SWP395. ncome ld'istribution and Poverty in MezD(c( Wovid. Bank Staff Working Paper No June 1980

Transcription

1 Public Disclosure Authorized ncome ld'istribution and Poverty in MezD(c( SWP395 Public Disclosure Authorized Wovid. Bank Staff Working Paper No. 395 June 1980 Public Disclosure Authorized Public Disclosure Authorized I y: Joel Bergsman Country Programs Department I Latin America and the Caribbean Regiona[\Officp 1980 The World Bank.1818 H Street, N.W. Washington, D.C , U.S.A. The views and interpretations in this document are those of the author and should not be attributed to the World Bank, to its affiliated organizations, or to any individual acting in their behalf.

2 The views and interpretations in this document are those of the author and should not be attributed to the World Bank, to its affiliated organizations, or to any individual acting in their behalf. WORLD BANK Staff Working Paper No. 39 Joint Library Wa;hington, D. C July 1980 INCOME DISTRIBUTION AND POVERTY IN MEXICO The recent publication of results from a household budget survey for the first six months of 1977 sheds new light on income distribution in Mexico. This paper reports on analysis of four such surveys, for 1963, 1968, 1975, as well as 1977, with adjustments for consistency and comparability. According to the results, the distribution of income has not changed much since Thus, although incomes are distributed very unequally in Mexico and poverty is widespread, the absolute income levels of the poorer households have risen more or less apace with those of their richer countrymen, at an annual rate of about 4% per year in real terms. The paper also touches on three important sources of poverty in Mexico, (a) rapid population growth, which has resulted in a high dependency ratio and explosive growth of the labor force; (b) neglect of productivity in a considerable portion of agriculture and the people who live from it; and (c) policies that biased Mexico's economic growth in ways that reduced demand for labor. Analysis of these data continues, both in Mexico and at the World Bank. Special thanks are due to Gabriel Vera, without whose expert guidance and assistance in the analysis this paper could not have been written, and to CENIET and its Director, Geronimo Martinez, for giving us special tabulations of the 1975 survey. Helpful comments from Montek Ahluwalia, Carlos Bazdresch, Surjit Bhalla, Leticia Calzada, David Felix, Ruben Gleason, Salvador Kalifa, Mark Leiserson, Ricardo Moran, Alexander Nowicki, Teresa Rendon, Carmen Sanchez Cordoba, Marcelo Selowsky, Leopoldo Solis, and Richard Webb are also gratefully acknowledged. Thanks also to Ruth MacCrae and Isabelle Wieviorka for their highly competent research assistance. Prepared by: Joel Bergsman Country Programs Department I Latin America and the Caribbean Regional Office, Copyright The World Bank 1818 H Street, N.W. Washington, D.C , U.S.A.

3 Table of Contents Page I. Size Distribution of Income 2 II. Characteristics of the Poor 20 III. Some Causes of Poverty and Inequality 30 IV. Mexican Poverty in Perspective 41 References 45

4 List of Tables and Charts Table Page No. 1 Basic Data on Size Distribution of Household Income, 1963, 68, 75, and ***... * Estimates of Disposable Household Income.5 3 Earlier Adjustments for Under-Reporting.6,7 4 Estimates of Decile Distributions of Household Income 13 5 Adjustments for Under-Reporting 14 6 Summary Measures of Income Distribution and Poverty 15 7 Number of "Poor" Families by Sector and Type of Worker: Number of "Poor" Families by Sector and by Type of Worker * Sources of Income of "Poor" Families: Number of "Poor" Families by Education of Head of Family: 1975 * Number of "Poor" Families, by Occupation of Head of Family: 1975 ********* *... ** Average Expenditures of "Poor" Families, by Type of Expenditure: 1975 *...**.* Patterns of Land Tenure and Production, Distribution of Land, Machinery, Value of Holdings, and Value of Production: Changes in Agricultural Production.Per Farm, Chart 1 International Comparisons of Income Inequality... 16

5 INCOME DISTRIBUTION AND POVERTY IN MEXICO, Joel Bergsman Introduction Mexico's problems of rapid population growth, low productivity workers and poverty are serious, and there are no quick solutions to them. The relevant data, which come from many different sources, are not always in agreement, and some estimates are difficult to accept. Given more time and access to the original data, it should be possible to put together a more accurate and complete pizture of trends in productivity and income distribution. in this task. Researchers at the World Bank and in Mexico are now engaged But even without the results of such work, the broad outlines of great inequality in productivity and in incomes, and little or no change in distribution over the last 15 years, are unmistakable.

