WPS6394 Policy Research Working Paper 6394 Rain, Agriculture, and Tariffs Paulo Bastos Odd Rune Straume Jaime A. Urrego The World Bank Development Research Group Trade and Integration Team March 2013 Policy Research Working Paper 6394 Abstract This paper examines whether and how rainfall shocks maker will respond to a rainfall shortage by reducing affect tariff setting in the agricultural sector. In a model import tariffs. These findings are robust to alternative of strategic trade policy, the authors show that the impact assumptions about market structure and the timing of of a negative rainfall shock on optimal import tariffs is the game. Using detailed panel data on applied tariffs and generally ambiguous, depending on the weight placed rainfall for 70 nations, the authors find robust evidence by the domestic policy maker on tariff revenue, profits that rainfall shortages generally induce policy makers to and the consumer surplus. The more weight placed on set lower tariffs on agricultural imports. domestic profits, the more likely it is that the policy This paper is a product of the Trade and Integration Team, Development Research Group. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at pbastos@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Rain, agriculture, and tari¤s Paulo Bastosy Odd Rune Straumez Jaime A. Urregox Keywords : Rainfall shocks; optimal tari¤s; strategic trade policy; agricul- ture. JEL: F1; L1; O1 We would like to thank the Editor, two anonymous referees, and Will Martin for very helpful comments. The views expressed herein are those of the authors and not of the institutions they are a¢ liated with. y Corresponding author. Development Research Group– Trade, The World Bank, 1818 H Street NW, Washington DC, United States. E-mail: pbastos@worldbank.org. Tel. +1 202 473 4332. z Department of Economics/NIPE, University of Minho, Portugal; and Department of Economics, Uni- versity of Bergen. E-mail: o.r.straume@eeg.uminho.pt x Research Department, Inter-American Development Bank, United States. E-mail: jaimeu@iadb.org 1 "...drought [in "rice countries"] is, perhaps, scarce ever so universal as neces- sarily to occasion a famine, if the government would allow a free trade."–Adam Smith (1776, IV.5.45) 1 Introduction Random ‡uctuations in weather may have severe adverse e¤ects on agricultural producers and consumers. How can policy makers mitigate these impacts? Recent empirical research suggests that greater openness to trade alleviates the adverse e¤ects of weather shocks on domestic consumers. Using district-level panel data for colonial India in 1875-1919, and exploiting the construction of the railroad network to identify the e¤ects of increased open- ness, Burgess and Donaldson (2011) document that the arrival of railroads dramatically constrained the ability of rainfall shortages to cause famine.1 But while this evidence points to the existence of a causal link between openness and weather-related famine, import tari¤s on agricultural goods remain high in many nations, and relatively little is known about the extent to which countries use trade policy strategically to mitigate the impacts of weather shocks on domestic welfare.2 In this paper, we examine whether and how rainfall shocks a¤ect tari¤ setting in the agricultural sector. To identify key mechanisms at play, we …rst set up a model of international oligopoly in which domestic and foreign agricultural producers compete in the home market. International trade is potentially costly due to import tari¤s optimally set by the domestic policy maker. A rainfall shortage increases the marginal costs of domestic producers, thus generating a shortfall in agricultural output that foreign producers have an incentive to meet. Consequently, the shock a¤ects the marginal e¤ects of import tari¤s on tari¤ revenue, domestic pro…ts and the consumer surplus, and thereby the optimal policy response. We …nd that the impact of a rainfall shock on optimal import tari¤s is not clear-cut. A rainfall shortage leads to a higher volume of agricultural imports and therefore higher marginal tari¤ revenues. All else equal, this increases the optimal tari¤. On the other hand, by making home production more costly, the shock reduces incentives for using tari¤s as an instrument to shift rents from foreign to domestic producers. This leads, ceteris paribus, to a lower optimal tari¤. Finally, a rainfall shortage also reduces domestic consumers’ surplus due to a lower total supply of the agricultural good. This increases (reduces) the negative e¤ect of import tari¤s on domestic consumers if food demand is 1 This evidence is consistent with Donaldson (2013), who shows that railroads contributed to reduce the exposure of agricultural prices and real incomes to rainfall shocks. 2 Gibson et al. (2001) emphasize that high protection for agricultural commodities in the form of tari¤s continues to be the major factor restricting world trade. 2 su¢ ciently (not too) convex, leading, all else equal, to lower (higher) tari¤s. The overall impact of the rainfall shock on optimal import tari¤s is, therefore, generally ambiguous, depending on the weight placed by the domestic policy maker on each of these policy objectives, and on the shape of the food demand function. A larger weight placed on domestic pro…ts will enlarge the scope for a rainfall shortage to cause tari¤ reductions. A stronger concern for domestic consumers may have a similar e¤ect, but only if demand is su¢ ciently convex, i.e., if the demand for the agricultural good becomes su¢ ciently inelastic at lower consumption levels. We proceed by verifying the extent to which the main predictions of our theory are sensitive to assumptions about the timing of the game and the mode of competition. In the basic model, we assume that a rainfall shortage increases the marginal costs of agricultural production, as the a¤ected producers must spend more resources on irrigation in order to achieve a given level of output. However, in some agricultural markets producers may have limited access to irrigation technology, and hence have limited ability to adjust output following the shock. We therefore modify the timing of the game, letting the policy maker, and subsequently the foreign producers, make their optimal decisions in response to an exogenous drop in domestic output. Reassuringly, we …nd that the results of the basic model remain qualitatively unchanged under this alternative setting, while also showing that a rainfall shortage is more likely to induce tari¤ cuts when producers have greater access to irrigation technology. In another extension to the basic model, we examine if and how our main results depend on the assumption of oligopolistic competition. Applying the theoretical framework of Bagwell and Staiger (2001), who analyze strategic trade policy under perfect competition, we re-examine the e¤ect of a rainfall shortage on optimal import tari¤s when …rms are competitive rather than oligopolistic. We …nd, once again, that the key predictions of our basic model remain qualitatively similar. We then examine empirically the impact of rainfall shocks on e¤ectively applied tari¤s. Using detailed panel data on applied tari¤s and rainfall for 70 nations over 1988-2006, we …nd that rainfall shortages generally lead to lower applied tari¤s on agricultural imports. The positive relationship between rainfall and import tari¤s is robust to several alternative empirical speci…cations, tends to be stronger in countries with greater access to irrigation, and applies fairly widely across the set of agricultural products and regions considered. The main exception refers to nations in Africa, for which we …nd no evidence of signi…cant tari¤ adjustment in response to rainfall shocks. This lack of response may plausibly re‡ect the fact that governments of African nations tend to depend more heavily on import tari¤s for revenue (Devarajan et al., 1999). In this case, tari¤ revenue is likely to carry a larger weight on the governments’objective