings of workers with
no experience; estimates
from Table A2.
=___________  1984    1994         1984        1994
Private Rates
1 -primary               15.6%        12.8%      13.3%        10.9%
2- secondary              12.7%       15.2%       10.9%       13.0%
3- preparatory            13.9%       16.3%       12.2%       14.3%
4- university              9.1%       17.6%        7.9%       15.2%
Social Rates
1- primary               14.4%        10.5%      10.8%         6.5%
2- secondary              11.3%       12.2%        8.1%        8.0%
3- preparatory            11.7%       12.2%        8.7%        8.7%
4- university              7.0%       12.1%        5.4%        8.1%
The Effects of Higher Wage Dispersion on the Returns to Education
Exogenous changes in the dispersion of wages alter the private rates of
return to education. In the absence of any policy responses, they will also change the
social rates of return in the same proportion. To see this, note from expressions (A.4) and
(A.5), that the social rate of return of investing in education level i can be expressed as a
function of the private rates of return:
rn; = r PKjP = r1SKYS = riS [Y(1P + (Cig/Ei- )].



24
After dividing through by K,P, this yields:
M/K"p  =  r,p  =  r1s [ 1 + (Cig/Ej, )/K,P]   ri [1 + GJi,
where G1 is a positive function of the amount of government spending per student in
level i of education. For simplicity's sake, assume that KJ, remains constant over time."9
Then by taking logs and differentiating, we obtain that,
A.6)         dn1/m, _  driP/ ri  =  d r1s/ r15  +   dGi /[l + GJ.
As observed earlier, over the interval from 1984 to 1994, the real wages of persons with
university education increased proportionately much more than the real wages of workers
with primary and secondary education. This means that the private rates of return for
higher levels of education would increase by more than for the lower levels; or using the
notation in expression (A.6), dr,P/ r"P > dr1.1P/ ri.,P. If the government were not to change
its resource allocation across the different education levels (i.e., dG; = 0), then the social
rates of return would also adjust by the same proportion and in the same direction as the
change in private rates of return.
Alternatively, assume that the government were applying a policy rule to
maintain the social rates of return constant across education levels (the efficient rule
discussed earlier). That would mean that dG; would have to adjust in the same proportion
as the change in private rates of return. In Mexico's case, the increase in the rate of return
to higher education would have to be met with an increase of G in higher education,
while the decline in the rate of return to primar, education would have to be met with a
reduction of G at the primary level. Only that way will the social rates of return be
maintained equal in the face of diverging movements in the private rates of return. Such
a policy response in G, however, also tends to exacerbate income inequalities, at least in
the short run, for the reasons discussed in Section E.
Annex Table A6:
Revealed Response to Change in Private Rates of Return, 1984-1994
Wages & Salaries            Monetary Income
Proportional change in:     proportional change in:
Schooling level                  private R   social R    G   private R   social R    G
1-primary complete               -18.3%    -27.1%    8.8%    -2.7%      -13.9%    11.2%
2 - secondary complete            20.1%       4.8%   15.3%    10.3%     -3.1%    13.4%
3-preparatorycomplete             17.4%       2.3%   15.1%    33.3%      18.0%   15.0%
4 - university and above          76.7%      73.7%    3.0%    60.9%      44.8%   16.1%
Source: Calculated from Tables 5 and Table 6.
Table A6 shows the actual response that took place in Mexico: even
though the private rates of return moved in divergent directions -- rates on primary
19 This assumption is less heroic than might appear, since it merely implies that the share of a full year's
income that would be foregone by continuing to study [i.e., variable x in expression (3.4)] remains
constant and that private out of pocket expenses move in line with earnings over time.



25
education declined, while rates on higher levels increased -- Government spending
increased at all levels, indicating that the "efficient" expenditure adjustment rule was not
followed. This apparent "egalitarian" response is consistent with the earlier observation
in Table 4, that per capita federal spending on different schooling levels had become
more equal in recent years. The unfortunate consequence of an egalitarian response is
that relatively fewer resources become invested in the education levels with the highest
rates of return, which eventually becomes translated into lower productivity growth.