6 I. SIZE DISTRIBUTION OF INCOME Data and Sources The main sources of data are four generally comparable household budget surveys-,-done in , 1968,_1975 and These surveys attempted to measure total household income after direct taxes, including income in kind, imputed income from owner-occupied dwellings, income from capital, and transfers, as well as normal monetary income. They are all national sample surveys. Although the concepts, methods, and reliability vary somewhat from one survey to another, they are much more comparable than sets of such surveys in most other LDCs, and at least three of the four appear to be at least as reliable. Moreover, they give direct estimates of the variables that are central to our concern, household income (rather than individual income), and they estimate total after-tax household income, including income in kind and various imputed components such as rent from owner-occupied dwellings. The reported estimates of the size distribution of income and expenditures are shown in Table 1. A comparison between the national total of disposable personal income (DPI) reported in the studies, and the same variable estimated from national income accounts data, is shown in Table 2. Among the interesting charateristics of these data are the following: (i) At lower income levels, reported expenditure is often much greater than reported income. In many cases the differences seem too large to be accepted as true and due to dis-saving by families whose income has fallen below their "permanent" income level.

7 -3- (ii) The national totals for DPI implied by the household budget studies are considerably less than that implied by national accounts data. Moreover, the difference increases with time--i.e. is smaller in 1963 and 1968 than in 1975 and These characteristics suggest that income may be underestimated by the studies, and that the degree of underestimation increased in the two more recent studies. If this is so, then neither the situation as of one year nor trends during the period are accurately reflected by the reported data. Under-reporting of income seems to occur in almost all household budget studies in all countries. It is generally thought to be mainly a reflection of two problems: income in kind and informal transfers are thought to be accidentally underestimated, and informants are thought to deliberately under-report. The first problem is presumably concentrated in the lower portion of the income spectrum; the second may well occur among informants at all income levels although many scholars believe it to be concentrated (in amount if not in percentage of informants) among those with higher income. Three different sets of adjustments to the reported data for various years are known to the author. Their results are shown in Table 3. Navarrete worked with data for 1950, 1958 and (Earlier work by the same author is not discussed here, on the assumption that her later publication supercedes the earlier one.) Felix made estimates going back to the nineteenth century, relied on Navarrete for 1950 and 1957, and made his own adjustments for 1963, 1968, and 1975; only the three latest years are discussed here. Altimir made adjustments for 1963 and 1968.

8 Table 1: BASIC DATA ON SIZE DISTRIBUTION OF HOUSEHOLD INCOME: 1963, 68, 75, and Number of Total Reported Number of T.t-l Rj1p-t-d Number of Total Reported Number of Total Reported Households Income Expenditure Households Income Expenditure Households Income Expenditure Households Income (millions) (million pesos/month) (millions) (million pesos/month) (millions) (million pesos/month) (millions) (million pesos/month) , , ,231 1, ,141 1, , , ,319 1, , ,309 2, , ,830 1, , ,218 1, ,035 1, ,120 2, , ,339 1, ,742 2, , ,459 1, ,697 3, , ,636 1, ,766 5, ,895 9, ,514 1, ,107 4, , ,367 9, ,321 1, , ,736 nl _ ,321 14, ,284 32, ,895 Sources: Encuesta Sobre Ingresos v Gastos Familiares en Mexico , Banco de Mexico. Estudio de Ingresos y Gastos de las Familias (1968), Secretaria de la Presidencia, Direccion General Coordinadora de la Programacion Economica y Social, July Special tabulation from Encuesta de Ingresos v Gastos Familiares (1975), Secretaria del Trabajo y Prevision Social, Centro Nacional de Informacion y Estadisticas del Trabajo, Special tabulation from the Encuesta Nacional de Ingresos y Gastos de los Hogares,1977, first semester, Direccion General de Estadistica, Secretaria de Programacion y Presupuesto, as reported by Hernandez and Cordoba (1979), Table 1.

9 Table 2: ESTIMATES OF DISPOSABLE HOUSEHOLD INCOME A. From NLational Income Accounts Wages, salaries, social security. ~~~~~~~~(billion pesos per year) payments a/ 64.o Transfers from abroad Mixed Income Rent and Interest Distributed Profits less: Direct taxes TOTAL B. From Household Budget Studies Ratio of "B" to "A" Source: Part A Unofficial estimates, Banco de Mexico a/ Includes net factor payments from abroad

10 Table 3: EARLIER ADJUSTMENTS FOR UNDER-REPORTING NAVARRETE (1970); average monthly incomes in current prices Percentiles Reported Adjusted Ratio Percentiles Reported Adjusted Ratio Percentiles Reported Adjusted Ratio Total Elasticity of adjusment I FELIX (1979); average annual incomes in 1950 prices Total Elasticity of adjustment continued,,,

11 Table 3: Continued I ALTIMIR (1979); average monthly income in current prices Percentiles Total Elasticity of adjustment : Author's calculations based on Tables 5, 6 & 7 of Navarrete (70), Table A-VI of Felix (79), and Tables 1 and 14 of Altimir (79).