function, and thereby o¤set opposite incentives via pro…ts and/or the consumer surplus. Overall, however, our empirical …ndings suggest 3 that policy makers around the world often adjust agricultural import tari¤s strategically in response to rainfall shocks. In addition to the work cited above, our paper is related to the theoretical and empirical literature on optimal tari¤s and strategic trade policy, including Bickerdike (1907), Graaf (1949–1950), Spencer and Brander (1983), Dixit (1984), Brander and Spencer (1985), Eaton and Grossman (1986), Grossman and Helpman (1994, 1995), Brander (1995), Gawande et al. (2000), Broda et al. (2008) and Bagwell and Staiger (2001, 2012a, b). We are not aware of previous research, either theoretical or empirical, focusing on the e¤ects of rainfall shocks on optimal tari¤s in the agricultural sector. Our work is also broadly related to the emerging literature on how weather shocks shape economic, social and po- litical outcomes (Deschênes and Greenstone, 2007; Maccini and Yang, 2009; Burgess et al., 2009; Brückner et al., 2011; Jones and Olken, 2010; Dell et al., 2012). The remainder of the paper is organized as follows. Section 2 develops a theoretical model in which rainfall shocks impact on the optimal agricultural tari¤s set by the domestic policy maker. Section 3 describes the empirical strategy, while section 4 presents the data employed in the empirical analysis. Section 5 presents the main results, before section 6 examines their robustness. Section 7 examines heterogeneity of e¤ects across products and countries. Section 8 concludes. 2 A model of weather shocks and optimal agricultural tari¤s Consider an oligopolistic domestic market for a homogeneous agricultural good that is supplied by n domestic and m foreign producers.3 Domestic demand for the agricultural good is given by the inverse demand function p=1 Q; (1) where n X m X Q= qi + bj q (2) i=1 j =1 bj being quantities supplied by the domestic is total supply of the good, with qi and q producer i and the foreign producer j , respectively.4 There is a constant marginal cost of 3 It is common to adopt oligopoly models to investigate international trade in agricultural markets. Early theoretical and empirical research in this literature includes Sarris and Freebairn (1983), Karp and McCalla (1983), Kolstad and Burris (1986), Paarlberg and Abbott (1986) and Pick and Park (1991). Krishna and Thursby (1992) note that agricultural trade is often conducted through marketing boards for the product in question, both in developed and developing countries. Building on this observation, they examine optimal tax/subsidy policies with export marketing boards that compete as oligopolists in world markets. 4 We adopt the simplifying assumption of linear demand in order to identify key mechanisms whereby a rainfall shortage shapes tari¤ setting incentives. Below we examine the implications of considering 4 production equal to c (b c) for domestic (foreign) producers. In addition, foreign producers must pay a per-unit tari¤ t for supplying the domestic market. This tari¤ is set by a domestic policy maker with the following objective function: n X =T + i + S; (3) i=1 where m X T =t bj q (4) j =1 is total tari¤ revenue, i = (p c) qi (5) is the pro…t of the domestic producer i, and 1 S = Q2 (6) 2 is domestic consumers’surplus. The speci…cation of is su¢ ciently general to encompass a variety of di¤erent policy objectives, where we allow the policy maker to place di¤erent weights on domestic pro…ts and consumers’surplus: 0 and 0, respectively.5 The pro…t of a foreign producer j is given by bj = (p b c bj : t) q (7) We consider the following two-stage game: Stage 1: The domestic policy maker sets the import tari¤ t. Stage 2: The domestic and foreign producers choose quantities simultaneously and non- cooperatively. We look for a subgame-perfect Nash equilibrium, solving the model by backwards induction. 2.1 Equilibrium supply of the agricultural good Each producer chooses its supply of the agricultural good to maximize its pro…ts. The …rst-order conditions for a domestic and foreign producer, respectively, are given by X m X @ i =1 2qi qk bj q c=0 (8) @qi k6=i j =1 alternative speci…cations for food demand, and discuss their empirical relevance. 5 Notice that these weights are relative to the weight on tari¤ revenues. For example, if < (>) 1, the policy maker places a smaller (larger) weight on domestic pro…ts than on tari¤ revenues. The case of > 1 is referred to by Bagwell and Staiger (2001) as a case where the government is in‡uenced by ‘political-economy considerations’(e.g., due to industry lobbying). 5 and X n X @ bj =1 bj 2q bs q qi b c t = 0: (9) @qbj s6=j i=1 bs = q bj = q Applying symmetry, qi = qk = q and q b, the Nash equilibrium output of domestic and foreign producers, respectively, are given by 1 (m + 1) c + m (b c + t) q= (10) m+n+1 and 1 (n + 1) (b c + t) + nc b= q : (11) m+n+1 Total output is therefore n (1 c) + m (1 b c t) Q= ; (12) m+n+1 which gives a domestic market price 1 + nc + m (b c + t) p= : (13) m+n+1 Pro…ts of domestic and foreign producers are given by b2 , respectively. = q 2 and b = q 2.2 Optimal import tari¤ At the …rst stage of the game, the domestic policy maker chooses the import tari¤, t, to maximize its objective function, given by (3). The optimal tari¤ is implicitly given by @ @T @ @S = + n + = 0: (14) @t @t @t @t The second-order condition, @2 2 (n + 1) (m + n + 1) (2 n + ) m = m < 0; (15) @t2 (m + n + 1)2 is satis…ed if 2 (n + 1) (m + n + 1) m < e := : (16) 2nm To ensure equilibrium existence with an interior solution in the production subgame, we assume that the sextuple (c; b c; n; m; ; ) belongs to a set , de…ned by := f(c; b c; n; m; ; ) jq (t ) > 0; b (t ) > 0; q < e: g (17) Solving (14) with respect to t, the explicit expression for the optimal import tari¤ is given by6 " # (1 b c + n (c b c)) (m + n + 1) + 2 n (1 c m (c b c)) (m (1 b c) + n (1 c)) t = : (18) 2 (n + 1) (m + n + 1) (2 n + ) m 6 The condition < e ensures that the denominator in (18) is positive. 6 The optimal tari¤ balances three di¤erent policy concerns: (i) raising tari¤ revenues, (ii) shifting oligopoly rents from foreign to domestic producers, and (iii) increasing con- sumers’ surplus. Notice that more rent shifting and an increase in consumers’ surplus are con‡icting policy targets. Thus, a larger weight on domestic pro…ts leads to a higher optimal tari¤, @t 2n (m + n + 1) = q (t ) > 0; (19) @ 2 (n + 1) (m + n + 1) (2 n + ) m while a larger weight on consumers’surplus leads to a lower optimal tari¤, @t (m + n + 1) = Q (t ) < 0: (20) @ 2 (n + 1) (m + n + 1) (2 n + ) m 2.3 Rainfall shortage A negative rainfall shock implies that the a¤ected producers must spend more resources on irrigation in order to achieve a given level of output, making agricultural production more expensive. We therefore model a domestic rainfall shortage as an increase in the marginal cost of domestic production, c. For simplicity, we assume that domestic and foreign producers have the same marginal production cost initially; i.e., we evaluate the domestic cost increase at c = b c. The e¤ect on the optimal tari¤ can then be summarized as follows:7 Proposition 1 Suppose that markets are oligopolistic and that domestic and foreign …rms have initially the same production costs. In this case, (i) a domestic rainfall shortage always leads to a higher import tari¤ if is su¢ ciently low, (ii) a domestic rainfall shortage leads to a lower import tari¤ if is su¢ ciently high and if m is su¢ ciently large relative to n, (iii) the higher is , the larger is the scope for a higher import tari¤ in response to a domestic rainfall shortage. Proof. (i) From (18) we have @t m+n+1+ 2 (m + 1) m+n+ +1 =n > 0 if < := : (21) @c 2 (n + 1) (m + n + 1) (2 n + ) m 2 (m + 1) (ii) From (21) we see that @t =@c < 0 if > . However, it remains to show whether this case is relevant, i.e., whether 2 for some parameter con…gurations. Setting c = b c and 7 We evaluate the e¤ect of a domestic rainfall shortage by considering a marginal increase in the domestic cost of agricultural production. However, it can easily be shown that the e¤ect on the optimal tari¤ is qualitatively identical if we instead consider a discrete increase in marginal production costs. 