100217 AUTHOR ACCEPTED MANUSCRIPT FINAL PUBLICATION INFORMATION Food Prices, Wages, and Welfare in Rural India The definitive version of the text was subsequently published in Economic Inquiry, (Forthcoming 2015), 2015-06-18 Published by Wiley and found at http://dx.doi.org/10.1111/ecin.12237 THE FINAL PUBLISHED VERSION OF THIS ARTICLE IS AVAILABLE ON THE PUBLISHER’S PLATFORM This Author Accepted Manuscript is copyrighted by the World Bank and published by Wiley. It is posted here by agreement between them. Changes resulting from the publishing process—such as editing, corrections, structural formatting, and other quality control mechanisms—may not be reflected in this version of the text. You may download, copy, and distribute this Author Accepted Manuscript for noncommercial purposes. Your license is limited by the following restrictions: (1) You may use this Author Accepted Manuscript for noncommercial purposes only under a For non- commercial use in accordance with Wiley Terms and Conditions license http://olabout.wiley.com/ WileyCDA/Section/id-820227.html. (2) The integrity of the work and identification of the author, copyright owner, and publisher must be preserved in any copy. (3) You must attribute this Author Accepted Manuscript in the following format: This is an Author Accepted Manuscript of an Article by Jacoby, Hanan G. Food Prices, Wages, and Welfare in Rural India © World Bank, published in the Economic Inquiry(Forthcoming 2015) 2015-06-18 For non-commercial use in accordance with Wiley Terms and Conditions http://olabout.wiley.com/WileyCDA/Section/id-820227.html http://dx.doi.org/10.1111/ecin.12237 © 2015 The World Bank FOOD PRICES, WAGES, AND WELFARE IN RURAL INDIA HANAN G. JACOBY∗ With soaring food prices in recent years has come alarm about rising poverty in the developing world. Less appreciated, however, is that many of the poor in agricultural economies may benefit from higher wages. This study finds that wages for manual labor in rural India, both within and outside agriculture, rose faster in districts growing more of those crops with large producer price run-ups over the 2004–2009 period. Based on a general equilibrium framework that accounts for such wage gains, rural households across the income spectrum are found, contrary to more conventional welfare analysis, to benefit from higher agricultural prices. (JEL Q17, Q18, F14) I. INTRODUCTION of the world’s poor (those living on less than $1.25/day), the answer to this question can have Elevated food prices over the last half decade momentous ramifications. After all, the vast have provoked a rash of government interven- majority of India’s rural population relies on tions in agricultural markets across the globe, the earnings from their manual labor, most of often in the name of protecting the poor. Of which is devoted to agriculture.3 Any thorough course, it is well recognized that many poor accounting of the global poverty impacts of households in developing countries, especially improved terms of trade for agriculture must, in rural areas, are also food producers and hence therefore, confront rural wage responses in India. net beneficiaries of higher prices.1 Even so, there Textbook partial equilibrium analysis (e.g., is another price-shock transmission channel, Deaton 1989; Singh et al. 1986) considers only potentially more important to the poor, which the direct income effect of a price change on has received far less attention in the literature: household welfare, which, to a first order, is pro- rural wages.2 To what extent do higher agri- portional to the household’s production of the cultural commodity prices translate into higher good net of consumption. While this approach wages? For rural India, home to roughly a quarter is useful for understanding the very short-run welfare impacts of price shocks, it ignores the ∗ I am grateful to David Atkin, Madhur Gautam, Denis inevitable labor market repercussions of persis- Medvedev, Rinku Murgai, Maros Ivanic, and Will Martin for tent price changes. Insofar as higher agricultural useful suggestions and to Maria Mini Jos for assistance in processing the data. The findings, interpretations, and con- prices lead to higher wages, then, there are three clusions of this paper are mine and do not necessarily reflect channels of general equilibrium welfare effects: the opinions of the World Bank, its executive directors, or the (1) higher labor income; (2) lower capital (land) countries they represent. income due to higher labor costs; and (3) higher Jacoby: Lead Economist, Development Research Group, The World Bank, Washington, DC 20433, Phone prices for nontradables. To quantify these effects (202)-458-2940, Fax (202)-522-1151, E-mail hjacoby@ and obtain the full welfare impact of changes worldbank.org in agriculture’s terms of trade, one needs, first and foremost, an estimate of the relevant wage- 1. Ivanic, Martin, and Zaman (2012), Wodon et al. price elasticity. (2008), and World Bank (2010) provide recent multi-country assessments of the welfare impacts of food price increases A few existing studies estimate wage-price accounting for such producer gains. See also the study by de elasticities using long aggregate time series data Janvry and Sadoulet (2009) for an analysis along these lines from countries that were effectively autarkic in using Indian data. 2. Ravallion (1990) surveys the debate in development on the nexus between the intersectoral terms of trade and poverty. Sah and Stiglitz (1987) provide an early theoretical treatment. In their cross-country study, Ivanic and Martin (2008) incor- porate price-induced changes in wages for unskilled labor 3. Indeed, rising wages are seen as the major driver of derived from nation-level versions of the GTAP computable rural poverty reduction in recent decades (Datt and Ravallion general equilibrium model. 1998; Eswaran et al. 2007; Lanjouw and Murgai 2009). 1 Economic Inquiry doi:10.1111/ecin.12237 (ISSN 0095-2583) © 2015 International Bank for Reconstruction and Development/The World Bank. 2 ECONOMIC INQUIRY the main food staple (pre-1980s Bangladesh in nontradable price elasticity without actually esti- Boyce and Ravallion 1991; the Philippines in mating it econometrically. Lasco, Myers, and Bernsten 2008), thus raising To evaluate the distributional impacts of serious endogeneity issues. Alternatively, Porto changes in agriculture’s terms of trade, I integrate (2006) estimates the wage impacts of changes a three-sector, specific factors, general equilib- in traded goods prices using several years of rium trade model (e.g., Jones 1975) into a first- repeated cross-sectional household survey data order welfare analysis.5 Appealing to the widely from Argentina. In the case of agricultural noted geographical immobility of labor across goods, which must somehow be aggregated, rural India (e.g., Topalova 2007, 2010),6 I apply Porto creates a price index using household this general equilibrium framework at the district expenditure shares as weights (as does Nicita level, treating each of these several hundred 2009). To appreciate the issue involved with this administrative units as a separate country with its strategy, consider an extreme example. Suppose own labor force but with open commodity trade that a country is a net exporter of cotton and net across its borders.7 This district-level perspective importer of wheat, its sole consumption item. has two implications for empirical implemen- Since the cotton industry is a major demander tation of my approach. First, since each district of labor, a rise in the cotton price should lead to produces a different basket of agricultural com- higher wages (and, ultimately, higher welfare); modities, differences in the magnitude of whole- conversely, a rise in the wheat price should sale price changes across crops (even if common have little impact on wages (i.e., only through across districts), generate cross-district variation an income effect on labor supply). Hence, in in agricultural price (index) changes. Second, this scenario, the correlation between changes following the logic of the model, the wage-price in wages and changes in the expenditure share elasticity itself is specific to a district, varying weighted agricultural price index may well be with characteristics of the local labor market. close to zero. Clearly, however, this is not the While my estimation strategy is related to the relevant wage-price elasticity for our purposes. “differential exposure approach” (Goldberg and Indeed, as I show in the context of a formal Pavcnik 2007) employed in studies of the local general equilibrium trade model, the relevant wage impacts of tariff reform (most recently in elasticity is one based on a production share Topalova 2010; McCaig 2011; and Kovak 2010, weighted agricultural price index.4 2013), there are several novel elements. Kovak, Even with the correct wage-price elasticity for example, uses the same type of theoreti- estimate in hand, one must still wrestle with what cal model to motivate his empirical specifica- to do about non-traded goods. One option is to tion, but he has many industrial sectors; there simply ignore them; that is, by assuming either is no distinctive treatment of agriculture. More- that