12 -8- All three scholars relied on estimates from the national accounts for the total amount of household income in each year (Altimir not completely, as explained below), and hence for estimates of changes over time in the totals. Even here, however, they differ to some extent because disposable household income is not reported in the national income accounts and was estimated differently by each author. But their major differences lie in how the shortfalls were distributed across income classes. Navarrete made a two-step adjustment. (See Navarrete, 1970, and Altimir, 1979, for full explanations.) She first increased the reported income of those classes that reported net dis-saving-by the class as a whole, to take into account presumed under-reporting of income in kind, gifts, transfers, and sporadic income. She then increased the reported income of those classes that reported net saving. The total of this second increase was such as to equate the total over all households with her estimate from the national income accounts, and it was distributed proportionally to the average income of each class so adjusted. As Felix and Altimir have pointed out, this distribution is not only arbitrary in concept, but also depends on the structure of disaggregation in the reported data. Given the way in which the results were originally reported, the adjustment distributes very large shares of the total discrepancy to the highest income classes because, since they covered only a small part of the very top of the distribution, they had very high average incomes. Felix also made a two-step adjustment, conceptually similar to Navarrete's. First he raised the income of the lowest 20% to 90% of its reported consumption and that of the second 20% to 92% of its reported

13 -9- consumption. The second step was to allocate the remaining gap between the (partly adjusted) survey totals and the national income accounts totals to the upper 60% of all households. Felix followed Navarrete in distributing the total proportionately to the average income of each income bracket. Felix notes that this adjustment is arbitrary, and that it implies an income elasticity of under-reporting greater than one (because the number of households in each bracket decreases as the average income increases).l/ Altimir takes a different approach. After detailed and careful analysis of what the various causes of the gaps may be, and of how Navarrete and Felix approached the problem, he concludes that the best method is to adjust separately each component of income: income, income from capital, and transfers. wages and salaries, business But because he did not have access to all the necessary data, he disaggregates families by socio-economic class and income size class. The correction factor for each group is a weighted average of ratios of income as estimated by the national accounts to income as estimated in the survey, and the weights are the incomes of the given type for that group, as estimated in the survey. For income from property, the discrepancy was distributed only to the top 20% of all households. For income from wages and salaries, the survey estimates exceeded the national account estimates and no adjustment was made. For other types of incomes, the discrepancy was distributed proportionally to income as estimated by the surveys. In the judgment of the present author, the methods of Navarrete and Felix (which conceptually are the same and in practice are rather similar) are not acceptable. Their more serious weakness is the totally 1/ "Income elasticity of under-reporting" is the slope of a log-linear regression of adjusted income on reported income.

14 arbitrary nature of the adjustment at the upper end. This adjustment depends not only on arbitrary assumptions about under-reporting (which are unavoidable) but also on the particular grouping by income size class (which is not unavoidable and which, as used, assigns almost all the discrepancy to the highest classes). Another aspect that seems hard to justify is the assumption that only low-income and high-income households under-report, while middle-income households report accurately and honestly. I see no reason to assume that middle-income families are especially accurate in their estimates of their income or especially honest in their reporting of it. Comparisons of reported income and reported consumption in the micro data, reported by Altimir, suggest under-reporting at all income levels. Altimir's logic and method seem far superior; however, his results seem dubious in two respects: for the lowest decile in 1968 his adjustments seem to have reduced the average income,l/ and in general the overall income elasticity of under-reporting implied by his results is rather high, especially in 1968 (see Table 3). In deciding how to deal with the adjustment question, I was guided by the following ideas: (i) Some adjustment seems to be required. Inter-year comparisons based on unadjusted survey results would be strongly biased (showing more of a downward trend in incomes than is likely to have occurred); even in one year it seems clear that the reported survey results miss a lot of income. l/ This result is, I think, logically impossible. It may be caused by a mistake in Altimir's calculations, in the typing of his paper, or in my calculations based on his work.