26
REFERENCES
Behrman, Jere R. and Anil B. Deolalikar, 1991, "School Repetition, Dropouts, and the
Rates of Return to Schooling: The Case of Indonesia", Oxford Bulletin of
Economics and Statistics, 53, 4, pp. 467-80.
Beristain, Javier, 1991, "Tres Opciones de Politica de Atenci6n a la Demanda de
Educaci6n Superior", in Francisco Gil Diaz and Arturo M. Femrndez, editors, El
Efecto de la Regulaci6n en Algunos Sectores de la Economia Mexicana, Fondo de
Cultura Econ6mica, Mexico.
Bosworth, Barry, 1997, "Productivity Growth in Mexico", background paper for the
World Bank CEM on Mexico: Enhancing Factor Productivity Growth.
Chiswick, Barry R. 1997, "Interpreting the Coefficient of Schooling in the Human
Capital Earnings Function", Policy Research Working Paper No. 1790, The
World Bank.
Cragg, Michael and Mario Epelbaum, 1994, "The Premium for Skills: Evidence from
Mexico", mimeo, (November).
De la Torre, Rodolfo, 1997, "Desigualdad, Pobreza y Polarizaci6n Social en Mexico",
processed.
Elias, Victor J., 1992, Sources of Growth: A Study of Seven Latin American Economies,
International Center for Economic Growth.
Feenstra, Robert C. and Gordon H. Hanson, 1996, "Globalization, Outsourcing, and
Wage Inequality", AEA Papers and Proceedings, Vol. 86, No. 2 (May), pp. 240-
45.
Hernandez Laos, Enrique, Nora Garro Bordonaro and Ignacio Llamas Huitr6n, 1998,
"Productividad y Mercado de Trabajo en Mexico", background paper for the
World Bank CEM on Mexico: Enhancing Factor Productivity Growth.
Johnson, George E., 1997, "Changes in Earnings Inequality: The Role of Demand
Shifts", Journal of Economic Perspectives, Vol. 11, No. 2 (Spring), pp. 41-54.
Katz, Lawrence F. and Kevin M. Murphy, 1992, "Changes in Relative Wages, 1963-
1987: Supply and Demand Factors", Quarterly Journal of Economics, (February).
Londoflo, Juan Luis, 1996, "Poverty, Inequality and Human Capital Development in
Latin America, 1950-2025", World Bank Latin American and Caribbean Studies,
(June).
Organization for Economic Cooperation and Development (OECD), 1997, Education at a
Glance: OECD Indicators, Paris.



27
Panuco-Laguette, Humberto and Miguel Szekely, 1996, "Income Distribution and
Poverty in Mexico", in Victor Bulmer-Thomas, editor, The New Economic Model
in Latin America and its Impact on Income Distribution and Poverty, St. Martin's
Press, New York, pp. 185-222.
Psacharopoulos, George, 1994a, "Returns to Investment in Education: A Global Update",
World Development, Vol. 22, No. 9 (September).
Psacharopoulos, George and Y.C. Ng, 1994b, "Earnings and Education in Latin America:
Assessing Priorities for Schooling Investment", Education Economics, Vol. 2,
No. 2.
Psacharopoulos, George, Eduardo Velez, Alex Panagides and Hongyu Yang, 1996,
"Returns to Education During Economic Boom and Recession: Mexico 1984,
1989 and 1992", Education Economics, Vol. 4, No. 3.
Ram, Rati, 1990, "Educational Expansion and Schooling Inequality: International
Evidence and Some Implications", The Review of Economics and Statistics.
Robbins, Donald, 1994, "Earnings Dispersion in Chile After Trade Liberalization",
mimeo, Northeast Universities Development Conference, Economic Growth
Center, Yale University (October).
Schultz, T. Paul, 1988, "Education Investments and Returns", in H. Chenery and T.N.
Srinivasan, editors, Handbook of Development Economics, vol. I, Elsevier
Science Publishers B.V., pp. 543-629.
Journal of Economic Perspectives, 1997, "Symposium on Wage Inequality", Vol. 11,
No. 2 (Spring).
World Bank Economic Review, 1997, "Symposium Issue on How International
Exchange, Technology, and Institutions Affect Workers", Vol. 11, No. 1
(January).
Presidencia de la Republica, Estados Unidos Mexicanos, Primer Informe de Gobiemo,
1989.
Presidencia de la Republica, Estados Unidos Mexicanos, Tercer Informe de Gobierno,
1997.