7 n(1+ )+1 inserting the optimal import tari¤ into (10)-(11), an interior solution exists if < 2n m+2(n+1) and < m . It is straightforward to verify that, if these two inequalities hold, then n(1+ )+1 the second-order condition < e is satis…ed. Furthermore, 2n > , implying 2 , n2 1 if m > n +1 . Part (iii) of the proposition follows directly from the de…nition of . For a given level of the import tari¤, a domestic rainfall shortage, by increasing the domestic cost of agricultural production, leads to lower market shares for domestic pro- ducers. Although some of the fall in domestic production is replaced by increased imports, there will also be a reduction in the total supply of the agricultural good to the domestic market. Thus, domestic producers as well as consumers are hurt by a negative rainfall shock. Since domestic producers (consumers) would su¤er (bene…t) from a lower import tari¤, it might seem somewhat counterintuitive that a lower import tari¤ can be an op- timal policy response to a rainfall shortage only if the policy maker places a su¢ ciently large weight on domestic pro…ts, and that the scope for such a policy response is larger the less weight the policy maker places on consumers’surplus. The intuition for this re- sult, though, can be found by considering how a domestic cost increase a¤ects the policy maker’s trade-o¤ among its three di¤erent policy targets. These are given by the three terms in (14), and we will consider each of the three di¤erent policy targets in turn. (i) The e¤ect of using import tari¤s to raise tax revenues is given by @T b @q n+1 b+ t =m q b =m q t : (22) @t @t m+n+1 b is An increase in domestic production costs leads to a higher import volume (since q increasing in c) and therefore higher marginal tari¤ revenues. All else equal, this increases the optimal tari¤. (ii) The e¤ect of using import tari¤s to shift oligopoly rents from foreign to domestic producers is given by @ @p @q 2m = q + (p c) = q: (23) @t @t @t m+n+1 Foreign producers will respond to a higher import tari¤ by reducing their supplies to the domestic market. This leads to a shift in oligopoly rents from foreign to domestic producers through two di¤erent channels, represented by the two terms in (23). The price increase that results from less imports makes domestic production more pro…table (term 1) and domestic producers will respond by expanding their output, which increases pro…ts in proportion to the price-cost margin (term 2). A domestic rainfall shortage reduces both of these e¤ects, since higher domestic production costs reduce both the output and the price-cost margin of domestic producers (notice that p c = q in equilibrium), which in turn reduces the marginal rent-shifting e¤ect of import tari¤s. In other words, import tari¤s become less e¤ective as a rent-shifting instrument if domestic producers become 8 more cost-disadvantaged. All else equal, this reduces the optimal tari¤ and, naturally, the e¤ect is stronger the larger weight the policy maker places on domestic pro…ts. (iii) The e¤ect of using import tari¤s to increase domestic consumers’surplus is given by @S @q b @q m =Q n +m = Q: (24) @t @t @t m+n+1 An increase in domestic production costs reduces total supply of food in the domestic mar- ket. If consumers’surplus is convex in output, which is true for linear demand functions, this means that the reduction in consumers’ surplus due to a marginal tari¤ increase is lower. All else equal, this increases the optimal tari¤ and, naturally, the e¤ect is stronger the larger weight the policy maker places on consumers’surplus. Thus, we can conclude that the policy maker will optimally respond to a domestic rainfall shortage by lowering import tari¤s if the second of the three above described e¤ects – the reduced e¤ectiveness of import tari¤s as a rent-shifting instrument – is su¢ ciently strong to outweigh the other two e¤ects. Otherwise, the import tari¤ will increase. The above analysis is based on the simplifying assumption of linear demand. Although the main mechanisms of the model, as given by the three di¤erent e¤ects described above, generalize well beyond the assumption of linear demand, it is worth considering the extent to which our results are robust to alternative demand assumptions. In particular, the result that a tari¤ increase has a larger negative impact on consumers at higher consumption levels relies on the assumption that consumers’surplus is convex in output. For a general inverse demand function p (Q), this requires that p0 (Q) + Qp00 (Q) < 0: In other words, consumers’ surplus is convex in output for concave, linear and ‘not-too- convex’ demand functions. However, since demand for food is likely to become quite inelastic for low consumption levels, consumers’ surplus might be concave in output for the range of output levels that are relevant for rainfall shortages that cause a serious cut-back in domestic production.8 If this is the case, a rainfall shortage will increase the negative e¤ect of import tari¤s on consumers’ surplus, thereby increasing the scope for tari¤ reductions as an optimal response to a negative rainfall shock. 8 In empirical studies, demand for food is generally found to be inelastic (Andreyeva et. al. 2010). O’Hare and Kammen (2008) argue that it seems reasonable to assume that demand for food becomes less and less elastic as agricultural output declines. As food consumption falls, consequences like malnutrition and starvation begin to appear, impacts much more compelling than the hedonic costs of consuming a less-preferred diet or wasting less food. Consistent with this hypothesis, Andreyeva et. al. (2010) survey a large body of evidence on the price elasticity of demand for food and conclude that demand is relatively more elastic for categories like "food away from home", "soft drinks" and "juice". 9 2.4 Extensions In this subsection we consider two di¤erent extensions to the model. First, we modify the timing of the game by considering the case where domestic producers are not able to adjust their output decisions in response to a rainfall shortage. Second, we abandon the oligopoly assumption and consider the case of perfectly competitive markets. 2.4.1 No domestic output adjustment In some agricultural markets, due to a lack of access to su¢ cient amounts of irrigated water, it might not be feasible to use irrigation in order to compensate for a rainfall shortage. If this is the case, domestic producers will have limited ability to adjust output in response to rainfall-induced tari¤ changes. We explore this scenario by modifying the timing of the game, letting the policy maker, and subsequently the foreign producers, make their optimal decisions in response to an exogenous drop in domestic production. P Let total domestic output be denoted by q := n i=1 qi . For a given tari¤ level, the best-response from foreign suppliers are given by (9). In the symmetric equilibrium, with b for all j = 1; :::; m, the output chosen by each foreign producer is given by bj = q q q t b c 1 b= q : (25) m+1 P b from (25) into (3), where n Substituting q i=1 i = (p c) q and S = 1 2 b)2 , and (q + mq maximizing with respect to t, the optimal import tari¤, as a function of an exogenous domestic output level, is given by (m + 1) (1 b c+( 1) q ) (q + m (1 b c)) t = : (26) 2 (m + 1) m 2(m+1) 9 The second-order condition of the maximization problem holds if < m . We also assume that q is su¢ ciently low to ensure an interior solution with strictly positive import volumes. We analyze the import tari¤ response to a domestic rainfall shortage by considering a marginal reduction in domestic production, q .10 Proposition 2 Suppose that markets are oligopolistic but domestic …rms are not able to adjust output in response to a rainfall shock. In this case, 9 The second-order condition is explicitly given by @2W (2 (m + 1) m ) = m < 0: @t2 (m + 1)2 10 Due to the linearity assumptions of the model, the tari¤ response is qualitatively identical if we instead consider a discrete fall in domestic production. 