15 (ii) In the absence of new information about under-reporting, any adjustment is arbitrary. Navarrete, Felix and Altimir agree. (iii) I have access only to the published results, plus a few special tabulations for Hence I could not hope to do as well as Altimir has done for 1963 and 1968, nor anything as good for 1975 and In this quandary I have opted for simplicity and conservatism. I do not even attempt to estimate "the" distributions, but rather make two sets of estimates for each year, one tending to overestimate inequality and the other tending to underestimate it, with the idea that the truth is probably somewhere in between. Each estimate adjusts total income to the level implied by the national accounts; the "low inequality" estimate, called method 1, assumes an income elasticity of under-reporting equal to 0.95 while the "high inequality" estimate, called method 2, uses a value of The two elasticities more or less span the range of those implied in the Navarrete, Felix, and Altimir adjustments. (In method 2, straightforward application of the elasticity of 1.20 results in reducing the average incomes of the lower-income groups; in such cases the added restriction that adjusted income be at least as great as reported income for each class was applied; therefore, in the final results the overall elasticity of under-reporting turns out to be less than 1.20.) In my estimates, reported in Table 4, the first step was to estimate decile distributions from the reported data for 1963, 1968, and (For 1977 the decile distribution was reported.) This was done by interpolating log-linearly, for each decile point on the cumulative distributions,

16 between the two reported points on either side of the decile point. (For 1963, the first reported point is above the first decile, and hence a rough approximation was made by freehand drawing of a graph.) The piece-wise log-linear interpolation seems to match the shape of the theoretical Pareto curve fairly well, while also being simple to do and giving minimal divergence from actual observations.-' (We had two different groupings from the 1975 survey. We used log-linear interpolation from one to estimate the other, and the results were very close to the actuals.) The second step was to apply the two different assumed elasticities of under-reporting to the estimated decile distributions. The results are presented in Tables 4, 5, and 6. The results of the 1975 survey--both before and after adjustments-- seem strange. They imply sharp drops in the shares of the lower income classes, even in their absolute income levels (see Tables 4 and 6). All three measures of inequality jump sharply from the 1968 levels. Then, the 1977 results imply a rapid restoration to a distribution similar to that of Under adjustment method 2 the real income of the lowest 40% drops 15% from the 1968 level, then rises 49% in the two years to The share of the lowest 20%, a less reliable measure, shows even more drastic and unlikely behavior under both adjustment methods, dropping between 27% and 46% from 1968 to 1975 and then rising between 82% and 97% from 1975 to These 1/ Note that in Kalifa's estimates, reported in Kalifa (1977) and in Hernandez and Cordoba (1979) a linear interpolation was used. Not only does this method give a less smooth Lorenz curve, in a few cases it results in identical estimates for average household income in two adjacent deciles, which is nonsense.

17 Table 4: ESTIMATES OF DECILE DISTRIBUTIONS OF HOUSEHOLD INODME Percent of income Cumulative percent of income Decile A. Unadjusted ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~a. a/ 1 1.3T/ B. Adjustment method l Income elasticity of under-reporting = C. Adjustment method 2: Income elasticity of under-reporting = Source: Author's estimates from household budget surveys.- Details may nof add to totals because of rounding. a/ Approximate

18 Table 5: ADJUSTMENTS FOR UNDER-REPORTING (average monthly incomes in current prices) Adjustment Method 1 Adjustment Method 2 ted [ Adtustment Method 1 Adjustment Method 2 Decile Reported Adjusted Ratio Adjusted Ratio Reor Adjusted Ratio Adjusted Ratio , , ,272 1, , ,107 1, , ,734 2, , ,592 1, , ,476 3, , ,923 4, , ,455 7, , ,367 11, , ,321 19, , Elasticity of adjustment , , ,280 2, , , , ,870 3, , ,218 2, , ,559 4, , ,674 3, , ,318 6, , ,180 4, , ,140 7, , ,816 5, , ,275 9, , ,814 6, , ,936 12, , ,615 10, , ,895 17, , ,498 24, , ,996 36, , ,284 59, , , , , Elasticity of adjustment Source: Author's estimated from household budget surveys.

19 Table 6: SUMMARY MEASURES OF INCOME DISTRIBUTION AND POVERTY A. Unadjusted B. Adjustment Method 1: C. Adjustment Method 2: Income Elasticity of Under-Reporting = 0.95 Income Elasticity of Under-Reporting = 1.20 Percent of Disposable Household Income Received by: Lowest 40 percent 10.3% 10.5% 8.1% 10.9% 11.2% 11.4% 8.9% 11.9% 8.4% 8.4% 5.4% 7.9% percent Highest 20 percent Measures of Inequaliry': Share of Highest 20 percent /share of lowest 20 percent Gini index a/ Theil index a/ '.582 Households with Real Incomes Less Than 1977 Minimum Wage: Number of households (millions) Percent of all households Average Real Income of Lowest 40 $565 $742 $523 $757 $753 $1009 $1034 $1426 $565 $742 $633 $944 percent (1977 dollars per year per household) Source: Author's estimates from household budget surveys. a/ calculated from decile distributions.