they constitute a negligible share of the bud- over, Kovak ignores intermediate inputs, whereas get or that their prices are fixed. Unfortunately, in this paper intermediates play a quantitatively the first assumption is counterfactual, at least in important role in transmitting food price shocks. the case of India, and the second assumption is Finally, Kovak does not consider the welfare or inconsistent with theory. As I will show, in a distributional implications of trade shocks, or of multisector general equilibrium model, in which food price shocks more particularly, which is a one of the sectors is nontradable, the price of point of departure for this paper. Topalova (2010) the nontraded good is increasing in the agricul- finds that tariff reductions during India’s trade tural price index. Recognizing this possibility, Porto (2006) provides one of the few, if only, 5. Another strand of the literature incorporates second- econometric estimates of the elasticity of non- order (substitution) effects of price increases on the consump- tion side based on demand-system estimation (most recently, traded goods prices with respect to traded goods Attanasio et al. 2013). Banks, Blundell, and Lewbel (1996), prices. In India, however, as in most developing however, provide evidence that first-order approximations countries, reliable data on prices of services and do reasonably well (relative error of around 10%) for price other nontradables are unavailable. One contri- changes on the order of 20%. bution of this paper, therefore, is to quantify the 6. Kovak (2010) finds no evidence that labor migration matters for local wage responses to trade reform in Brazil, a country with much higher inter-regional labor mobility 4. A related issue is that prices or unit values obtained than India. from household expenditure surveys (as in Marchand 2012; 7. Capital (land, in agriculture) is also assumed immo- Porto 2006) may not reflect the wholesale prices faced by bile across both districts and production sectors. Longer-run farmers in a particular region, especially where government Stolper–Samuelson effects are not of paramount concern in intervention is heavy (as in India). policy discussion of food price shocks. JACOBY: FOOD PRICES, WAGES, AND WELFARE IN RURAL INDIA 3 liberalization led to a fall in wages, including In the next section, I sketch the theoreti- agricultural wages, and to a rise in rural poverty. cal framework and develop my empirical testing Although Topalova’s analysis is reduced-form strategy. Section III discusses the econometric and ex-post,8 she interprets her findings through issues and the estimates. Section IV presents the the lens of a specific-factors trade model with sec- distributional analysis of food price shocks, com- torally immobile labor (and mobile capital). Such paring the general to partial equilibrium scenar- a model, however, implies that nonagricultural ios. I conclude, in Section V, with a discussion wages would fall with higher food prices and, of the Government of India’s responses to the hence, that households would be affected very 2007–2008 food price spike, notably its export differently by rising food prices according to the ban on major foodgrains. sector in which their members are employed. My evidence will show the contrary, that the wage benefits of higher food prices are similar across II. GENERAL EQUILIBRIUM FRAMEWORK employment sectors. More broadly, Topalova’s A. Model Assumptions results do not speak directly to the impact of shifts in agriculture’s terms of trade.9 This study Consider each district as a separate economy is thus the first to adapt the differential exposure with three sectors: agriculture (A) and manu- approach specifically to the agricultural sector facturing (M ), both of which produce tradable and to the question of food-price crises. goods, and services (S), which produces a non- My empirical analysis finds that nominal tradable. The reason it is necessary to distinguish wages for manual labor across rural India services from manufacturing is simple. Combin- respond elastically to higher (instrumented) agri- ing the two into one nontradable nonagricultural cultural prices.10 In particular, wages rose faster sector is tantamount to allowing changes in agri- in the districts growing relatively more of the cultural prices to affect the prices of both manu- crops that experienced comparatively large run- factured goods and services. Since manufactured ups in price over the 2004–2005 to 2009–2010 goods are, in fact, tradable, this approach would period. Importantly, the magnitude of these wage overstate the welfare impact of changes in agri- responses is broadly consistent with the quanti- culture’s terms of trade. tative predictions of the specific-factors model. Continuing with the assumptions, output Y i in These results have striking distributional impli- each sector i = A, M , S is produced with a spe- cations. Improved terms of trade for agriculture, cific (i.e., immobile) type of capital K i , along rather than reducing the welfare of the rural poor with manual labor Li and a tradable intermediate as indicated by the conventional approach (which input I i , using sector-specific production function ignores wage impacts), would actually benefit Y i = F i (Li , I i , K i ). In the case of agriculture, K A both rich and poor alike, even though the latter is land and I A is, for example, fertilizer. Interme- are typically not net sellers of food.11 diate inputs do not play an essential role, except insofar as the model provides quantitative predic- tions, in which case (as we will see) they make a 8. In other words, rather than predicting distributional impacts from a model based on estimated elasticities, it looks big difference. at changes in poverty rates directly. By contrast, Marchand In India, as in most developing countries, (2012) uses an ex-ante simulation of household consumption agricultural production largely takes place on along the former lines to find that the fall in India’s trade barriers during the 1990s would have reduced rural poverty household-farms using family and hired labor. in India. Moreover, in a given year, these farms typi- 9. In particular, as high nontariff barriers on agricultural cally produce several crops on the same land products remained in force well after India’s initial trade (contemporaneously via multicropping and/or liberalization (Anderson 2009), it is not clear what actually happened to the relative price of agriculture in Topalova’s sequentially in multiple cropping seasons) with post-reform period. largely the same workers and intermediate 10. In rural India, manual labor by far predominates over inputs. Hence, following, for example, Strauss nonmanual labor in terms of annual days worked, and much (1984), I treat the representative farm as a of the latter is in the public (i.e., nonmarket) sector. Thus, unlike, e.g., Porto (2006) or Nicita (2009), I do not attempt multiproduct firm that chooses among a fixed to estimate separate wage-price elasticities for skilled and set of c crops {Y 1 ,..., Y c } to grow, transform- unskilled workers. ing between them according to the function 11. To be sure, the increase in rural wages may lag the increase in consumer prices, and so the conventional analysis Y A = G(Y 1 ,..., Y c ), where G is assumed to be may be more appropriate for the very short run. This article homogeneous of degree one. To account for does not speak to the timing issue. the huge agroclimatic variation across India, one 4 ECONOMIC INQUIRY TABLE 1 Summary Statistics for Major Crops Area Share Value Share No. of Districts pj − price Rice 0.380 0.408 447 0.000 (0.320) (0.328) Wheat 0.225 0.199 390 −0.032 (0.183) (0.165) Soyabean 0.092 0.099 153 0.056 (0.151) (0.159) Bajra 0.076 0.037 287 −0.064 (0.146) (0.091) Cotton 0.076 0.128 206 −0.130 (0.112) (0.175) Maize 0.067 0.054 410 −0.011 (0.112) (0.103) Jowar 0.065 0.024 317 −0.041 (0.110) (0.040) Ragi 0.052 0.030 192 0.052 (0.123) (0.092) Groundnut 0.046 0.050 349 −0.112 (0.115) (0.115) Gram 0.043 0.045 385 −0.195 (0.072) (0.087) Sugarcane 0.035 0.090 386 0.001 (0.082) (0.164) Rapeseed/Mustard 0.034 0.038 367 −0.199 (0.073) (0.090) Urad 0.028 0.012 409 0.364 (0.042) (0.018) Moong 0.025 0.014 424 0.586 (0.041) (0.030) Arhar 0.021 0.019 428 0.253 (0.033) (0.033) Potato 0.019 0.053 312 −0.146 (0.061) (0.105) Sunflower 0.014 0.009 271 −0.083 (0.048) (0.032) Sesamum 0.012 0.008 387 0.053 (0.022) (0.022) Notes: Means (standard deviations) of district-level data and number of districts growing each crop in 2003–2004. Log-price changes for 2004–2009 are averages across the 18 major states of India weighted by state production shares. should think of the set of feasible crops as varying Now, we may write profit per acre in agricul- across districts. ture as ΠA = [PA F A (LA , I A , K A ) − PI I A − WLA ]/ Farmers then choose the particular quanti- K A , with analogous expressions for average ties to grow, the Y j , to maximize total revenue, profit per unit capital in manufacturing, ΠM and ∑c j=1 Pj Yj , where Pj is the price of crop j, subject in services ΠS , given respective output prices in to the constraint that G(Y 1 ,..., Y c ) = Y A for any these sectors, PM and