Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas

Transcription

1 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas Mark Klee 12/11/06

2 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 2 1 Introduction Goals Examine response of human capital investment to uctuations in family income What are effects of nancial market constraints in developing countries? Estimate structural model of human capital investment to quantify cost of using child labor as self-insurance Distinguish effects of anticipated, unanticipated income shocks

3 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 3 Results Child labor plays signicant self-insurance role for poor families in rural India, though not costly in the long-run Reject intra-village complete markets Only small-farm households are inadequately insured ex ante

4 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 4 1. Introduction 2. Theoretical Framework 3. Data 4. Estimation Strategy 5. Results 6. Conclusion

5 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 5 2 Theoretical Framework Value of Education Average grade attainment in 6 ICRISAT villages rose steadily by cohort for both sexes, suggests education valuable Education as route to more lucrative employment outside agriculture Education to enhance farm management ability Estimate conditional farm prot function with household xed effects, with education of head shifting prot level and interacting with use of high-yielding seed varieties Signicantly positive returns to education increasing with fraction of land under HYVs

6 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 6 Model Child from household i is eligible for school for 0 t T consecutive agricultural seasons S it, school attendance given cumulative history of shocks at time t, augments beginning of period human capital stock H it according to H it = g(s it ; H it ; it ) Functional form of g determines cost of child labor as insurance (ex. g it = H it f(s it ; it ), > 0 it is education productivity shifter, may reect illness or aggregate effects like school closings

7 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 7 Consumers solve max fc it ;S it g E 0 Household 0 s Problem " TX # t U(c t ) + (H it +1 ; B it +1 ) t=0 B it +1 = end of schooling period bequest of nancial assets c it = household consumption Dynamics of school attendance governed by Wt E t 1 it z it = 1 (1) W t 1 W t = child wage determined in child labor market, independent of human capital stock while still in school z it = g H (t) g s(t 1) g s (t) = Marginal Rate of Transformation between school attendance in adjacent periods it = shadow price of date t consumption relative to date t-1 consumption, determined by nancial market structure

8 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 8 Assume: Impact of Income Fluctuations on School Attendance g(s it ; H it ; it ) = (1 )H it f(s it ; it ) 1 f(s it ; it ) = exp exp [S it it ], > 0 where f(0; 0) = 1, f 0 > 0; 8 and f is chosen so that log z it = (S it it ). To focus on income uncertainty, if all changes in it are anticipated, taking logs and rearranging equation (1) yields S it = [log it + log W it log(1 + it )] + it (2) W where it is non-zero error in forecasting t it W t 1 z it

9 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 9 Role of Financial Markets: Complete Markets If households can trade in state-contingent consumption claims, can reallocate resources across time and states at xed prices (lifetime budget constraint) With complete markets, it constant across households in the same market at ratio of consumption claim in state " t to that in state " t 1, scaled by Separation of household's human capital investment and consumption decisions If markets are complete across villages, it = t for each village. In particular, t = 1 1+r f t Income shocks (either aggregate or idiosyncratic) do not affect school attendance If markets are only complete within villages, it = t in the village. In particular, t is the intertemporal MRS in consumption Aggregate income shocks (to F it ) transmitted to school attendance by increasing shadow price of consumption Idiosyncratic income shocks have no effect on school attendance

10 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 10 Role of Financial Markets : Incomplete Markets If consumption is not ex ante insurable, separation breaks down, extent to which depending on household's ability to transfer resources across time Relation between assets at beginning of t and end of t-1 given by A it = R t(a it 1 ) Assume MC of storage is increasing, so R 00 t < 0 for A it 1 > 0, relevant for village India Household budget constraint is now C it + A it = R t (A it ) + F it W t S it it is household specic intertemporal MRS in consumption (scaled by ), governed by E t 1 [ it Rt] 0 = 1, or equivalently log it = log Rt(A 0 t 1 ) + log(1 +! it ) where! is mean 0 forecast error in it Rt 0 from unanticipated income shocks Presence of it in equation 2 means that school attendance depends on unanticipated idiosyncratic income shocks, not merely unlike complete markets Functional form of R t determines how school attendance responds to anticipated income shocks

11 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 11 is affected by anticipated income shocks, con- With increasing MC of storage, Rt 0 sequently school attendance If R 000 < 0, magnitude of response of school attendance to increase in income is larger than response to decrease in income

12 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 12 3 Data ICRISAT Village Level Studies survey, data on time allocation of children 5-18 from 40 households in 6 villages Aurepalle and Dokur: red soil with limited water storage capacity, highly unpredictable rainfall Shirapur and Kalman: black soil with better water storage capacity, level and timing of rainfall erratic Kanzara and Kinkheda: more reliable rainfall Data based on one day recall from June 1975 to the end of 1978 rainy (Kharif) and post- No appreciable summer cultivation, so two crop seasons: rainy (Rabi) Focus on response of school attendance to income uctuations during season changes since full income only well dened over complete production cycle Two seasonal averages of school hours (including travel) S 1 averages all child's monthly observations in a season, including zeros (despite