10 (i) a domestic rainfall shortage leads to a higher (lower) import tari¤ if is su¢ ciently low (high), (ii) the higher is , the larger is the scope for a higher import tari¤ in response to a domestic rainfall shortage. Proof. (i) From (26) we have @t ( 1) (m + 1) = < (>) 0 if < (>) 1 + : (27) @q 2 (m + 1) m m+1 An interior solution always exists for > 1 + m+1 if q is su¢ ciently low, which we assume to be the case. Part (ii) of the proposition follows directly from (27). Comparing Propositions 1 and 2, it is clear that the e¤ect of a domestic rainfall shortage on import tari¤s, in qualitative terms, does not depend on whether domestic producers are able to adjust their output in response to the rainfall shock or not. Notice also that the results in Proposition 2 hold for any initial level of q (that yields positive import volumes). The intuition behind these results follows closely the intuition given for Proposition 1. The e¤ects that go through the policy maker’s incentive for raising tari¤ revenues and increasing consumers’surplus are almost identical under the two di¤erent scenarios, as the equilibrium e¤ect of higher marginal production costs is similar to an exogenous cut-back in production. The incentives for using import tari¤s to shift oligopoly rents from foreign to domestic producers are also qualitatively similar, although the quantitative e¤ect is somewhat di¤erent when domestic producers cannot respond strategically to a change in import tari¤s. Formally, the e¤ect of a tari¤ increase on domestic pro…ts in the case of no domestic output adjustment, is given by P @( n i=1 i ) @p mq =q = : (28) @t @t m+1 A higher import tari¤ increases the domestic market price, which increases domestic prof- its. However, if domestic producers are not able to adjust their output in response to the tari¤ change, the positive pro…t e¤ect is smaller than it would have been if the domestic producers could meet the price increase by increasing their output. Thus, the incentive to use import tari¤s to increase domestic pro…ts is smaller in the absence of domestic output adjustment. Nevertheless, a drop in domestic production due to rainfall shortage will reduce this policy incentive even further, contributing, all else equal, to a lower import tari¤. As before, this e¤ect establishes a negative relationship between rainfall shortage and import tari¤s if the policy maker places a su¢ ciently large weight on domestic pro…ts in his objectives. 11 2.4.2 Perfect competition The oligopoly assumption used in the main version of the model may be less suitable for some agricultural markets, particularly in developing countries. We will therefore analyze if and how our main results might be a¤ected by the assumption of imperfect compe- tition. Suppose instead that the domestic market is supplied by competitive (domestic and foreign) …rms. As in Bagwell and Staiger (2001), we assume that the domestic and foreign industries are characterized by positively sloped supply functions and associated pro…t (producer surplus) functions. The domestic and foreign supply functions are given by, respectively, p c q (p) = (29) 2 and b p q b) = ; b (p (30) 2 b is the export price for foreign suppliers. where p is the domestic price, given by (1), and p The parameter c is a shift parameter that shifts down the domestic supply function in case of a rainfall shortage.11 The prices p and p b satisfy the arbitrage condition b+ t p=p (31) and the competitive equilibrium is characterized by the market-clearing condition b) : b (p Q = q (p) + q (32) From (1) and (29)-(32), the domestic price in the competitive equilibrium is given by 1 1 p= 1+ (c + t) : (33) 2 2 The domestic policy maker optimally sets the import tari¤ to maximize an objective function similar to (3), now given by =T + + S; (34) b) is tari¤ revenue, b (p where T = tq is domestic producer surplus, and S is consumers’ surplus, given by (6). Since @ =@p = q (p), domestic producer surplus is given by Z p p p (p) = q (s) ds = c : (35) 0 2 2 11 In contrast to the cost assumption in the oligopoly model, these supply functions imply decreasing returns to scale. This di¤erence is consistent with the often held notion that market structure (oligopoly versus perfect competition) is endogenously determined by technology. However, it is straightforward to show that the results from the oligopoly model would be qualitatively similar if we changed the cost functions to a linear-quadratic form C (q ) = cq + q 2 , with c being the cost-shifting parameter, which would correspond to a supply function given by (29). Details of these calculations are available upon request. 12 s maximization problem, it is straightforward to derive an Solving the policy maker’ explicit expression for the optimal import tari¤: 8 + 4c 4 + 2c + 2 3c t = : (36) 24 2 The second-order condition is @2W 1 = (24 2 ) < 0; (37) @t2 32 which is satis…ed if < 24 2 .12 As in the oligopoly model, optimal import tari¤ setting balances three di¤erent considerations: (i) raising tari¤ revenues, (ii) increasing domestic producer surplus, and (iii) increasing consumer surplus. Once more, (ii) and (iii) are con‡icting policy targets, as domestic producers bene…t from a higher import tari¤ while consumers bene…t from lower tari¤ level. Thus, @t 8q (t ) @t 8Q (t ) = > 0; = < 0: (38) @ 24 2 @ 24 2 A negative rainfall shock leads to a downward shift in the domestic supply curve. As before, due to the linearity of the model, the e¤ects on optimal tari¤ setting of a marginal increase in c and a discrete increase in c are identical. In line with our previous analysis, we also assume that domestic and foreign production costs are ex ante identical; i.e., we evaluate the e¤ect of the rainfall shock at c = 0. Proposition 3 Suppose that markets are perfectly competitive and that domestic and for- eign …rms have initially the same production costs. In this case, (i) a domestic rainfall shortage leads to a higher (lower) import tari¤ if is su¢ ciently low (high), (ii) the higher is , the larger is the scope for a higher import tari¤ in response to a domestic rainfall shortage. Proof. (i) From (36) we have @t 4+2 3 1 = > (<) 0 if < (>) (4 + 2 ) : (39) @c 24 2 3 1 We also need to show that an interior solution always exists for > 3 (4 + 2 ). For c = 0, an interior solution requires < 7 and < 3 + . The latter inequality also ensures that the second-order condition < 24 2 is satis…ed. Equilibrium existence for 1 1 > 3 (4 + 2 ) is con…rmed by noticing that 3 + > 3 (4 + 2 ) for all 0. Part (ii) of the proposition follows directly from (39). 12 14 17c In addition, an interior solution with positive domestic and foreign supply requires that < 2(1 c) 2 +3c 2c +6 and < 2(1 c) . 