PS . I assume that manual given Y A . Thus, in∑ this set-up, production value labor is perfectly mobile across the three sectors shares sk = Pk Yk ∕ c j=1 Pj Yj are determined by but its overall supply is fixed at L = LA + LM + LS both agroclimatic conditions and by relative crop within each district. Thus, in each district econ- prices. Given the homogeneity of G, there omy, there is one type of labor with a single ∑ exists a price index PA such that PA YA = c j=1 Pj Yj , nominal wage, W , and a unique wage-price which upon differentiation yields elasticity ∑ ̂A ̂ ∕P (1) ̂A = P ̂j sj P (2) ψ≡W j that must be solved for. where “hats” denote proportional changes; that is, Because this is a general equilibrium frame- ̂ x = d log x. This establishes our production value work, income effects of changes in factor prices share-weighted agricultural price index. are fully accounted for. Thus, total income y JACOBY: FOOD PRICES, WAGES, AND WELFARE IN RURAL INDIA 5 consists of the sum of value-added (revenue net by considering the special case αI = βS = 0; a input expenditures) across sectors i = A, M , S two-input, two-sector economy (without non- ∑ tradables). According to Equation (5), in this (3) y= Pi Yi − PI Ii + E case ψ = βA , where βA is the share of the rural i labor force in agriculture. Referring to Figure 1, with an additional exogenous component, E. compare equilibrium A, with a high share of Although a technical nuisance, the presence of labor in agriculture to equilibrium B with a low E suits an important empirical purpose: A sig- agricultural share. At A the value of marginal nificant portion of household income in rural product curve in manufacturing (the supply India comes from (salaried) nonmanual labor; curve of labor to agriculture) is necessarily very for example, teachers, police/army, and other steep; at B it is very flat. Thus, in moving from ′ civil servants. The exogeneity assumption on this A to A , a 50% increase in the agricultural price income can be motivated by thinking about entry translates into an almost 50% increase in the ′ into these professions as requiring an advanced wage, whereas, in moving from B to B , the same level of education (relative to unskilled labor), price increase leads to virtually no wage increase which cannot be acquired in the short-run.12 whatsoever (in proportional terms). If we now let αI > 0, then we have B. Solution and Intuition ψ = βA /(αL + αK ) > βA . So, while the qualita- tive prediction is the same, the magnitude of We are interested in what happens to the equi- the wage-price elasticity can increase quite a lot librium wage in this model when the agricultural after accounting for the cost share of intermedi- price index changes, holding other tradable prices ate inputs. The source of this amplification effect ̂M = P constant; that is, P ̂I = 0. Given that farm- is the increase in intermediate input use induced ers are price-takers in all markets, we have (from by higher agricultural prices, which boosts the price equals unit cost) marginal product of labor in agriculture. Because ̂ + αK Π ̂A ̂A = P of a greater exodus of labor from manufac- (4) αL W turing in response to agriculture’s improved where, under constant returns to scale, the input terms of trade, there must be an even larger cost shares in agriculture, the αl , l = K , L, I , are wage increase than was the case in the absence such that αK + αL + αI = 1. Similar equations of intermediates. hold for the other sectors, each with its own set Finally, let us return to δ in Equation (5). of input cost shares. In the interest of clarity and To solve out this parameter, we must equate the because it will make no appreciable difference demand and supply of services, which I discuss empirically (see below), I assume equal input in the Appendix. For purposes of exposition, set cost shares across sectors from now on. αI = 0 again and consider the special case E = 0, As I show in the Appendix, in which there is no exogenous source of income ( ) ( ) outside of the three sectors. As shown in the (5) ψ = βA + δβS ∕ αL + αK Appendix, ψ = δ = βA /(1 − βS ) > βA in this case. where the βi = Li /L are the sectoral labor shares Thus, the introduction of a nontradable sector ̂S ∕P̂A . Note that δ, the elasticity of also amplifies the wage-price elasticity. In this and δ ≡ P economy, a rise in the wage induced by higher the nontradables price with respect to the price agricultural prices reduces the supply of services; of agriculture, is endogenous and needs to be it also increases the demand for services due to solved out.13 an income effect. Both forces put upward pres- Before doing so, however, we can gain some intuition for the mechanics of the model sure on the price of services, so that δ > 0. With the expansion of the service sector as agricul- 12. We can also think of rural nonmanual labor as paid tural prices rise, the supply curve of labor to for out of a central government budget financed by urban agriculture becomes even more inelastic, mak- taxpayers and not contributing directly to output in any rural ing the rural wage even more sensitive to these sector. price changes. ̂ ̂ ( 13.) Combining ( ) (4) and (5) also gives ΠA ∕PA = 1∕αK 1 − αL ψ . So the elasticity of the return on land with respect to the agricultural price index incorporates C. Empirical Validation the direct (positive) effect of price changes on farm profits as well as the indirect (negative) effect of price induced wage The advantage of the above machinery is changes. twofold: First, the model tells us what the relevant 6 ECONOMIC INQUIRY FIGURE 1 Labor Market Equilibrium wage-price elasticities are and, second, it delivers statistics).16 Generally speaking, the estimated explicit expressions for these elasticities in terms elasticities are high (ψ = 1.15), reflecting large of input cost shares, sectoral labor shares, and values of βA . Indeed, for the average rural dis- other parameters, all of which can be computed trict, around three-quarters of manual labor days from nationally representative data collected by (adjusted for efficiency units; see Sectoral Labor India’s National Sample Survey (NSS) Organiza- Shares section in Appendix) are spent in agri- tion.14 I thus calculate district (d) specific wage- culture. Note also that intermediate inputs play a price elasticities, ψ d , assuming equal input cost quantitatively important role in the elasticity cal- shares across sectors,15 for 472 districts in the 18 culation; if I assume that αI = 0, then ψ would major states of India (see Table A2 for descriptive drop to 0.85. In other words, the input amplifi- cation effect on the wage-price elasticities, dis- 14. An exception is the share of aggregate income cussed in the previous section, is substantial. from exogenous sources, or E/y (cf., Appendix), In principle, one could econometrically esti- which is computed at the state-level from IHDS data mate separate wage-price elasticities for each described below. district and compare them to their theoretically 15. While it is straightforward to allow for sector-specific input cost shares using the results in Appendix, it is somewhat messy. Fortunately, it hardly matters, because they yield vir- 16. Excluded are the peripheral states of Jammu/Kashmir tually identical elasticity results as in the equal shares case. in the far north and Assam and its smaller neighbors to the Cost shares of value-added for Indian manufacturing and ser- north and east of Bangladesh. Included states, organized into vice sectors based on national accounts are available from five regions, are North: Harayana, Himachal Pradesh, Pun- Narayanan, Aguiar, and McDougall (2012). As it turns out, jab, Uttar Pradesh, and Uttaranchal; Northwest: Gujarat and however, the ratio of capital to labor shares is what is most Rajastan; Center: Chhattisgarh, Madhya Pradesh, Maharash- relevant to our calculations, and these are quite similar across tra, and Orissa; East: Bihar, Jharkhand, and West Bengal; sectors. South: Andhra Pradesh, Karnataka, Kerala, and Tamil Nadu. JACOBY: FOOD PRICES, WAGES, AND WELFARE IN RURAL INDIA 7 implied counterparts above. In practice, however, can be, procured for eventual release into the this would require long time-series of wages and nationwide public distribution system (PDS). prices for each district over a period of structural In practice, however, the level of procurement, stasis.17 In lieu of such data, I estimate the regres- and thus the extent to which the MSPs are sion analog to the identity given by Equation (2), binding, varies greatly by crop and state, and or even within states (Parikh and Singh 2007). The ∑ principal foodgrains, rice and wheat, have, in (6) Δwd ∕ψd = c + γ sd,j Δpj + εd recent years, been the overwhelming focus of j government procurement efforts, concentrated in where c is an intercept, γ is a slope parameter, and the states of Punjab and Haryana, often for lack εd is a disturbance term for each district d. Thus, of storage capacity and marketing infrastructure Equation (6) replaces W ˆ and P ̂A by their empirical elsewhere. By contrast, procurement of pulses counterparts; Δwd is the difference in log wages and oilseeds has been minimal, as market prices between years t − k and t and the Δpj are the have consistently exceeded MSPs.18 corresponding time-differences in log prices of During and after the sharp run-up in interna- crop j, which are weighted by production value tional food prices in 2007–2008, the Government shares sd,j as already discussed. of India imposed export bans on rice, wheat, and a Under the null hypothesis, which is that the few other agricultural commodities in an attempt model and all its auxiliary assumptions holds to tamp down domestic price increases. Mean- true on average, we have γ = 1. In other words, while, over several consecutive years, MSPs for under the null, the magnitude of observed wage rice and wheat (and most other major crops) were responses to actual changes in the agricultural raised substantially, partly in response to interna- price index corresponds (in an average sense) to tional prices; huge stockpiles of foodgrains were what the theory says it should be. Several econo- subsequently accumulated through government metric issues arise in implementing Equation (6), procurement (Ahmed and Jansen 2010; Himan- including potential endogeneity of price changes. shu and Sen 2011). These are left for Section III.D. The upshot of these interventions is that output prices faced by Indian agricultural producers do not always perfectly track those in international III. EMPIRICAL ANALYSIS markets.19 Moreover, as domestic market integra- tion is somewhat limited (especially in the case A. Domestic Agricultural Markets of rice), there is considerable variability across Since at least the 1960s, Indian govern- states in crop price movements. On the one hand, ments, both at the national and state level, have this variation may reflect differential transmis- intervened extensively in agricultural markets. sion of exogenous price pressure (e.g., because of Interstate trade in foodstuffs is often severely varying levels of state procurement or exposure circumscribed through tariffs, taxes and licens- to trade, both with other countries and with other ing requirements (see Atkin 2010, for a review) states); on the other hand, it may reflect localized with some states (e.g., Andhra Pradesh) going supply or demand shocks, which can also drive so far recently as to prohibit the exportation of rural wages directly. rice to other states (Gulati 2012). The Govern- ment of India also sets minimum support prices B. Crop Prices (MSPs) at which major food crops are, or at least Wholesale crop price data averaged at the state level from observations at several district markets 17. India’s quinquennial labor force survey, available since 1983, would yield, at best, five first-differenced wage per state (and weighted by district production), observations per district. Alternatively, Jayachandran (2006) are compiled by the Ministry of Agriculture, examines a 30-year agricultural wage series for Indian dis- tricts. Although these data fall entirely within the pre-reform (largely autarkic) trade regime, Jayachandran estimates a 18. See the reports by the Commission for Agricultural national-level wage elasticity with respect to agricultural TFP, Costs and Prices on http://cacp.dacnet.nic.in/ for more details. instrumented with rainfall shocks. One of the difficulties with 19. This is true for the principal intermediate input in interpreting this as an estimate of ψ , however, is that the year- agriculture as well. Despite a substantial upsurge in the inter- to-year TFP shocks induced by annual rainfall deviations are national prices of chemical fertilizers beginning in 2007, retail unanticipated and thus are unlikely to give rise to the sectoral prices in India, which are set by the central government, labor reallocations underlying the general equilibrium model remained uniform and unchanged over the 2004–2009 period of this paper. (Sharma 2012). 8 ECONOMIC INQUIRY TABLE 2 Rural Wage Impacts of Crop Price Changes: 2004–2009 (1) (2) (3) (4) (5) (6) (A) Wages for all manual labor (N = 462) γ 0.429 0.547 0.864 0.822 0.847 0.846 (0.100) (0.105) (0.305) (0.302) (0.318) (0.320) ΔPWa 0.042 (0.215) p-values H0 : γ = 1 0.000 0.014 0.672 0.579 0.660 0.663 H0 : γ = 0 0.000 0.001 0.006 0.009 0.014 0.018 Cragg-Donald F -stat (weak identification test) 1384.1 61.0 50.0 39.0 38.5 (B) Wages for nonagricultural manual labor (N = 445) γ 0.672 0.779 0.988 0.844 0.900 0.851 (0.109) (0.104) (0.263) (0.245) (0.249) (0.250) ΔPW a −0.228 (0.242) p-values H0 : γ = 1 0.010 0.461 0.969 0.585 0.733 0.613 H0 : γ = 0 0.000 0.000 0.004 0.008 0.006 0.010 Cragg-Donald F -stat (weak identification test) 1522.6 73.1 59.0 48.2 49.8 Instrument — IV 1d IV 2d IV 3100 d IV 3200 d IV 3200 d Notes: Standard errors robust to spatial dependence in parentheses. All p-values based on Bester et al. (2014) bootstrapped critical values (R = 10000). Dependent variable is the change in log wage district fixed effect between 2004 and 2009 scaled by the district wage-price elasticity. All regressions include a constant term and are weighted by the inverse estimated sampling variance of the dependent variable. See text for definition of instruments. a Difference in average days of public works employment per week in district between 2004 and 2009. as are production and area data at the district major field crops according to the criteria that level. So as to focus on a period of substan- they cover at least 1% of total cropped area tial price movement, as well as to match the nationally or that at least five districts had no NSS wage data (see below), I consider state- less than 10% of their cropped area planted to level price changes between the 2004–2005 and them in 2003–2004. These 18 crops, listed in 2009–2010 crop marketing seasons. Given the Table 1 in descending order of planted area, relative ease of moving produce across district comprise some 92% of area devoted to field (as opposed to state) lines, state-level wholesale crops in 2003–2004 in the major states of India. prices appear the appropriate measure of farmer Table 1 also reports national average log-price production incentives.20 changes (weighted by the state share of total I base the crop value shares, the sd,j in production) relative to rice. Thus, in the first Equation (6), on production data from the row, the relative price change for rice is zero, 2003–2004 crop-year, which has the best dis- quite negative for several important crops (e.g., trict/crop coverage for the pre-2004–2005 cotton, gram, groundnut, and mustard/rapeseed) period. Value of production is calculated at and highly positive for pulses (Urad, Moong, 2004–2005 state-level prices. Note, however, and Arhar). that I do not take the value-weighted sum of price changes across every single agricultural product grown in India. Price data for many of the minor C. Wages field crops and the tree crops are incomplete Wage data are derived from the NSS or not reliable. Moreover, the associated pro- duction data are often inaccurate (especially Employment-Unemployment Survey (EUS), for vegetables and tree products). I thus select normally conducted every 5 years. The most recent round, the 66th, collected in 2009–2010, is the first conducted in the wake of the food price 20. As sugarcane is sold mostly to mills and not in “crisis” of 2007–2008, whereas the 61st round of wholesale markets, I use the national MSP or, when relevant, “State Advised Prices,” which tend to be much higher and, 2004–2005 most closely preceded it. Once again, hence, closer to international cane pricing standards (see in the spirit of the theoretical model, I focus on Gulati 2012). manual labor, which constitutes nearly 83% of JACOBY: FOOD PRICES, WAGES, AND WELFARE IN RURAL INDIA 9 days of paid employment in rural areas.21 The and price changes. For instance, suppose that a first-stage of the estimation takes individual log particular district has been industrializing rela- daily wages in the last week and regresses them tively rapidly over the 2004–2009 period, or that on district fixed effects as well as a quadratic in it has experienced comparatively rapid techno- age interacted with gender. Thus, I estimate the logical improvement in agriculture. Both types of respective log-wage district fixed effects, wd,09 shocks would tend to raise district wages. And, and wd,04 , separately for each round, removing, they may influence crop prices as well insofar as via the constant terms, year effects due to, for the state’s agricultural markets are insulated from example, general inflation. Estimates ( of the ) the rest of India (and the world) and the district is standard errors of the fixed effects σ wd,09 important relative to that market, or the shocks ( ) are strongly spatially correlated. and σ wd,04 , which I use below to construct regression weights, are obtained following the The next step, therefore, is to develop an procedure of Haisken-DeNew and Schmidt instrument that is uncorrelated both with district- (1997). level wage shocks and with measurement error in price changes (and crop value shares). Consider, then, D. Identification ∑ Rewriting Equation (6) to reflect the price data (9) IV 2d = ad,j ΔpSTATEd ,j j discussed above, I wish to estimate ∑ where ΔpSTATEd ,j is the production share weighted (7) Δwd ∕ψd = c + γ sd,j ΔpSTATEd ,j + εd . j mean change in the log-price of crop j across states excluding the state to which district d where STATEd denotes the state in which district belongs.22 In other words, IV 2d replaces the state d is located. There are two endogeneity issues to price changes in IV 1d with a national average contend with: measurement error and simultane- price change uncontaminated by state-specific ity between wage and price changes. shocks or measurement error because no price As to the first issue, both the crop value data from that state or production data from that shares, sd,j , and the crop-specific log-price district are used in its construction. The idea, changes, ΔpSTATEd ,j , may be measured with error. then, is that ΔpSTATEd ,j reflects exogenous inter- Putting aside the latter concern momentarily national price changes transmitted to other states and assuming that measurement error is con- of India as well as shifts in demand and sup- fined solely to value shares, I could deploy the ply in the vast domestic market outside of the instrument particular state. ∑ A problem with IV 2d , however, is that it does (8) IV 1d = ad,j ΔpSTATEd ,j not meet the exclusion restriction if εd are cor- j related across state boundaries. In other words, where ad,j is the area share of crop j in district d. if industrialization or agricultural innovation (or To be sure, cropped areas may also be measured even weather) in, say, southern Andhra Pradesh with error, but these errors should not be corre- and northern Tamil Nadu move together, then the lated with those of crop production and prices. ΔpSTATEd ,j for a district in Andhra Pradesh may Clearly, IV 1d does not deal with measure- reflect these shocks inasmuch as price changes ment error in price changes, which could arise from Tamil Nadu contribute to the weighted aver- if, for example, the marketed varieties or grades age. To deal with this concern, I first establish some notation: Let BSTATEd r be the set of states of a certain crop in a certain state change over time. Another concern is unobserved district- within a radius of r kilometers around district d; level shocks (or trends) correlated with both wage of course, STATEd ⊆ BSTATEd r . Thus, BSTATEr d for the district in southern Andhra Pradesh, 21. The NSS-EUS categorizes jobs in terms of manual depending on r, may include Karnataka and and non-manual labor only for rural, not urban, workers. Tamil Nadu (in addition to AP itself), whereas, Based on the 61st round sample of nearly 39,000 individ- uals, the population-weighted proportions in each category ∑ are as follows: 58% in manual-agricultural; 24% in man- 22. To be precise, ΔpSTATEd ,j = C k∈STATEd ωkj Δpk,j , ual nonagricultural; and 18% in nonmanual (virtually all in C is the set of states excluding STATE and ω where STATEd d kj nonagriculture). For the 66th round sample of some 30,000 individuals, the corresponding proportions are 51%, 30%, and is state k’s share of total production of crop j among all states in STATEd .C 19%, respectively. 10 ECONOMIC INQUIRY if d were instead in northern AP, BSTATEd r might the clustered covariance estimator. But clustering include Maharashtra and Chhattisgarh. With this standard errors by state or region assumes inde- definition, my instrument becomes pendence of errors across state or regional bound- ∑ aries, a serious lacuna given the large fraction of (10) IV 3r d = ad,j ΔpBSTATEr ,j d districts bordering an adjacent state.23 j Bester et al. (2014) show that the asymptotic where ΔpBSTATEr ,j is the production share normal distribution, typically used to obtain crit- d weighted mean change in the log-price of ical values for inference in HAC estimation, is crop j across states excluding those in BSTATEd r. a poor approximation in finite samples. I thus Here, again, the logic is that the price instru- follow their suggestion of bootstrapping the dis- ment should not directly, or, in this case, even tribution of the relevant test-statistics. For this indirectly, be driven by local shocks that also reason, inference should be guided by p-values determine differential wage growth across rather than by standard errors, although I will districts (and states). follow convention and report both. In particu- The choice of r, the radius of “influence” of lar, bootstrapped p-values are much less sensitive local wage shocks on prices in bordering states than standard errors to choice of the tuning or may seem arbitrary. As, on average, districts bandwidth parameter (i.e., the degree of kernel are 57 kilometers apart (centroid-to-centroid), smoothing).24 at r = 100 kilometers, the sets BSTATEd r and Both numerator, Δwd = wd,09 − wd,04 , and STATEd differ only for districts relatively close denominator, ψ d , of the dependent variable in to their state’s border with another Indian state. Equation (7) are district-level summary statistics Indeed, IV 3100 d = IV 2d for half of the 462 dis- derived from micro-data. This gives rise to a tricts in my estimation sample (those in the deep particular form of heteroskedasticity and renders interior of states or along the coasts or interna- least-squares estimation inefficient. The standard tional borders). By contrast, IV 3200 d = IV 2d for solution is to use weighted least-squares, taking fewer than 10% of sample districts. This suggests the inverse of the estimated sampling variances a strategy of comparing alternative estimates of as weights. ( While ( ) the sampling ) variance of Δwd γ from Equation (7) based on IV 3r d with succes- is σ2 wd,09 + σ2 wd,04 (see above), there is sively higher values of r to determine at what no equally straightforward “plug-in” estimate point increasing the radius of influence ceases of the sampling variance of ψ d . I, therefore, to matter. bootstrap this variance as well by drawing Finally, as Equation (10) makes evident, dif- 1000 random samples of individuals from each ferences in price trends across crops is key to district’s original sample and computing ψ d identification; if the ΔpBSTATEr ,j are the same for repeatedly. From these two components, then, I d all j, then IV 3rd collapses to ΔpBSTATEr , essen- obtain the sampling variance of Δwd /ψ d using d tially a constant. Given the inclusion of the con- the delta-method.25 stant term c, γ is virtually nonidentified in this scenario. Equally as important is variation in F. Estimation Results crop composition across districts (see Table 2). If ad,j = aj for all d, then even if the ΔpBSTATEr ,j Estimates of γ based on Equation (7) are d reported in Table 2A, in which identifying are not all equal, IV 3rd again essentially collapses to a constant. The adjusted R2 s of the first-stage 23. Also note that with only a single (5-year difference) regressions using IV 1d , IV 2d , IV 3100 d , and IV 3200 d observation per district, serial correlation is not an issue in my are, respectively, 0.788, 0.121, 0.103, and 0.091. set-up. 24. Bandwidth here is the distance cutoff, in degrees of E. Inference lat/long, beyond which spatial dependence is assumed to die out. Based on simulation evidence from Bester et al. (2014), I As already alluded to, the error term εd is choose a bandwidth of 16; i.e., given the area of my “sampling region” (the 18 major states of India), this choice should likely to be correlated across neighboring dis- yield minimal test-size distortion across a range of possible tricts, if only because geographically proximate spatial correlations. I find these p-values to be highly robust regions experience similar productivity shocks to bandwidth deviations of at least ± 4. over time. I use a nonparametric covariance 25. Although this procedure ignores correlation between numerator and denominator arising from the fact that these matrix estimator or spatial HAC (Conley 1999) to two statistics are calculated from partially overlapping sam- account for heteroskedasticity and spatial depen- ples of the same underlying micro-data, it should serve ade- dence. A familiar alternative to the spatial HAC is quately as a first approximation. JACOBY: FOOD PRICES, WAGES, AND WELFARE IN RURAL INDIA 11 assumptions become progressively less restric- G. Robustness: NREGA tive across columns. Thus, column 1 estimates India’s National Employment Rural Guaran- are by ordinary (weighted) least squares, column tee Act (NREGA) is meant to provide every 2 uses IV 1d as an instrument, column 3 uses rural household with 100 days of manual labor IV 2d , column 4 uses IV 3100d , and column 5 uses at a state-level minimum wage, which is typi- IV 3200 d . Instrument diagnostics are problem- cally above the market wage. Imbert and Papp atic given the spatial error structure discussed (2012), using NSS-EUS data and exploiting the above. However, for lack of a better alternative, gradual phase-in of the program since 2006, I report Cragg-Donald F -stats, which assume find that NREGA increased overall public works i.i.d. errors, in Table 2 for all IV regressions. The employment while (modestly) raising private- critical value for the associated weak instrument sector wages in rural India. As these labor mar- test, based on 10% maximal size for a 5% Wald ket changes were contemporaneous with rising test, is 16.4 in all cases (Stock and Yogo 