20 SHARE OF LOWEST 40 PERCENT ~~~~~~0 0 TAIWAN SRI LANKA. I YUGOSLAVIA INDIA KOREA.4 SPAIN ARGENTINA. CHILE COSTA RICA PHILIPPINES MEXICO PREDICTED TURKEY MALAYSIA VENEZUELA HONDURAS PERU i._._ 5_._I_._! BRAZI L., m m m 0~ z > 0 >I a -4 N) > m r _ Z I ' ~~~~m -n r :c RATIO OF SHARES, ; C, -I HIGHEST 20 PERCENT/LOWEST 20 PERCENT C _ co aei7u-~n.ny - r7nf YE..~~~~~~~~~~~~~~~~~ Z.Z TAIWAN ~0 SRILANKA YUGOSLAVIA SPAIN INDIA KOREA ARGENTINA CHILE PHILIPPINES MEXICO PREDICTED COSTA RICA 0. co TURKEY MALAYSIA VENEZUELA HONDURAS PERU BRAZI L I 0 m>, M Oo 0 c~~~~~~~~~ -I >o.p. Ir- 0

21 - 17 reported changes may in part reflect the truth; 1975 was a year of high inflation and a slight drop (0.7%) in crop production. The shares of low income households may well have dropped, and even their absolute incomes may have fallen. But the magnitude of the reported changes is very hard to accept. The 1975 survey was somewhat lower quality than those of 1968 and 1977, in terms of sample size and training of enumerators, and there was also some greater imprecision in the definition of the concepts of income. For all these reasons, we conclude that the drastic changes implied by the 1975 results were probably in small part actual but short-run, and in large part due to errors in the survey. Therefore, most of our subsequent analysis is based on the (adjusted) results of the other three surveys. Inequality Income in Mexico is distributed very unequally. The share of the poorest 40% of households is between 8 and 12% of the total. This is probably lower than the 11.5% "expected" on the basis of cross-country patterns (see Chart 1).!/ It is lower than in at least 9 (or at most 13) of 16 other less-developed countries for which data are presented in the World Development Report for 1979; only in Honduras, Peru, and Brazil is the share clearly lower than in Mexico.!/ The ratio of the share of the highest 20% to the share of the lowest 20% is similarly higher than in most other countries. Changes in inequality in Mexico since 1963 have been small or nil. Disregarding the questionable data for 1975, the three different measures of inequality shown in Table 6 show no significant trend or changes when considered together. The decile structure has changed but little, although 1/ "Expected" values calculated by the author from results reported in Ahluwalia (1976), Table 3. 2/ More recent estimates for Brazil suggest slightly less inequality there, but still more than the present estimates for Mexico. I

22 there has probably been some increase in the share of middle income groups at the expense of the upper and lower income groups (see Table 4). This conclusion about no significant change in the d1istribution does not appear to be subject to the problem of under-reporting and how to distribute the unreported income (discussed earlier). There is no significant change in distribution in any of our three sets of estimates-- the unadjusted and the two differently adjusted ones. Thus, the conclusion stands regardless of *how the unreported income was distributed, as long as that distribution was more or less similar in each of the three surveys. This last assumption seems very likely to be true; whatever the reasons for under-reporting were, there is no reason to think that they changed from 1963 to Poverty Many Mexicans still live in dire poverty, as anyone who knows the country can attest. However, the rise in per capita incomes in the nation as a whole, combined with the more or less unchanged share of the lowest 40% increased the real income of that stratum considerably--perhaps 70 or 80% from 1963 to 1977, an annual rate of around 4% per year (see Table 6). This estimate of 4% per year is considerably higher than the growth in GDP per capita, estimated by the Banco de Mexico to have risen at 2.7% per year over the same period. The 4% is derived from unofficial estimates by Banco de Mexico staff of disposable personal income at current prices, deflated by the consumer price index since 1968 and a linked wholesale price index from 1963 to 1968, and divided by World Bank staff estimates of population. The difference between the 4 and the 2.7 is due to two