13 A comparison of Propositions 1 and 3 reveals that the e¤ect of domestic rainfall shortage on import tari¤s does not depend crucially on whether agricultural markets are oligopolistic or competitive. In either case, a su¢ ciently large weight on domestic producer surplus in the policy maker’s objective function is necessary for import tari¤s to fall in response to a negative rainfall shock. The decomposition of e¤ects shown in Section 2.2 also applies in the case of perfect competition. A downward shift in the domestic supply curve implies that some domestic production is replaced by imports but total supply to the market goes down. The increased import volume increases marginal tari¤ revenues, while the lower total supply implies that the reduction in consumer surplus due to a tari¤ increase is lower (since consumers’surplus is convex in output). Both of these e¤ects contribute to a higher import tari¤ in response to a domestic rainfall shortage. As before, the counteracting e¤ect is the concern for domestic producer surplus. A higher import tari¤ increases the domestic producer price, and the corresponding increase in domestic producer surplus is positively correlated with total domestic output. Techni- cally, this e¤ect is given by @ @ @p @p q 2 + t 3c = =q = = : (40) @t @p @t @t 4 32 A negative rainfall shock (i.e., an increase in c) reduces domestic output and therefore reduces the marginal pro…t gain of increasing the domestic price through higher import tari¤s. All else equal, this reduces the optimal tari¤, and, if this e¤ect is su¢ ciently strong (which requires a su¢ ciently high value of ), it will outweigh the two other e¤ects and enforce a negative equilibrium relationship between rainfall shortage and import tari¤s. 3 Empirical strategy To estimate the impact of rainfall shocks on applied import tari¤s we adopt the following econometric speci…cation: ln tari¤ict = + ln rainct + ln rainct agrici + i + c + t + ict ; (41) where tari¤ict is the mean of import tari¤s applied on product i by country c in year t; rainct is the amount of rainfall recorded in country c in year t; agrici is a dummy variable that takes the value of one for agricultural products; i is an industry …xed-e¤ect, c is a country …xed-e¤ect and t is a year …xed-e¤ect. ict is an idiosyncratic error term. Our main interest lies on , which reveals whether and how countries use trade policy strategically to mitigate the adverse e¤ects of rainfall shortages on the agricultural sector. Conditional on industry and year …xed-e¤ects, identi…cation of the causal e¤ect of interest 14 relies on the plausible assumption that within-country variation in rainfall over time is orthogonal to other determinants of import tari¤s. 4 Data Our empirical analysis makes use of the following sets of data: 1. Import tari¤ s We use annual data on bilateral applied tari¤s by country and product (SITC 2-dig., Rev. 4) from the TRAINS database of the United Nations Conference on Trade and Development. These data are generally available from 1988 onwards, but the extent of time coverage di¤ers considerably across countries. We construct both simple and import-weighted averages of bilateral applied tari¤s by country- product-year. For a given country-product pair, the former measures are de…ned as the simple average of applied tari¤s on imports from the various source countries in year t. The latter measures are de…ned as the weighted average of such bilateral import tari¤s, where the weights are the share of each source country in total country- product imports in a base period.13 Data on bilateral imports used to construct these weights come from the United Nations Commodity Trade Statistics Database (COMTRADE). To account for tari¤ binding, we will further make use of analogous information on bound tari¤ rates.14 These data also come from TRAINS, but are only available from 1995. 2. Rainfall and irrigation We use the data set on population-weighted rainfall levels by country-year constructed and made available online by Dell et al. (2012). The original source of the historical rainfall data is the Terrestrial Air Temperature and Precipitation: 1900-2006 Gridded Monthly Time Series (Version 1.01), compiled by Kenji Matsuura and Cort Willmott (2007) in conjunction with NASA. This database provides worldwide (terrestrial) monthly precipitation information at 0:5 0:5 degree resolution. Dell et al. (2012) use geospatial software to aggregate these rainfall data to the country-year level, weighting by the population distribution within the country.15 We further use country-level data on the proportion of agricultural land that is irrigated. These data come from the World Development Indicators of the World Bank, and are generally available for the period 2001-2009. We combine these sets of data to construct an unbalanced panel at the product-country- year level, covering 70 nations over the period 1988-2006. Table A.1 in the appendix 13 The base period is the …rst year for which tari¤ data are available for the tari¤-setting country. 14 The most-favored-nation tari¤ rate resulting from negotiations associated with GATT/WTO member- ship. 15 These data are also used by Jones and Olken (2010). 15 summarizes the data sources and variable de…nitions. To ensure adequate coverage for each country, we excluded information for nations that do not have at least three con- secutive years of tari¤ and rainfall data. Table A.2 lists the countries and corresponding time periods covered in our sample, while Table A.3 details the product classi…cation we employ. To identify agricultural products we use the UN statistics division of the SITC classi…cation. For robustness, we also consider a broader de…nition that includes as well other agricultural products.16 Irrigation data are available for 32 of the 70 countries ini- tially considered. These data generally refer to the 2001-2009 period, but the extent of time coverage di¤ers across countries. We therefore use the average value of this variable among non-missing observations for each of the 32 countries as a (time-invariant) proxy for access to irrigation at the country-level. Figure 1 provides illustrative evidence on the rainfall variable we employ for a subset of countries. The dashed lines in these diagrams refer to major, widely-documented droughts observed in each nation. Many of these extreme weather events were associated with the El Niño or La Niña phenomena, including Argentina (1999), Bolivia (1998), Brazil (1998, 2003), Colombia (1992, 1997), Chile (1998), Costa Rica (2001), Guatemala (1997, 2001) and Peru (2000). [Figure 1 about here] Our theory applies to countries that have discretion to set import tari¤s on agricultural goods. A potential concern is that, in reality, tari¤ setting might be constrained by limits associated with GATT/WTO membership. As shown in Figure 2, however, countries typically set their tari¤s on agricultural goods well below the bound tari¤s imposed by such multilateral tari¤ agreements; see Gibson et. al (2001, pp. 20-21) for further data and discussion. This implies that policy makers have ample policy space to adjust tari¤s strategically in response to rainfall shortages. We will nevertheless address this concern directly in the econometric analysis below. [Figure 2 about here] Table 1 reports summary statistics on the …nal panel data set employed in the regres- sion analysis. As can be seen, most observations refer to developing countries in Latin America and Caribbean, Asia-Paci…c and Africa.17 [Table 1 about here] 16 As shown in Table A.3, this latter de…nition includes as well SITC product categories "08", "09", "11", "12" and "29". 17 We adopt a broad de…nition of the term "Asia-Paci…c", comprising all nations in Asia and Oceania. 16 5 Main results Table 2 reports our baseline results. Columns (1) and (2) report the estimates obtained using a simple average of import tari¤s as the dependent variable, while columns (3) and (4) report results from using import-weighted tari¤s. By de…nition, the former measures give equal weights to all bilateral applied tari¤s within each country-product-year. They are therefore less sensitive to potential bias arising from the fact that a higher tari¤ im- posed on a given source country may lead to zero or little equilibrium imports from that nation. The latter measures have the advantage of placing a larger weight on tari¤s im- posed on countries for which import volumes of that product are larger to begin with. If such di¤erential import shares re‡ect heterogeneity in fundamentals across source coun- tries (e.g., due to stronger comparative advantage in agriculture), the policy maker may optimally favor larger adjustments to import tari¤s imposed on major food exporters, so as to obtain larger impacts from the policy response. It is therefore important to use both measures. [Table 2 about here] The estimates reported in columns (1) and (3) of Table 2 show no signi…cant e¤ects of rainfall shocks on overall import tari¤s. This result is not surprising, considering that rainfall shortages would not be expected