2002). food prices, they are worth taking seriously as Hence, subject to the caveat already noted, I possible confounding factors. Given my estima- can strongly reject weak identification, even tion strategy, however, NREGA will only affect using IV 3200 d . the results insofar as the local expansion of the While a comparison of the first two columns program was systematically related to the (instru- suggests that measurement error in crop shares mented) change in the agricultural price index. leads to a modicum of attenuation bias, even Based on 7-day employment recall informa- the column 2 estimate is well below unity as tion in the NSS-EUS, I compute the population indicated by the p-values from the bootstrapped- weighted district average days spent in public based t-test of H 0 : γ = 1.26 Relaxing the works employment (both NREGA and other) for assumption of no measurement error or simul- rounds 61 and 66.27 Including the 2004–2009 taneity bias in price changes in columns 3–5 change in this public works employment vari- delivers a ̂ γ much closer to unity, albeit one able (ΔPW ) in regression (Equation (7)) results much less precisely estimated. The specifica- in no appreciable changes in my estimates of tions in columns 4 and 5, however, which allow γ (compare columns 5 and 6 of Table 2). Of shocks to be correlated across state borders, course, the coefficient on ΔPW does not neces- do not give much different results from that sarily reflect the causal impact of NREGA or any of column 3, which ignores such correlation. other public works employment program in India The pattern of coefficients across columns on rural wages; this specification merely serves suggests a rough balance between measure- as a robustness check. ment error in prices (attenuation bias) and simultaneity bias. H. Sectoral Labor Mobility None of the p-values for H 0 : γ = 1 in columns 3–5 are anywhere near rejection levels, evidence My framework assumes perfect mobility of in favor of the specific-factors model. To assess labor across production sectors over the rele- power, I use the bootstrapped t-distribution to vant horizon. However, as noted above, Topalova answer the question: How likely would I have (2010) proposes an alternative specific-factors been to reject H 0 : γ = 1 had the true γ been model to rationalize her empirical results for at or very near zero? Based on this empirical India in which labor is perfectly immobile, but power functions, at a true γ of zero, H 0 : γ = 1 capital moves freely, across sectors. It is easy to would be rejected with 95% certainty in the col- see that, in this set-up, agricultural wages respond umn 3 specification, and with closer to 90% cer- positively to an increase in food prices but nona- tainty in the column 5 specification. In this sense, gricultural wages respond negatively, as capital is then, power is reasonably good: The evidence reallocated away from the sector whose terms of does not support the view that rural wages are trade have deteriorated and toward agriculture. unresponsive to agricultural price changes over a To test perfect intersectoral mobility of labor, half-decade period. I use the same procedure just employed to 27. This is essentially the same variable considered 26. The p-value is the proportion of times the boot- by Imbert and Papp (2012). In 2004–2005, public-works strapped, re-centered, t-statistic of Bester et al. (2014) employment accounted for just 0.22% of a day of work on exceeds the conventional t-statistic for the null in ques- average, increasing to a still minuscule 1.44% of a day in tion computed for the original sample. I use 10,000 2009–2010. Note, however, that NREGA employment is con- bootstrap replications. centrated in the agricultural off-season. 12 ECONOMIC INQUIRY construct log-wage district fixed effects for the of nontradables. There are also several differ- 2004–2005 and 2009–2010 NSS-EUS rounds, ences between Equation (11) and the compen- except in this case using only wage data for sating variation formula used by Porto (2006), nonagricultural jobs. The dependent variable is and earlier by Ravallion (1990). First, Ω allows again the time difference of these district fixed not just for changes in labor earnings but for effects scaled by ψ d . Relative to the previous changes in capital (land) income, which is obvi- analysis, 17 districts are dropped for lack of data ously critical in my setting. Second, whereas the on nonagricultural wage jobs. The estimates, in λs vary by household, as in Porto’s application, panel (B) of Table 2, differ little from their coun- the elasticities δ and ψ vary in my case by the terparts in panel (A), nor can I reject H 0 : γ = 1 sectoral composition of the district labor market. in the specifications with the least restrictive Moreover, rather than plugging in reduced-form identifying assumptions. Hence, it appears that econometric estimates of these elasticities (which nonagricultural wages, contra Topalova’s impli- are infeasible for reasons already discussed), I cation, respond as positively to higher food compute them based on an empirically validated prices as do wages overall. Consequently, the theoretical model. resulting welfare gains accruing to manual labor- In what follows, I consider the distributional ers (through wages) should not depend on the consequences of a uniform percentage increase in sector in which they happen to be employed. all agricultural commodity prices relative to the price of manufactures, the numeraire. According to Equation (11), the corresponding household IV. FOOD PRICES AND WELFARE welfare elasticity is simply ε = Ω − νA , where νA is the expenditure share of food crops. A. Welfare Elasticities Now consider a rural household embedded B. Distributional Analysis within the economy sketched out in Section II. Its contribution to aggregate income consists of The India Human Development Survey value-added from its enterprises, both farm and (IHDS) of 2005 is a nationally representative nonfarm, its net earnings from manual labor, and household survey of both rural and urban India its exogenous income E. The second of these (Desai, Vanneman, and National Council of components, which I will denote by W (LS − LD ), Applied Research 2008). Within the 18 major is not present in Equation (3) because manual states already discussed, the IHDS covers nearly labor supply (LS ) and demand (LD ) are equal in 24 thousand rural households spread over 254 the aggregate. districts, collecting information on consumption Household indirect utility is a function of expenditures and income, including revenues income and prices, PM , PS , and Pj , j = 1,..., c. Fol- and costs from household enterprises, both agri- lowing the conventional derivation, the propor- cultural and nonagricultural. Figure 2 shows the tional change in money-metric utility m is patterns of λA and λS smoothed across percentiles of per-capita expenditures, as represented by the ∑( ) (11) m̂= Ω s j − νj P̂j IHDS rural sample. Relative to total house- j hold income, gross revenues from both farming and service enterprises increase by percentile, where Ω = λA + (λS − νS )δ + λL ψ , υj is the expen- though the former increases much faster. By diture share of good j (S in the case of services), contrast, because the demand for hired labor λA = PA Y A /y is the ratio of gross farm revenue across household enterprises increases with to income, λS = PS Y S /y is the ratio of gross rev- wealth, λL decreases and essentially goes to zero enue from service enterprises to income, and for the highest percentile. On the consumption λL = [W (LS − LD )]/y is the ratio of the net earn- side (Figure 3), the behavior of the food share ings of manual labor to income. The term Ωsj − νj is familiar, falling steadily and quite rapidly by is reminiscent of Deaton’s (1989) well-known percentile, whereas the share of expenditures on net consumption ratio (revenue minus expendi- nontraded goods has the opposite, though a less tures on crop j divided by total consumption steep, distributional gradient.28 expenditures) except that, unlike Deaton’s par- tial equilibrium result, it fully accounts for the 28. Nontraded goods expenditure categories include: fire- changes in factor income induced by a given wood, entertainment, conveyance, house rental, repair and price change, as well as for changes in the price maintenance, medical care, education, and other services. JACOBY: FOOD PRICES, WAGES, AND WELFARE IN RURAL INDIA 13 FIGURE 2 λA , λS , and λL by Percentile .8 proportion of total income .6 .4 .2 0 0 20 40 60 80 100 per capita consumption percentile Gross crop revenue Gross service enterprise rev. Net earnings manual labor Turn now to the main results in Figure 4, show- beneficial adjustment in rural wages, the poorest ing the relationship between the welfare elastic- rural households in India would experience a ity with respect to food prices, ε, and per capita welfare loss of around 0.2% for a 1% uniform expenditure percentile. Observe that ε is positive increase in agricultural prices. However, the across the income spectrum, never falling below relative advantage of the general equilibrium sce- 0.4. Thus, higher food prices confer substantial nario erodes rapidly with