Measurement errors and the permanent income hypothesis : evidence from rural India

REP1 00 World Bank Reprint Series: Number 100 June 1970 Surjit S. Bhalla Measurement Errors and the Pennanent Income Hypothesis: Evidence from Rural India Reprinted with permission from The American Economic Review, vol. 69, no. 3 (June 1970), pp. 295-307   Measurement Errors and the Permanent Income Hypothesis: Evidence From Rural India By SURJIT S. BHALLA* The traditional, or Keynesian, consumption The above two questions have been exten- hypothesis postulated that (a) current income sively investigated since the publication of was a prime determinant of consumption; and Friedman's theory two decades ago.' Nev- (b) that the average propensity to consume ertheless, with one exception, all these tests declined with income. This hypothesis was (including Friedman's) have been indirect rejected by Milton Friedman's permanent and nonrigorous. The exception was an inge- income hypothesis (PIH) which contended nious study by Nissan Liviatan, who used that (a) permanent income, and not current two-year panel Israeli data to test the PIH, income, was the relevant determinant of and, in particular, the proportionality hy- consumption; and (b) that permanent con- pothesis.4 His study, however, suffered from a sumption was proportional to permanent major drawback in that it failed to account income (the proportionality hypothesis).' for the bias caused by common measurement This paper attempts to test (and distinguish errors in consumption and income. This paper between) the two theories. Answers to two extends Liviatan's analysis by explicitly questions are sought: How important is the allowing for the presence of measurement current income-permanent income distinc- errors in all the variables-income, consump- tion; and is the proportionality hypothesis, an tion, and saving. The empirical estimation of important and controversial aspect of Fried- the variances of these errors (source of bias) is man's theory, valid? The resolution of the possible due to the availability of independent former question has important implications estimates of consumption, savings, and for the understanding of habit and lags in income.5 The data base is a three-year panel consumption behavior and the efficacy of survey, conducted by the National Council of short-run macro-economic policies. The valid- Applied Economic Research (NCAER), New ity of the proportionality hypothesis has an Delhi, of 4,118 households in rural India. important bearing on the recently popular (See Appendix A for description of data and controversy regarding the tradeoff between definitions of variables.) growth and equity in developing countries.2 The incorporation of error variances in the *World Bank. This is a shortened version of a paper therefore growth. (This is predicated on the assumption which was written while I was a recipient of a Rockefeller that growth in less developed countries (LDCs) is Post-Doctoral Fellowship at the Rand Corporation The constrained by a lack of finance rather than of investment financial support is gratefully acknowledged. I would like opportunities.) Thus, the PIH contends that there is no to thank William Rogers for helpful discussions, Thomas tradeoff between redistribution (equity) and growth. Mayer and Charles E. Phelps for comments on an earlier 3An excellent summary and interpretation of the volu- draft, and David Harrison for research assistance. The minous literature on consumption behavior is contained views expressed in this paper are mine alone. in Thomas Mayer's book. 'Analogous conclusions are reached by the Modigliani- 4Friedman accepted the compelling logic of Liviatan's Brumberg life cycle hypothesis (LCH) of consumption tests and stated that " . . if these [Liviatan's] results behavior. The two theories, LCH and PIH, use different should be confirmed for other bodies of data, they would empirical methods but employ a common theoretical constitute relevant and significant evidence that the model of consumer behavior. The empirical techniques elasticity of permanent components is less than unity [the used in this paper apply directly to Friedman's formula- proportionality hypothesis]" (1963a, p. 63). tion of the theory. Hence, the emphasis is on testing the 5These independent estimates allow for two measures PIH. of consumption-direct consumption and residual con- 21f, as the proportionality hypothcsis implies, the sumption (income minus savings). This "excess" of propensity to consume permanent income is independent information is what makes the estimation of error vari- of the level of permanent income, then redistribution ances possible. See Section 11 (and Appendix B) for policies will be neutral with respect to savings, and details of methodology. 295  296 THE AMERICAN ECONOMIC REVIEW JUNE 1979 analysis, a unique aspect of this study, insures (2a) Z* = Z' + Z" that unlike Liviatan's tests, the tests of the (2b) Z Z* + Z