Policy Research Working Paper Series
Contact
Title                            Author                  Date              for paper
WPS1925 Half a Century of Development     Jean Waelbroeck          May 1998           J. Sweeney
Economics: A Review Based on                                                31021
the Handbook of Development
Economics
WPS1926 Do Budgets Really Matter?          Emmanuel Ablo          June 1998           K. Rivera
Evidence from Public Spending    Ritva Reinikka                             34141
on Education and Health in Uganda
WPS1927 Revenue-productive Income Tax      Fareed M. A. Hassan     June 1998          A. Panton
Structures and Tax Reforms in                                               85433
Emerging Market Economies:
Evidence from Bulgaria
WPS1928 Combining Census and Survey Data  Jesko Hentschel          June 1998          P. Lanjouw
to Study Spatial Dimensions      Jean Olson Lanjouw                         34529
of Poverty                       Peter Lanjouw
Javier Poggi
WPS1929 A Database of World Infrastructure   David Canning         June 1998          A. Abuzid
Stocks, 1950-95                                                             33348
WPS1930 The Main Determinants of Inflation in liker Domac          June 1998          F. Lewis
Albania                          Carlos Elbrit                              82979
WPS1931 The Cost and Performance of Paid   Ariel Dinar             June 1998          F. Toppin
Agricultural Extenion Services: The   Gabriel Keynan                        30450
Case of Agricultural Technology
Transfer in Nicaragua
WPS1932 Air Pollution and Health Effects:  Bart D. Ostro           June 1998          C Bernardo
A Study of Respiratory Illness   Gunnar S. Eskeland                         31148
Among Children in Santiago, Chile   Tarhan Feyzioglu
Jose Miguel Sanchez
WPS1933 The 1997 Pension Reform in Mexico  Gloria Grandolini      June 1998           C. Zappala
Luis Cerda                                 87945
WPS1934 WTO Accession for Countries       Constantine Michalopoulos June 1998         L. Tabada
in Transition                                                               36896
WPS1935 Explaining the Increase in Inequality  Branko Milanovic   June 1998           G. Evans
during the Transition                                                       85734
WPS1936 Determinants of Transient and     Jyotsna Jalan            June 1998          P. Sader
Chronic Poverty: Evidence from   Martin Ravallion                           33902
Rural China
WPS1 937 Aid, the Incentive Regime, and    Craig Burnside          June 1998          E. Khine
Poverty Reduction                David Dollar                               37471



Policy Research Working Paper Series
Contact
Title                           Author                   Date             for paper
WPS1938 What Explains the Success         David Dollar            June 1998          E. Khine
or Failure of Structural Adjustment   Jakob Svensson                       37471
Programs?
WPS1939 Second Thoughts on Second         Arturo J. Galindo       June 1998          M. Cervantes
Moments: Panel Evidence on      William F. Maloney                         37794
Asset-Based Models of Currency
Crises
WPS1940 The Structure of Labor Markets in    William F. Maloney   June 1998          M. Cervantes
Developing Countries: Time Series                                          37794
Evidence on Competing Views
WPS1941 Are Labor Markets in Developing   William F. Maloney      June 1998          M. Cervantes
Countries Dualistic?                                                       37794
WPS1942 Poverty Correlates and Indicator-    Christiaan Grootaert  July 1998         G. Ochieng
Based Targeting in Eastern Europe  Jeanine Braithwaite                     31123
and the Former Soviet Union
WPS1943 The Implications of Hyperbolic    Maureen Cropper         July 1998          A. Maranon
Discounting for Project Evaluation    David Laibson                        39074
WPS1944 Detecting Price Links in the World   John Baffes          July 1998          J. Baffes
Cotton Market                                                              81880
WPS1 945 Evaluating a Targeted Social     Martin Ravallion        July 1998          P. Sader
Program When Placement Is       Quentin Wodon                              33902
Decentralized
WPS1946 Estonia: The Challenge of Financial  Carlos Cavalcanti    July 1998          L. Osborne
Integration                     Daniel Oks                                 38482
WPS1 947 Patterns of Economic Growth: Hills,  Lant Pritchett      July 1998          S. Fallon
Plateaus, Mountains, and Plains                                            38009
WPS1948 Comparative Advantage and the     Aart Kraay              July 1998          A. Kraay
Cross-Section of Business Cycles   Jaume Ventura                           35756