to a¤ect tari¤ setting in the non-agricultural sector, and that agricultural goods represent a relatively small proportion of the full spec- trum of products. Columns (2) and (4) point, however, to signi…cant impacts of rainfall shocks on import tari¤s of agricultural products: the coe¢ cient of interest is positive and statistically signi…cant at the 1% level, suggesting that a rainfall shortage induces policy makers to set lower tari¤s on agricultural imports. When seen in conjunction with our theoretical model, this result suggests that policy makers place a considerable weight on domestic producer surplus when setting import tari¤s. 6 Robustness We conduct a number of checks to verify the robustness of our empirical …ndings. First, we examine the extent to which the results are sensitive to composition of our sample. To do this, we restrict the sample to include only countries that report at least four consecutive years of tari¤ and rainfall data. This restriction reduces the number of countries in the estimation sample to 51, down from the 70 initially considered. Reassuringly, the estimates yielded by this restricted sample remain very similar, both qualitatively and quantitatively (Table 3). [Table 3 about here] 17 We proceed by verifying the extent to which the results are sensitive to the de…nition of agricultural products we employ. In our baseline analysis, we rely on the UN statistics division of the SITC classi…cation. For robustness, in Table 4 we consider a broader de…nition that includes as well other agriculture-related products (Table A.3). The results reported in this table show that our …ndings remain very similar when this alternative de…nition is used. [Table 4 about here] Table 5 reports the estimates obtained when considering this broader de…nition of agricultural products and the restricted sample of countries that report at least four con- secutive years of tari¤ and rainfall data. Once again, the results remain very similar. [Table 5 about here] As mentioned above, a potential concern for our empirical analysis is that tari¤ setting might be constrained by limits associated with GATT/WTO membership. We adopt two alternative strategies to verify the extent to which our results are sensitive to tari¤ binding. First, we restrict the sample to include only country-product-year observations in which applied tari¤s are considerably smaller than the corresponding bound tari¤s. Second, we account for censoring by adopting a Tobit random e¤ects model on the full estimation sample. Table 6 reports the results yielded by the …rst of these strategies. While the analysis in the previous section spanned the 1988-2006 period, information on bound tari¤s is only available from 1995. Hence we begin by showing in column (1) that our baseline results remain qualitatively similar when we restrict the sample to include this sub-period only. In columns (2)-(4) we report results from analogous regressions, but now imposing progressively more stringent restrictions on the distance between applied and bound tari¤s. Reassuringly, we …nd that the coe¢ cient of interest remains remarkably stable across all the subsamples considered. [Table 6 about here] A potential concern with the results reported in Table 6 is that this approach intro- duces sample selection, which may a¤ect the estimated coe¢ cients directly. We therefore turn to the Tobit random e¤ects estimator. This method makes it possible to account for censoring due to tari¤ binding without introducing sample selection. A downside of the Tobit model, however, is that it is not possible to correct the standard errors for potential idiosyncratic correlation across products and over time within each country, implying that the standard errors are likely to be biased downwards. Table 7 reports the estimated mar- ginal e¤ects. Columns (1) and (3) refer to models with country random-e¤ects (including 18 also product and year dummies), while columns (2) and (4) refer to models with product …xed-e¤ects (including also country and year dummies). We …nd that the marginal e¤ects of interest are always remarkably similar to the OLS point estimates reported in column (1) of Table 6. Taken together, the results from these two strategies provide evidence that the potential bias imposed by tari¤ binding does not appear to be an important concern in our application. In the analysis thus far, we have adopted a linear function of rainfall. To account for potential nonlinearity in the policy response to extreme events such as drought, we modify out basic speci…cation by considering instead a set of dummy variables that are intended to capture extreme realizations of rainfall. In columns (1) and (4) of Table 8, we consider a dummy variable that takes the value of one when a realization of rainfall for a given country-year is below the 15th percentile of the rainfall distribution in the period 1988-2006.18 Consistent with the results reported earlier, the estimates indicate that severe rainfall shortages induce policy makers to set lower tari¤s on agricultural imports. In columns (2) and (5), we introduce a dummy variable that equals one when the level of rainfall for a given country-year is above the 85th percentile of this distribution. The results indicate that larger volumes of rainfall are associated with higher agricultural tari¤s, what is again consistent with the linear speci…cation.19 The results in columns (3) and (6) reveal that these conclusions remain unchanged when including both these dummy-variables and interaction terms together in the same regression. 7 Heterogeneity of e¤ects across products and countries We now examine the extent to which the impacts of rainfall shocks on agricultural tari¤s are heterogeneous across products and countries. Table 9 presents the results from re- gressions in which the rainfall variable is interacted with dummy variables for each of the agricultural products included in the UN statistics division of the SITC classi…cation (see Table A.3). The results in columns (1) and (3) suggest that a rainfall shortage leads to tari¤ reductions in most of the products considered. When considering both simple and weighted import tari¤s, we …nd positive and signi…cant e¤ects for all product categories, 18 We de…ne this variable based on the full country-level data on the log of rainfall for this period. We have considered alternative thresholds to de…ne the dummy variable and have obtained qualitatively similar results. 19 Extremely high realizations of rainfall in a short period of time can cause ‡oods. Floods may (or may not) reduce agricultural output in a given year. While a ‡ood typically reduces agricultural output in the months it occurs, it also tends to increase agricultural productivity in the following growing season, what has the opposite e¤ect (Banerjee, 2007). Based on annual data on country-level rainfall, it is however di¢ cult to separate ‡oods from consistently higher volumes of rainfall spread over the year. 19 with the exception of the categories "live animals" and "oilseeds and oleaginous fruits". In columns (2) and (4), we conduct a similar analysis, but including now interactions terms between log rainfall and individual dummy variables for each agricultural product considered in the broader de…nition de…ned earlier. [Table 9 about here] Table 10 reports separate estimates by region. Since our sample comprises a very small number of jurisdictions in North America and Europe, we report results for countries in Africa, Latin America and Caribbean and Asia-Paci…c. The estimates for the latter two regions are very similar to those obtained for the full sample. For Africa, however, we obtain insigni…cant coe¢ cients on the interaction term of interest.20 In the context of our model, how can we rationalize these insigni…cant results? One potential explanation is that agricultural producers in Africa tend to have less access to irrigation technologies. As discussed in section 2.4, for a given weight placed by domestic policy maker on domestic