23 factors: (a) the GDP deflator rose more rapidly than the consumer price index and (b) World Bank population estimates are lower than official Mexican estimates for years since 1970 (the last census), because the official Mexican figures apparently make no allowance for emigration. The difference in price deflators accounts for about four-fifths of the total discrepancy and the difference in population estimates accounts for about one-fifth. The greater rise in the GDP deflator is reasonable; many consumer prices were controlled in 1977 and hence probably rose much less than the general price level did in the mid-1970's; moreover the import substitution in intermediate goods probably meant that prices of all goods rose more than prices of consumer goods over the last 15 or 20 years. However, if the consumer price index errs, it is probably in the direction of under-estimation of inflation; moreover, the controlled prices of 1977 conceal inflationary pressures that will probably make themselves felt in the future. Therefore, the estimates of real income growth presented here may err on the high side. Taking the 1977 minimum wage as a poverty line, the percentage and even the absolute number of households whose incomes fell below the line have been decreasing (see table 6). As of 1977, only 20 to 30% of all households earned less than the minimum wage. The World Bank conventionally considers an income equal to onethird the national average as a "relative poverty" line. It turns out that such an income, in 1977, was close to the minimum wage. The percentage of households with income less than one-third the national average was between 23 and 34% in 1977.

24 II. CHARACTERISTICS OF THE POOR Unfortunately, disaggregated information relating to total disposable household income from the 1977 survey only became available after this paper was substantially completed; analyzing the data will take some additional time. We did have some descriptions based on the 1975 survey, and present them here. Even though the 1975 survey seems less accurate than the others, we may learn something from what it tells us about poor families. In 1975 the largest number of "poor" Mexican families--52%-were in the agricultural sector (see Table 7).l/ Of these, 33% are listed as self-employed--presumably ejidatarios and other small proprietors--while 18.5% are listed as salaried--presumably landless agricultural workers. Thus the most important single group of the poor are land-owning peasants million families--and the second most important are the even less fortunate landless rural workers--850,000 families. These 2.4 million families are the core of Mexican poverty. They include 76% of all families in agriculture in Mexico. 1/ "Poor" families were defined in this source as those households with incomes less than half of the estimated national mean-mex$1,621 per month, equivalent in purchasing power to approximately US$1,315 per year per family. This cut-off point classified some 4.6 million Mexican families--45% of total--as "Poor". This is a rather larger number than we would choose; both the higher cut-off criterion and the apparently exaggerated extent of poverty in the 1975 survey contribute to the difference. Characteristics such as sector, occupation, and education refer to the head of the household.

25 Table 7: NUMER OF "POOR" FAMILIES BY SECTOR AND TYPE OF WORKER: 1975 Sector and Type Percentage of Average Income of Worker Thousand Families Total Families (dollars per year) Total Urban Rural Total Urban Rural Total Urban Rural No Occupation Agriculture 2,410 X 2, Owner Self-employed 1, , o Non-salaried in family business Salaried employee Mining , ,152 Owner Self-employed Salaried employee Petroleum and Electricity ,077 1,077 Owner _ Self-employed I _ Salaried employee Manufacturing Owner Self-employed Salaried employee Construction Owner Self-employed o.4 Salaried employee Commerce, Services, Transport Owner Self-employed Salaried employee Government g , Owner Self-employed Salaried employee Insufficient Information Total ,121 3, Unemployed Owner o.4 o.8 Self-employed 2, , Non-salaried in family business Salaried employee 1, , Source: Author's estimates from household budget survey data. a/ Defined as families whose monthly income was less than 1,621 pesos. L

26 The sectoral structure of Mexican poverty has not changed much since The percentage distribution by sector of "poor" families in 1963 was virtually identical to that in 1975; in 1963 agriculture accounted for 54.5%. In 1963 there were proportionally more landless workers and fewer of those who owned land (see Table 8). If these data are correct, they reflect a failure of Mexico's land redistributions during the period to improve the incomes of its beneficiaries. Families whose head is not working account for only 12% of the "poor"--about 5% of all families--and their average incomes, while low, are one-fourth again as much as those of the poor employed in agriculture. This confirms the often-repeated idea that the poorest Mexicans cannot afford to be unemployed. Other important sectoral-occupational concentrations of the poor are both self-employed and salaried workers in manufacturing and construction. Tables 9 through 11 give more details on characteristics of poor families in Mexico, as estimated by the 1975 survey. Among the urban poor about two-thirds of income is from wages and salaries, and about 20% from their own business; for the rural poor self-employment is just as important as wages and salaries; each accounts for about 45% of income (Table 9). Heads of poor families have very little formal education, especially in rural areas where 55% had no school at all and another 30% more had less than 4 years. Overall, more than three-quarters of the heads of poor families had less than 4 years of schooling (Table 10).1/ 1/ These percentages are probably over-estimates, because some households without any adult male name the oldest male as head of household. Thus, for example, a 2 year old boy, obviously without any schooling, could be named as head of household.