13 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 13 possible measurement error) S 2 averages only positive observations, capturing adjustments in school day length rather than attendance After deleting entries which don't allow study of school attendance in consecutive seasons, 1256 observations on 258 children 84 cases of zero school attendance for an entire season coming between two seasons of positive attendance 20% child labor market participation based on original sample

14 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 14

15 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 15 Measures of Income F = (w m N m + w f N f + w c N c ) + + V Y = w m L m + w f L f + w c L c + + V F = full income Y = conventional income N i = number of type i people residing in the household = 156 for each season = seasonal prot from crop cultivation V = income from homemade handicrafts, sale of livestock products, land rental

16 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 16

17 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 17 4 Estimation Strategy S ivt = f logf ivt + 2 logf vt log W vt + " ivt S ivt = change in school attendance of child i, in village v, between season-years t and t 1 f logf ivt = deviation of change in log full income from village-season-year mean change logf vt = village-season-year mean log full income change " ivt captures school productivity shocks ( it ) and is assumed orthogonal to changes in full income and child wages 1 = 2 = 0 implies complete markets both within and across villages 1 > 0 suggests incomplete markets within village, though no inference on cause of market failure

18 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 18 Separating Inuence of Credit and Insurance Markets F ivt = 0vt + X 0 ivt (X ivt 1 DR vt ) u ivt 0vt is a season-village-year xed effect X ivt 1 is a vector of farm characteristics lagged one season DR vt is deviation of season-year t rainfall from long-run village mean 2 identied by assumption that effect of rain shocks varies across farms

19 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 19 c logf a ivt = X 0 ivt 1 b 1 c logf u ivt = (X ivt 1 DR vt ) 0 2 b S ivt = 0vt + 1a logf c ivt a + 1u logf c ivt u + ivt 0vt = village-season-year xed effect to purge effect of aggregate shock, focus on idiosyncratic Provides separate test of perfect intravillage credit markets ( 1a = 0) and perfect insurance markets ( 1u = 0) Increasing MC of storage consumption testable by estimating S ivt = 0vt + + 1aD + ivt c logf a ivt + 1a (1 where D + ivt = 1 if c logf a ivt > 0 D + ivt )c logf a ivt + 1u c logf u ivt + ivt

20 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 20 5 Results Both coefcients signicant, have expected sign using both trimmed (removing zeros from beginning and end of each child's time series) and untrimmed data Weaker wage effect and insignicant income effect using S 2, suggests meaningful variation comes in days of attendance rather than length of school day Replacing full income with conventional, income effect is much smaller due to ex post adjustments Wage effect still strongly signicant, negative after including adult wage: child wage doesn't reect other village shocks Smaller wage effect, insignicant income effect including village-season dummies Larger wage effect including child effects (accounts for self-selectivity IV estimates of log W vt (due to correlation with ) similar to row 7 Instruments: lagged child wage, lagged rainfall, village-season dummies

21 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 21

22 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 22 Financial Markets Structure Aggregate income changes affect school attendance more than do idiosyncratic (i.e. 2 > 1 ), though not after controlling for seasonal effects Either aggregate income shocks are fully insured though idiosyncratic are not, or aggregate seasonal uctuations are predictable (collinear w/ dummies) Insignicant effect of idiosyncratic shocks on school attendance in Kanzara may represent highest, least erratic rainfall, relative abundance of formal credit institutions Responses to aggregate shocks differ across villages Reject intra-village complete markets after adding village-season-year xed effects After instrumenting income, small farm households worse insured than large

23 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 23

24 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 24 Anticipated v: Unanticipated Instrument income with lagged farm characteristics, interactions with current deviations of quarterly rainfall, deviations squared, onset day of monsoon Both anticipated and unanticipated idiosyncratic income shocks have signicant effect on school attendance for small farms not well insured ex ante, nor have perfect access to seasonal borrowing, lending opportunities Only anticipated idiosyncratic income shocks matter for large farms, less effect than small farm + 1a > 1a, as predicted by increasing MC of storage, though not signicantly different at 5%

25 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 25

26 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 26 Impact of Market Incompleteness Attendance of uninsured child B differs in a given period from that of insured child A (S vt ) by e Bt ^ 1. With same initial HC, learning technology, expected percentage difference in HC at

27 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 27 end of education is: H A E log T +1 = E H B T +1 exp 1 e Bt TX 1 If e Bt ~N(0; 2 ), difference in terminal HC increases with 1 and 2 t=0 exp 1 S vt Small estimated Human Capital losses for uninsured children, regardless of how 1, determined Losses largest in village with most variable rainfall Even though large farm households have larger variance of idiosyncratic income shocks, ^ 1 estimate is half that of small farm households These gains represent gain from eliminating "excess" variability of attendance, ignore likely increase in average level of school attendance

28 Risk, Financial Markets, and Human Capital in a Developing Country, by Jacoby and Skouas 28 6 Conclusion Reject complete markets within village Credit market constraints, perhaps combined with limited storage, affect human capital investment decisions Costs of variable school attendance are relatively low Prospective policies must understand relevant economic risks, constraints Compulsory schooling laws may lower household welfare Policies promoting short-term credit and insurance may affect economic growth via human capital investment decisions