pro…ts, the inability to adjust output in response to a rainfall shortage will reduce the likelihood that such shock leads to a tari¤ reduction. Alternatively, policy makers in Africa may generally place a smaller weight on domestic pro…ts, perhaps because agricultural producers there tend to be relatively smaller and less organized collectively. On the other hand, as also discussed above, a negative rainfall shortage is more likely to induce tari¤ reductions when demand for food is more inelastic for low consumption levels. This would appear to be more likely the case in African countries, where food consumption tends to be closer to subsistence levels, making tari¤ reductions in response to a rainfall shortage more likely there. The lack of response revealed by the data may however re‡ect the fact that governments of African nations tend to depend more heavily on import tari¤s for revenue (Devarajan et al., 1999). In such a case, tari¤ revenue is likely to carry a larger weight on the governments’objective function, and thereby o¤set opposite incentives via pro…ts and/or the consumer surplus. [Table 10 about here] We proceed by examining more directly the predictions of our theory about the role of access to irrigation in shaping the e¤ects of a rainfall shock on agricultural import tari¤s. As noted above, our theory predicts that a greater ability to adjust output in response to a rainfall shortage increases the incentives to shift oligopoly rents from foreign to domestic 20 Nevertheless, there are reasons to remain cautious in interpreting these insigni…cant results as evidence that African nations do not lower import tari¤s strategically in response to rainfall shortages. First, although still sizable, sample size is considerably smaller for Africa than for the other two regions. Second, when using the main sample and weighted import tari¤s as the dependent variable, the magnitude of the coe¢ cient of interest is fairly similar across regions. 20 producers. All else being equal, greater access to irrigation would therefore be expected to magnify the positive relationship between rainfall and agricultural import tari¤s. To examine this prediction, we use country-level data on the proportion of total agricultural land that is irrigated. As mentioned above, these data are available for 32 of the 70 countries initially considered. Hence our estimation sample is reduced accordingly. Table 11 reports the corresponding results. In line with the predictions of our theoretical model, the estimates in columns (2) and (4) suggest that greater access to irrigation contributes to magnify the positive relationship between rainfall and agricultural import tari¤s. [Table 11 about here] 8 Concluding remarks We have examined theoretically and empirically the e¤ects of rainfall shocks on tari¤ setting in the agricultural sector. In a model of strategic trade policy, we have shown that the impact of a negative rainfall shock on optimal tari¤s is generally not clear-cut, depending crucially on the weight placed by the domestic policy maker on each of the various policy objectives, and on the shape of the food demand function. We have then estimated the impact of rainfall shocks on applied tari¤s. Using detailed panel data on applied tari¤s and rainfall for 70 nations, we have found robust evidence that rainfall shortages generally lead to lower tari¤s on agricultural imports. When interpreted in conjunction with our theory, this result suggests that policy makers place a considerable weight on domestic producer surplus when setting import tari¤s. Appendix [Tables A.1-A.3 about here] References [1] Andreyeva, T., Long, M.W. and Brownell, K.D. (2010). "The Impact of Food Prices on Consumption: A Systematic Review of Research on the Price Elasticity of Demand for Food." American Journal of Public Health, 100: 216-222. [2] Banerjee, L. (2007). "E¤ects of ‡ood on agricultural wage formation in Bangladesh: An Empirical Analysis." World Development, 35: 1989-2009. [3] Bagwell, K. and Staiger, R. (2001). "Strategic trade, competitive industries and agri- cultural trade disputes." Economics and Politics, 13: 113-128. 21 [4] Bagwell, K. and Staiger, R. (2012a). "The Economics of Trade Agreements in the Linear Cournot Delocation Model." Journal of International Economics, 88(1): 32- 46. [5] Bagwell, K. and Staiger, R. (2012b). "Pro…t Shifting and Trade Agreements in Im- perfectly Competitive Markets." International Economic Review, 53(4): 1067-1104. [6] Bickerdike, C. (1907). "Review of A. C. Pigou’s Protective and Preferential Import Duties." Economic Journal, 17: 98–108. [7] Brander, J. (1995), "Strategic Trade Policy." in G. Grossman and K. Rogo¤ (eds.): Handbook of International Economics, Amsterdam: North-Holland, 3: 1395-1455. [8] Brander, J. and Spencer, B. (1985). "Export subsidies and international market share rivalry." Journal of International Economics, 18: 83-100. [9] Brückner, M. and Ciccone, A. (2011). "Rain and the Democratic Window of Oppor- tunity." Econometrica, 79: 923-947. [10] Broda, C., Limão, N., and Weinstein, D. (2008). "Optimal Tari¤s and Market Power: The Evidence." American Economic Review, 98: 2032-2065. [11] Burgess, R. and Donaldson, D. (2011). "Can Openness Mitigate the E¤ects of Weather Fluctuations? Evidence from India’s Famine Era" American Economic Review, 100: 449-453. [12] Burgess, R., Dêschenes, O., Donaldson, D. and Greenstone, M. (2009). "Weather Related Death in India", MIT. [13] Deschênes, O. and Greenstone, M. (2007). "The Economic Impacts of Climate Change: Evidence from Agricultural Output and Random Fluctuations in Weather." American Economic Review, 97: 354-385. [14] Devarajan, S., Go, D.S., Li, G. (1999). "Quantifying the …scal e¤ects of trade reform: A general equilibrium model estimated for 60 countries." The World Bank, Policy Research Working Paper 2162. [15] Dixit, A. (1984). "International Trade Policy for Oligopolistic Industries." Economic Journal, 94: 1–16 [16] Dell, M., Jones, B., Olken, B. (2012). "Temperature Shocks and Economic Growth: Evidence from the Last Half Century." American Economic Journal: Macroeco- nomics, 4(3), 66-95. 22 [17] Donaldson, D. (2013). "Railroads of the Raj: Estimating the Impact of Transportation Infrastructure." American Economic Review, forthcoming. [18] Eaton, J., Grossman, G. (1986). "Optimal Trade and Industrial Policy under Oligopoly." Quarterly Journal of Economics, 101: 383–406. [19] Gibson, P., Wainio, J., Whitley, D. and Bohman, M. (2001). "Pro…les of Tari¤s in Global Agricultural Markets". Agricultural Economic Report No. 796, Market and Trade Economics Division, Economic Research Service, U.S. Department of Agricul- ture. [20] Karp, L. S. and McCalla, A.F. (1983). "Dynamic Games and International Trade: An Application to the World Corn Market." American Journal of Agricultural Economics, 65: 641-56. [21] Krishna, K. and Thursby, M. (1992). "Optimal policies and marketing board objec- tives." Journal of Development Economics, 38, 1-15. [22] Kolstad, C. D. and Burris, A.E. (1986). "Imperfectly Competitive Equilibria in In- ternational Commodity Markets." American Journal of Agricultural Economics, 68: 25-36. [23] Maccini, S. and Yang, D. (2009). "Under the Weather: Health, Schooling, and Economic Consequences of Early-Life Rainfall." American Economic Review, 99(3): 1006–1026. [24] Matsuura, K. and Willmott, C.J. (2007). Terrestrial Air Temperature and Precipitation: 1900-2006 Gridded Monthly Time Series, Version 1.01. http://climate.geog.udel.edu/ climate/. [25] Gawande, K. and Bandyopadhyay, U. (2000). "Is Protection for Sale? Evidence on the Grossman-Helpman Theory of Endogenous Protection." Review of Economics and Statistics, 82(1): 139–52. [26] Graaf, J. (1949–1950). "On Optimum Tari¤ Structures." Review of Economic Studies, 17(1): 47–59. [27] Grossman, G. and Helpman, E. (1994). "Protection for Sale." American Economic Review, 84: 833–50. [28] Grossman, G. and Helpman, E. (1995). "Trade Wars and Trade Talks." Journal of Political Economy, 103: 675–708. 