27 Table 8: NUMBER OF "POOR"-/ FAMILIES BY SECTOR AND BY TYPE OF WORKER Type of Activity of Percentage of Head of Family J Tho.usan4dsof Families Total Families Total Urban Rural Total Urban Rural Agriculture , Owner Self-employed 1, Admin. or Tech. Employee o.6 Worker 1, ,286.o Mining Owner _ Self-employed Admin. or Tech. Employee - _ Worker Industrial Manufacturing Owner _ Self-employed o Admin. or Tech. Employee o Worker Construction Owner Self-employed Admin. or Tech. Employee _ Worker Electricity, Water, etc. 16.o Owner _ Self-employed o Admin. or Tech. Employee Worker Commerce lL Owner Self-employed Admin. or Tech. Employee o Worker o Transport and Communications Owner Self-employed Admin. or Tech. Employee Worker Services Owner _ Self-employed Admin. or Tech. Employee o Worker Total 4, , Owner Self-employed 1, , Admin. or Tech. Employee Worker 2, , l Defined as families whose monthly income was less than 1,000 pesos. This cut-off level is not exactly comparable to the one used for 1975; data availability does not permit estimation of these results with the same cut-off level for the two years. Source:!Author's estimates from household-budget survey data.

28 _TableAt SOURCES OF INCOME OF "POOR" FAMILIES: 1975 Family Income (dollars per year) Percentage of Total Family Income Total Urban Rural Total Urban Rural Income from Work Wages and salaries Own business, nonagricultural Own business, agricultural Transfers Buying and Selling of Assets, and Other Total Source: Author's estimates from household budget studies. add to totals because of estimating.procedures.- Details may not

29 Table 10, NUMBER OF "POOR" FAMILIES BY EDUCATION OF HEAD OF FAMILY: 1975 Percentage of Average Income Years of Schooling Thousand Families Total Families (dollars per year) Total Urban Rural Total Urban Rural Total Urban Rural- No education 2, , , , , and more o.4 o.4 _ Not specified 5 5 o Total 4,599 1,121 3, Source: Author's estimatet from household budget studies. a/ The average incomes for rural families headed by persons with considerable education are subject to possibly large sampling area.

30 Over half of all poor families are headed by agricultural workers. 15% are in industries or handicrafts, only 10% are service workers, and 12% describe themselves as without occupation. Thus, according to the 1975 survey, the service sector does not account for a large share of poverty in Mexico; even among the urban poor only about one-fourth are in services (Table 11). Studies based on the 1968 and the 1975 surveys have identified education of the head of household as the variable most closely associated with income differences. For 1968, van Ginnekin (n.d.) performed a decomposition analysis of the Theil index of inequality. He found that of five possible variables, the one accounting for the largest share of observed inequality in incomes was inequality in education. Urban/rural location was second in importance, occupation and sector were about tied for third place, and age was a distant last. For 1975, Vargas and Vera (n.d.) used a very different technique, the "automatic interaction detector," and also found education the dominant variable associated with income differences. Urban-rural location and number of workers per household were also important; neither age, sex, sector, or occupational status of the head of household was significant. The statistical importance of the education variable is common to such data in most countries; it does not tell us to what degree more school-." ing is a cause of higher income, or simply a concommitant. More interesting is the lack of importance of the age variable; this is consistent with Felix's (1979) conclusions that mobility of workers is low in Mexico, and that therefore comparisons of different point-in-time distributions of

31 Tab6e 11: NUMBER OF "POOR" FAMILIES, BY OCCUPATION OF HEAD OF FAMILY: 1975 I. Percentage of Average Income Thousand Families Total Families (do4lars per year) Total Urban Rural Total Urban Rural Total Urban Rural No occupation Professional and technical o Managers o , White dollar workers ,083 Tradesmen and salesmen Service workers go9 760 Agricultural workers 2, , Industrial and handicraft workers Not classified Total 4,598 1,122 3, Source:'Author's estimates from household budget studies.

32 income may not seriously understate actual increases in income of the particular families whose income was low in the initial year. Finally, the relatively small effects of both sector and occupation reflect the duality-- or better, the dispersion of incomes within broad occupational and sectoral categories. There are rich farmers and poor farmers, middle-income salaried workers and poor salaried workers, etc. Table 12 shows the expenditures of poor families. Not surprisingly, about half are on food and the lion's share of the rest are on housing and clothing. However, all other items than these three account for about 20% of expenditures; not a very small amount for what are largely discretionary expenditures.