income as manual labor and broad-based benefits to the rural population earnings become progressively less important in of India, although the pattern of proportional the higher percentiles. Indeed, because in partial welfare gains is mildly hump-shaped, with the equilibrium, the richest households do not have poorest and richest households gaining least. This to pay higher prices for services or higher wages latter feature is driven by changes in non-traded to hired labor, they would benefit even more than goods prices and the relatively large share of in general equilibrium from higher food prices. expenditures devoted to these goods by the rich. In other words, if δ is artificially set to zero, then ε would be essentially flat across the top per capita V. CONCLUSIONS expenditure quintile.29 Finally, let us compare the general equilibrium In reaction to the food price spike of welfare analysis to a more conventional partial 2007–2008, the Government of India imposed equilibrium one. Of course, the latter assumes export bans on certain major crops. Such efforts that ψ = δ = 0 so that, from Equation (11), to restrain consumer prices can have the unfortu- Ω = λA . The distribution of partial equilibrium nate side-effect of restraining producer prices as welfare elasticities looks dramatically different well. My analysis shows that, in the face of higher than that of ε (Figure 4). Without the large and agricultural commodity prices, a stand-alone export ban, or any policy that mimics its effects, 29. In Jacoby (2013), I further account for India’s vast would reduce welfare for the vast bulk of India’s Public Distribution System (PDS), under which eligible population. Moreover, it is precisely the poorest households (generally, those below the poverty line) can pur- rural households (and, hence, the poorest in India chase fixed rations of either rice, wheat, or sugar in “Fair Price Shops” at below-market prices. If I assume that PDS prices as a whole) that are most harmed by forestalling, remain stable even as market prices rise, I obtain modestly or at least delaying, the substantial trickle-down better welfare outcomes for all but the top two deciles. effects of higher crop prices via rural wages. 14 ECONOMIC INQUIRY FIGURE 3 .6 .5 .4 .3 .2 .1 Expenditure Shares by Percentile 0 20 40 60 80 100 per capita consumption percentile food crops non-traded goods FIGURE 4 Welfare Elasticities by Percentile .6 welfare elasticity w.r.t. food price 0 .2–.2 .4 0 20 40 60 80 100 per capita consumption percentile General equilibrium Partial equilibrium JACOBY: FOOD PRICES, WAGES, AND WELFARE IN RURAL INDIA 15 ∑ Partial equilibrium analysis, which assumes where ωj = (1 + αKj /αLj )βj /(1 + i βi αKi /αLi ). fixed wages, provides a highly misleading picture On the supply side, from the services production function and the specificity of capital, we have: of the distributional impacts of food price shocks among India’s vast rural population. To be sure, (A4) LS + αIŜ ̂S = αLS ̂ Y IS . the story may be quite different in metropolitan Meanwhile, the condition that input prices equal respec- India, where the poor, arguably, benefit little from tive marginal value products delivers W ̂ =P̂S + F̂L = P̂S − rising rural wages.30 Even though not much more ̂ ̂S and P̂ S = −F ̂I = ̂ ̂ S LS + Y I S − YS , where the second equal- than a quarter of India’s population resides in S ity in each case follows from the total differentiation of the cities, urban constituencies are obviously more marginal product functions FLS and FIS . Solving these two concentrated than rural ones and, hence, from equations, after first substituting out Î S from the second using a political-economy standpoint, are likely to be Equation (A4), yields more pivotal in shaping government policy on αLS + αIS α (A5) ̂S = Y ̂. ̂S − LS W P such matters as food security. αKS αKS Finally, this study speaks to the broader debate Substituting Equation (A2) into Equation (A5), equating on the link between trade and poverty. Consistent the result to Equation (A3), and solving gives: with the WTO’s Doha agenda, my results imply that lowering barriers to trade in agricultural αKS (1 − E∕y) ωA D + αLS βA ∕αKA (A6) δ= ( ) . goods on the part of developed countries, if only D 1 − αKS (1 − E∕y) ωS − αLS βS ∕αKS by improving the lot of the rural poor in India, With equal input cost shares, Equation (A6) can make a significant dent in global poverty. simplifies to δ = RβA /(αK + αL − RβS ) where R = αL + αK (αK + αL )(1 − E/y). Finally, as mentioned in the text, E = 0 and αI = 0 ⇒ R = 1 ⇒ δ = βA /(1 − βS ). APPENDIX MODEL SOLUTION PARAMETERS COMPUTED FROM NSS DATA I assume Cobb–Douglas production functions with input cost shares αLi + αIi + αKi = 1 in each sector i = A, M , S. The Input Cost Shares in Agriculture first step is to solve the following system of four equations: The 59th round of the National Sample Survey (NSS59) ̂ + αKA Π αLA W ̂A ̂A = P collected nationally representative farm household data in 2002–2003, including information on agricultural inputs ̂ + αKM Π αLM W ̂M = 0 and outputs for over ∑ 40,000 farms. The labor cost share is αL = W ( h + f )/ j Pj Y j , where h and f are, respectively, ̂ + αKS Π αLS W ̂S ̂S = P hired and family labor in agriculture and the denominator is the value of crop production. We may write the numer- (A1) ̂ A + βM Π βA Π ̂ M + βS Π ̂ ̂S = W ator as W h (1 + f ), where f = f / h is the ratio of family to hired labor. For a labor market in equilibrium, f should for W ˆ and Π ̂M = P ̂ i (recall, P ̂I = 0 by assumption). The equal the ratio of the number of agricultural laborers work- first three equations are the sectoral price-equals-unit-cost ing on their own farm to the number working for wages on conditions, whereas the last∑equation is derived from the labor other farms. Thus, we can calculate f for each of the five constraint (which implies βi ̂ Li = 0) and the fact that ̂ Li = regions (north, northwest, center, east, and south) from indi- i vidual employment data in NSS61-EUS. Comparable data ̂i − W Π ̂ in the Cobb–Douglas case. on hired labor expenses (for regular and casual farm work- The solution for the wage-price elasticity is given as ers), W h , and on total value of crop production are available follows: at the farm-level by season from NSS59. Summing up W h ( ) (A2) Ŵ ∕P̂A = βA ∕αKA + βS δ∕αKS ∕D, across seasons and households within each region (using sam- ∑ pling weights) multiplying by (1 + f ) and dividing by a sim- where D = 1 + i βi αLi /αKi . In the case of equal input cost ilarly computed sum of production value gives the regional shares across sectors, D = 1 + αL /αK and Equation (A2) labor shares. I use the ∑ same approach for the intermediate reduces to Equation (5) in the text. input shares αI = PI I A / j Pj Y j , where the numerator is the Solving for the elasticity of the services sector price with total expenditures on non-labor variable inputs as reported in respect to the agricultural sector price, δ, involves equating NSS59 (seed, fertilizer, pesticide, and irrigation). The results changes in service sector supply Y ˆ S and demand X ̂S . If the of these calculations are as follows: Marshallian demand function for services takes the form X S = ηy/PS (i.e., Cobb–Douglas preferences), where η is a share parameter, then TABLE A1 ( ) Estimated Input Shares (A3) ̂S = ̂ X y−P ̂S = (1 − E∕y) ωA P ̂S − P ̂A + ωS P ̂S North Northwest Center East South αL 0.331 0.304 0.258 0.317 0.260 30. A full analysis of rural–urban labor market linkages αI 0.264 0.325 0.258 0.250 0.238 is beyond the scope of this study, but is an important topic for future research. 16 ECONOMIC INQUIRY TABLE A2 Summary Statistics for Major States of India Annual No. of PC Expend. A S M Districts North Haryana 4.559 1.177 0.785 0.142 0.073 19 (0.817) (0.136) (0.138) (0.111) (0.081) Himachal Pradesh 4.094 1.161 0.772 0.165 0.063 12 (0.551) (0.112) (0.120) (0.095) (0.049) Punjab 4.535 1.145 0.731 0.203 0.067 17 (0.891) (1.145) (0.160) (0.135) (0.041) Uttaranchal 3.296 1.177 0.761 0.187 0.052 13 (0.474) (0.163) (0.180) (0.147) (0.058) Uttar Pradesh 3.108 1.182 0.781 0.149 0.070 70 (0.596) (0.127) (0.124) (0.090) (0.068) Northwest Gujarat 3.136 1.316 0.835 0.088 0.078 25 (0.579) (0.169) (0.136) (0.084) (0.093) Rajasthan 3.317 1.266 0.758 0.168 0.075 31 (0.503) (0.103) (0.091) (0.073) (0.057) Center Chhattisgarh 2.244 1.253 0.870 0.092 0.038 13 (0.481) (0.085) (0.113) (0.093) (0.041) Madhya Pradesh 2.489 1.249 0.860 0.104 0.035 45 (0.608) (0.092) (0.117) (0.103) (0.050) Maharashtra 2.752 1.204 0.825 0.118 0.057 33 (0.558) (0.137) (0.120) (0.071) (0.058) Orissa 1.964 1.131 0.759 0.151 0.090 30 (0.557) (0.135) (0.137) (0.114) (0.076) East Bihar 2.408 1.183 0.802 0.142 0.055 37 (0.391) (0.155) (0.167) (0.126) (0.068) Jharkhand 2.257 1.040 0.697 0.210 0.093 18 (0.441) (0.282) (0.243) (0.210) (0.073) West Bengal 2.667 0.951 0.603 0.223 0.174 17 (0.363) (0.189) (0.139) (0.075) (0.113) South Andhra Pradesh 2.486 1.073 0.717 0.174 0.109 22 (0.308) (0.081) (0.090) (0.099) (0.052) Karnataka 2.595 1.159 0.828 0.094 0.078 27 (0.593) (0.208) (0.176) (0.081) (0.117) Kerala 4.355 0.686 0.370 0.458 0.172 14 (0.877) (0.289) (0.223) (0.172) (0.096) Tamil Nadu 2.386 0.891 0.588 0.225 0.187 29 (0.369) (0.228) (0.174) (0.114) (0.139) Notes: Means (standard deviations) of district-level data. Annual per capita expenditures are in thousands of 2004 Rupees. Sectoral Labor Shares week) by sector, Dd,m , m = MA (manual ag. labor), MNA (manual nonag. labor), and MNAS (manual nonag. labor Despite being a so-called “thin” round, NSS64, col- in services). lected in 2007–2008, fielded the standard Employment– There is a persistent daily wage gap between agriculture Unemployment Survey questionnaire on a “thick”-round and nonagriculture, present across all NSS-EUS rounds, sample of nearly 80,000 rural households. I use these data which suggests that days spent in agriculture are substantially to compute district-level sectoral labor shares at roughly the less productive than those spent in nonagriculture. In partic- mid-point between 2004–2005 and 2009–2010. As the sur- ular, an agricultural sector dummy included in a log-wage vey was carried out throughout the whole year in most dis- regression using the NSS64 rural sample attracts a coefficient tricts, agricultural labor seasonality is not a major issue at of −0.243, after controlling flexibly for gender, age, edu- the district level. For each individual, I compute the total cation, and district. Thus, labor productivity is around 24% manual labor days in the last week in both agricultural and lower per day in agriculture. To account for this productivity nonagricultural jobs, apportioning the latter (based on indus- difference, I incorporate an efficiency units assumption into try codes) between services and manufacturing sectors. I the model. In other words, the labor constraint becomes then take a population-weighted sum of days across indi- L = LA′ + L + L , where L′ = L e−0.243 . The district-level M S A A viduals in each district to get total district labor days (per sectoral labor shares, in efficiency units, can hence be JACOBY: FOOD PRICES, WAGES, AND WELFARE IN RURAL INDIA 17 calculated using Imbert, C., and J. Papp. “Equilibrium Distributional Impacts of Government Employment Programs: Evidence from e−0.243 Dd,MA India’s Employment Guarantee.” Working Paper No. βd,A = e−0.243 Dd,MA + Dd,MNA 2012-14, Paris School of Economics. 2012. Ivanic, M., and W. Martin. “Implications of Higher Global Dd,MNAS Food Prices for Poverty in Low-Income Countries.” βd,S = . Agricultural Economics, 39, 2008, 405–16. e−0.243 Dd,MA + Dd,MNA Ivanic, M., W. Martin, and H. Zaman. “Estimating the Short- Descriptive statistics for sectoral labor shares and other Run Poverty Impacts of the 2010–11 Surge in Food key variables are shown in Table A2. Prices.” Policy Research Working Paper No. WPS5633. Washington, DC: World Bank, 2012. Jacoby, H. “Food Prices, Wages and Welfare in Rural India.” REFERENCES Policy Research Working Paper No. 6412 (April). Ahmed, S., and H. G. P. Jansen. 2010. Managing Food Price Washington, DC: World Bank, 2013. Inflation in South Asia. Washington, DC: World Bank. Jayachandran, S. “Selling Labor Low: Wage Responses to Anderson, K., ed. Distortions to Agricultural Incentives: Productivity Shocks in Developing Countries.” Journal A Global Perspective, 1955–2007 . Washington, DC: of Political Economy, 114(3), 2006, 538–75. World Bank Publications, 2009. Jones, R. “Income Distribution and Effective Protection in a Atkin, D. “Trade, Tastes and Nutrition in India.” Yale Multicommodity Trade Model.” Journal of Economic University Economic Growth Center Discussion Paper Theory, 2, 1975, 1–15. No. 986, 2010. Kovak, B. “Regional Labor Market Effects of Trade Policy: Attanasio, O., V. Di Maro, V. Lechene, and D. Phillips. Evidence from Brazilian Liberalization.” Research “Welfare Consequences of Food Prices Increases: Evi- Seminar in International Economics, University of dence from Rural Mexico.” Journal of Development Michigan Working Paper No. 605, 2010. Economics, 104(September), 2013, 136–51. . “Regional Effects of Trade Reform: What is the Cor- Banks, J., R. Blundell, and A. Lewbel. “Tax Reform and Wel- rect Measure of Liberalization.” American Economic fare Measurement: Do We Need Demand System Esti- Review, 103(5), 2013, 1960–76. mation?” The Economic Journal, 106, 1996, 1227–41. Lanjouw, P., and R. Murgai. “Poverty Decline, Agricultural Bester, A., T. Conley, C. Hansen, and T. Vogelsang. “Fixed-b Wages, and Nonfarm Employment in Rural India: Asymptotics for Spatially Dependent Robust Nonpara- 1983–2004.” Agricultural Economics, 40, 2009, metric Covariance Matrix Estimators.” Econometric 243–63. Theory, 2014, in press. Lasco, C., R. Myers, and R. Bernsten. “Dynamics of Rice Boyce, J., and M. Ravallion. “A Dynamic Econometric Model Prices and Agricultural Wages in the Philippines.” Agri- of Agricultural Wage Determination in Bangladesh.” cultural Economics, 38, 2008, 339–48. Oxford Bulletin of Economics and Statistics, 53(4), Marchand, B. “Tariff Pass-Through and the Distributional 1991, 361–76. Effects of Trade Liberalization.” Journal of Develop- Conley, T. “GMM Estimation with Cross Sectional Depen- ment Economics, 99, 2012, 265–81. dence.” Journal of Econometrics, 92, 1999, 1–45. McCaig, B. “Exporting Out of Poverty: Provincial Poverty in Datt, G., and M. Ravallion. “Farm Productivity and Rural Vietnam and U.S. Market Access.” Journal of Interna- Poverty in India.” Journal of Development Studies, tional Economics, 85, 2011, 102–13. 34(4), 1998, 62–85. Narayanan, B., A. Aguiar, and R. McDougall, ed. Global De Janvry, A., and E. Sadoulet. The Impact of Rising Food Trade, Assistance, and Production: The GTAP 8 Data Prices on Household Welfare in India. Berkeley, CA: Base. Purdue, IN: Center for Global Trade Analysis, Institute for Research on Labor and Employment, 2009. Purdue University, 2012. Deaton, A. “Rice Prices and Income Distribution in Thailand: Nicita, A. “The Price Effect of Tariff Liberalization: Mea- A Non-Parametric Analysis.” The Economic Journal, suring the Impact on Household Welfare.” Journal of 99, 1989, 1–37. Development Economics, 89(1), 2009, 19–27. Desai, S., R. Vanneman, and National Council of Applied Parikh, J., and C. Singh. “Extension of MSP: Fiscal and Wel- Research. India Human Development Survey, 2005. fare Implications.” A Study for the Planning Commis- Ann Arbor, MI: Inter-university Consortium for Polit- sion [of India] by Integrated Action and Research for ical and Social Research, 2008. Development. New Delhi, India, 2007. Eswaran, M., A. Kotwal, B. Ramaswami, and W. Wadhwa. Porto, G. “Using Survey Data to Assess the Distributional “How Does Poverty Decline? Evidence from India, Effects of Trade Policy.” Journal of International Eco- 1983–1999.” Discussion Paper No. 07-05. Indian Sta- nomics, 70, 2006, 140–60. tistical Institute, New Delhi, India, 2007. Ravallion, M. “Rural Welfare Effects of Food Price Changes Goldberg, P., and N. Pavcnik. “Distributional Effects of Glob- under Induced Wage Responses: Theory and Evidence alization in Developing Countries.” Journal of Eco- for Bangladesh.” Oxford Economic Papers, 42, 1990, nomic Literature, XLV, 2007, 39–82. 574–85. Gulati, A. “Chapter 3: How to Create Economic Freedom for Sah, R., and J. Stiglitz. “Price Scissors and the Structure of the Agriculture: Can Agriculture Be Unshackled from Gov- Economy.” Quarterly Journal of Economics, 102(1), ernment Controls?, ” in Economic Freedom Ranking for 1987, 109–134. the States of India, 2012, edited by B. Debroy, L. Bhan- Sharma, V. “Dismantling Fertilizer Subsidies in India: Some dari, S. Aiyar, and A. Gulati. Washington, DC: The Cato Issues and Concerns for Farm Sector Growth.” Work- Institute, 2012. ing Paper No. 2012-09-01. Ahmadabad, India: Indian Haisken-DeNew, J., and C. Schmidt. “Interindustry and Inter- Institute of Management, 2012. region Differentials: Mechanics and Interpretation.” Singh, I., L. Squire, and J. Strauss, ed. Agricultural Household The Review of Economics and Statistics, 79(3), 1997, Models. Baltimore: Johns Hopkins University Press, 516–21. 1986. Himanshu, and A. Sen. “Why Not a Universal Food Security Stock, J., and M. Yogo. “Testing for Weak Instruments in Lin- Legislation?” Economic & Political Weekly, XLVI(12), ear IV Regression.” NBER Working Paper No. t0284, 2011, 38–47. Department of Economics, Harvard University, 2002. 18 ECONOMIC INQUIRY Strauss, J. “Joint Determination of Food Consumption Wodon, Q., C. Tsimpo, P. Backiny-Yetna, G. Joseph, F. and Production in Rural Sierra Leone: Estimates of Adoho, and H. Coulombe. “Potential Impact of Higher a Household-Firm Model.” Journal of Development Food Prices on Poverty: Summary Estimates for a Economics, 14(1), 1984, 77–103. Dozen West and Central African Countries.” Policy Topalova, P. “Trade Liberalization, Poverty, and Inequality: Research Working Paper No. 4745, Washington, DC: Evidence from Indian Districts,” in Globalization and World Bank, 2008. Poverty, edited by A. Harrison. Chicago: University of World Bank. Food Price Increases in South Asia: National Chicago Press, 2007. Responses and Regional Dimensions. Washington, DC: . “Factor Immobility and Regional Impacts of Trade South Asia Region, Agriculture and Rural Development Liberalization: Evidence on Poverty from India.” Amer- Unit, World Bank, 2010. ican Economic Journal: Applied Economics, 2, 2010, 1–41.