33 Table 12: AVERAGE EXPENDITURES OF "POOR" FAMILIES, BY TYPE OF EXPENDITURE: 1975 Expenditures- (dollars per year) Percentage Distribution (percent) Total Urban Rural Total Urban Rural Food Beverages Tobacco o Housing Domestic Services Clothing, Transportation Education o Medical Services Other Services Furniture Home Appliances o Vehicles Other Expenditures o Total 1,255 1,808 l,o84 1oo O Source: Author's estimates from-household-budget studies.

34 III. SOME CAUSES OF POVERTY AND INEQUALITY Even though a fully satisfactory analysis of the causes of the situation just described is beyond our reach, we want to explore some of the more obvious factors that lie behind Mexican poverty and inequality. These are to be found in three related processes: (a) rapid population growth, especially since about 1950, with a resulting high dependency ratio and a rapid growth of the labor force; (b) neglect of the productive capacity of many agricultural laborers and owners of poor land, (including many ejiditarios) who now comprise over 2 million of Mexico's 11 million families and some three-fourths of all Mexicans engaged in agriculture; (c) policies that made capital equipment cheaper and labor more expensive than would otherwise have been the case. Rapid Population Growth 1/ Since 1950, Mexico has experienced remarkable demographic growth. Life expectancy increased from about 51 years in 1950 to 64 years in The decline in mortality was not offset by any fall in fertility until the mid-1970s, and consequently the population grew by about 3.4X per year during the 1960s and about 3.3% in the early 1970s. Although crude birth rates and total fertility rates remained high through the mid-1970s, recently released data indicate that the crude birth rate has declined from 45 per thousand in 1970 to about 40.6 per thousand by mid-1976,2/ and may have declined still further since then. 1/ This section is mostly taken from Pierce (1978). 2/ Estimates by El Colegio de Mexico.

35 The decline in mortality and the persistence of high fertility through the mid-1970s produced a markedly young population. Between 1950 and 1970, the median age declined by almost two years to 16.6 years; by 1970 about 47% of the population was below 15 years of age. These changes in age structure have brought about a sharp increase in the dependent population. The age dependency ratio rose from 85.6 in 1950 to in The large proportion of the population under age 15 poses serious problems for the country's capacity to educate these young people and to absorb them into the labor force. The population of working age (12 years and over) expanded at a much slower pace than total population in the period , but somewhat more rapidly than it since 1970, because the large cohorts born during the earlier decades are now reaching working age. The labor force itself grew more slowly than the population of working age (and hence even more slowly than total population) from 1950 to 1969, as participation rates dropped sharply during these two decades. The labor force expanded at an annual rate of 2.0% from 1950 to 1960, 2.7% from 1960 to 1969, and currently seems to be expanding at a rate of over 3% even after deducting rough "guesstimates" for emigration. Rural Poverty Land reform in Mexico dates from the revolution of 1910, but proceeded fastest in the Cardenas period ( ), when massive land redistributions were made, input and credit needs of ejidatarios were recognized, and ejidos were organized. After Cardenas, land distribution continued at a reduced pace, and most of its beneficiaries were given

36 little in the way of credit and other inputs. Rather, emphasis shifted to large-scale irrigated farming, and to industrialization. From 1942 to 1969, government investment in the rural sector was concentrated on roads and on regional river basin and other irrigation programs which focused on building large dams and other infrastructure. Agricultural research and extension were devoted (very successfully) to large-scale irrigated agriculture, while better seeds and techniques for rainfed agriculture were never attempted. Nevertheless, from 1940 through the mid-1960s the demand for labor in agriculture grew fairly rapidly. Using a technical labor/output coefficient approach, Rendon (1976) estimates that the demand for labor in agriculture grew 3.1% a year during This growth was due to two basic trends: a massive increase in irrigation, both on new land and on already cultivated land, increased both the need for labor per acre and the number of acres under cultivation, and land redistribution also resulted both in bringing previously fallow acreage under cultivation and in greater labor use per acre. A rate of increase of labor requirements of 3.1% per year would have been about equal to the rate of natural increase of the rural population (for the nation as a whole, population grew at 2.9% per year over the period); meanwhile employment in non-agriculture activities was increasing at 4.8% per year, providing considerable employment for emigrants from rural areas. So while there was almost certainly some rural under-employment in 1940, and many peasants had little or no land, the pressures of population on the land at least did not increase in the period.