23 [29] Jones, B. and Olken, B. (2010). "Climate shocks and exports." American Economic Review, 99: 454-459. Hare, M. and Kammen, D. (2008). "Memorandum on Land Use Change and the [30] O’ Global Food System.", UC Berkeley. [31] Paarlberg, P. L. and Abbott, P.C. (1986). "Oligopolistic Behavior by Public Agencies in International Trade: The World Wheat Market." American Journal of Agricultural Economics, 68: 528-42. [32] Pick, D.H. and Park, T.H. (1991). "The Competitive Structure of U.S. Agricultural Exports." American Journal of Agricultural Economics, 73: 133-141. [33] Sarris, A. H. and Freebairn, J. (1983). "Endogenous Price Policies and International Wheat Prices." American Journal of Agricultural Economics, 65: 214-24. [34] Smith, A. (1776). An Inquiry into the Nature and Causes of the Wealth of Nations. Methuen & Co., Ltd. [35] Spencer, B. and Brander, J. (1983). "International R&D rivalry and industrial strat- egy." Review of Economic Studies, 50: 707–22. 24 Table 1: Descriptive statistics Variable Obs Mean Std. Dev. ln simple tariffs 41,043 2.034 1.221 ln weighted tariffs 40,606 1.582 1.680 ln rain 41,043 2.469 0.757 agric 41,043 0.129 0.335 agric2 41,043 0.209 0.406 irrig 18,812 13.626 17.476 North America 41,043 0.046 0.211 Latin America and Caribbean 41,043 0.369 0.482 Europe 41,043 0.053 0.225 Africa 41,043 0.200 0.400 Asia-Paci�c 41,043 0.329 0.469 Number of countries 70 Number of products 65 Country-level data on the proportion of agricultural land that is irrigated are available for 32 countries. Table 2: Baseline estimates Dep. variable ln simple tariffs ln weighted tariffs (1) (2) (3) (4) ln rainjt 0.064 0.035 0.125 0.082 [0.076] [0.076] [0.094] [0.093] ln rainjt ∗ agrici 0.230*** 0.338*** [0.047] [0.064] Observations 41,043 41,043 40,606 40,606 Number of countries 70 70 70 70 Number of products 65 65 65 65 R-squared 0.572 0.574 0.401 0.403 The estimation method is OLS. Each model includes product, country and year �xed effects. Robust standard errors clustered by country in brackets. *** p ≤ 0.01 ,** p ≤ 0.05, * p ≤ 0.1 1 Table 3: Robustness, restricted sample Dep. variable ln simple tariffs ln weighted tariffs (1) (2) (3) (4) ln rainjt 0.039 0.006 0.106 0.061 [0.092] [0.091] [0.113] [0.111] ln rainjt ∗ agrici 0.263*** 0.354*** [0.049] [0.068] Observations 32,436 32,436 32,117 32,117 Number of countries 51 51 51 51 Number of products 65 65 65 65 R-squared 0.579 0.582 0.415 0.418 The estimation method is OLS. Each model includes product, country and year �xed effects. Robust standard errors clustered by country in brackets. *** p ≤ 0.01 ,** p ≤ 0.05, * p ≤ 0.1 Table 4: Robustness, alternative de�nition of agricultural goods Dep. variable ln simple tariffs ln weighted tariffs (1) (2) (3) (4) ln rainjt 0.064 0.015 0.125 0.062 [0.076] [0.074] [0.094] [0.092] ln rainjt ∗ agrici 0.235*** 0.304*** [0.051] [0.059] Observations 41,043 41,043 40,606 40,606 Number of countries 70 70 70 70 Number of products 65 65 65 65 R-squared 0.572 0.574 0.401 0.403 The estimation method is OLS. Each model includes product, country and year �xed effects. Robust standard errors clustered by country in brackets. *** p ≤ 0.01 ,** p ≤ 0.05, * p ≤ 0.1 2 Table 5: Robustness, alternative de�nition of agricultural goods and restricted sample Dep. variable ln simple tariffs ln weighted tariffs (1) (2) (3) (4) ln rainjt 0.039 -0.015 0.106 0.040 [0.092] [0.089] [0.113] [0.110] ln rainjt ∗ agrici 0.261*** 0.317*** [0.056] [0.063] Observations 32,436 32,436 32,117 32,117 Number of countries 51 51 51 51 Number of products 65 65 65 65 R-squared 0.579 0.583 0.415 0.418 The estimation method is OLS. Each model includes product, country and year �xed effects. Robust standard errors clustered by country in brackets. *** p ≤ 0.01 ,** p ≤ 0.05, * p ≤ 0.1 3 Table 6: Accounting for tariff binding Panel A Dep. variable: ln simple tariffs Full sample Applied < 0.9 Bound Applied < 0.8 Bound Applied < 0.7 Bound (1) (2) (3) (4) ln rainjt 0.022 0.004 0.007 0.011 [0.070] [0.071] [0.071] [0.071] ln rainjt ∗ agrici 0.218*** 0.228*** 0.225*** 0.223*** [0.053] [0.050] [0.049] [0.048] Observations 33,454 31,128 30,538 29,798 Number of countries 69 69 69 69 Number of products 65 65 65 65 R-squared 0.580 0.582 0.582 0.583 4 Panel B Dep. variable: ln weighted tariffs Full sample Applied < 0.9 Bound Applied < 0.8 Bound Applied < 0.7 Bound (1) (2) (3) (4) ln rainjt 0.071 0.069 0.072 0.065 [0.087] [0.094] [0.095] [0.098] ln rainjt ∗ agrici 0.271*** 0.277*** 0.272*** 0.265*** [0.068] [0.066] [0.065] [0.064] Observations 33,114 29,980 29,403 28,726 Number of countries 69 69 69 69 Number of products 65 65 65 65 R-squared 0.416 0.415 0.414 0.411 The estimation method is OLS. Due to availability of data for bound tariffs, the period of analysis is restricted to 1995-2006. Each model includes product, country and year �xed effects. Robust standard errors clustered by country in brackets. *** p ≤ 0.01 ,** p ≤ 0.05, * p ≤ 0.1 Table 7: Accounting for tariff binding, tobit random effects Dep. variable ln simple tariffs ln weighted tariffs (1) (2) (3) (4) ln rainjt 0.030 0.034 0.091** 0.105** [0.026] [0.026] [0.043] [0.043] ln rainjt ∗ agrici 0.220*** 0.220*** 0.273*** 0.267*** [0.018] [0.017] [0.029] [0.028] Observations 33,454 33,454 33,114 33,114 Number of countries 69 69 69 69 Number of products 65 65 65 65 Country effects Random Fixed Random Fixed Product effects Fixed Random Fixed Random Log likelihood -38777.222 -38718.094 -52882.104 -52832.222 The estimation method is Tobit random effects. Due to availability of data for bound tariffs, the period of analysis is restricted to 1995-2006. Each model includes year �xed-effects. Marginal effects resported. Standard errors in brackets. *** p ≤ 0.01 ,** p ≤ 0.05, * p ≤ 0.1 Table 8: Non-linear speci�cation of rainfall Dep. variable ln simple tariffs ln weighted tariffs (1) (2) (3) (4) (5) (6) low rainjt 0.137 0.131 0.081 0.073 [0.086] [0.086] [0.098] [0.098] low rainjt ∗ agrici -0.340** -0.296** -0.407** -0.343* [0.143] [0.143] [0.182] [0.186] high rainjt -0.062 -0.060 -0.044 -0.042 [0.068] [0.068] [0.094] [0.094] high rainjt ∗ agrici 0.265* 0.253* 0.379** 0.366** [0.143] [0.144] [0.181] [0.183] Observations 41,043 41,043 41,043 40,606 40,606 40,606 Number of countries 70 70 70 70 70 70 Number of products 65 65 65 65 65 65 R-squared 0.572 0.572 0.572 0.401 0.401 0.402 The estimation method is OLS. Each model includes product, country and year �xed effects. Robust standard errors clustered by country in brackets. *** p ≤ 0.01 ,** p ≤ 0.05, * p ≤ 0.1 5 Table 9: Estimates by individual agricultural product Dep. variable ln simple tariffs ln weighted tariffs (1) (2) (3) (4) ln rainjt 0.034 0.015 0.082 0.062 [0.076] [0.074] [0.093] [0.092] ln rainjt ∗ live animals 0.118 0.137 0.306 0.326* [0.114] [0.115] [0.184] [0.182] ln rainjt ∗ meat products 0.286*** 0.305*** 0.507** 0.526*** [0.074] [0.075] [0.195] [0.194] ln rainjt ∗ dairy products and birds eggs 0.262** 0.281** 0.536*** 0.555*** [0.114] [0.118] [0.159] [0.161] ln rainjt ∗ cereals 0.320*** 0.339*** 0.348*** 0.368*** [0.062] [0.066] [0.088] [0.092] ln rainjt ∗ vegetables and fruit 0.296*** 0.315*** 0.336*** 0.355*** [0.059] [0.062] [0.090] [0.094] ln rainjt ∗ sugar and honey 0.301*** 0.320*** 0.354** 0.374** [0.072] [0.077] [0.148] [0.151] ln rainjt ∗ cofee, tea, cocoa and spices 0.245*** 0.264*** 0.266*** 0.286*** [0.062] [0.063] [0.092] [0.094] ln rainjt ∗ oilseeds and oleaginious fruits -0.015 0.004 0.045 0.064 [0.085] [0.084] [0.108] [0.110] ln rainjt ∗ feeding stuff for animals 0.001 0.205 [0.125] [0.151] ln rainjt ∗ miscellaneous edible products 0.252*** 0.256*** [0.065] [0.078] ln rainjt ∗ beverages 0.335*** 0.126 [0.075] [0.150] ln rainjt ∗ tobacco 0.292* 0.379** [0.159] [0.151] ln rainjt ∗ crude animal and vegetable materials 0.180* 0.127 [0.102] [0.118] Observations 41,043 41,043 40,606 40,606 Number of countries 70 70 70 70 Number of products 65 65 65 65 R-squared 0.574 0.576 0.404 0.405 The estimation method is OLS. Each model includes product, country and year �xed-effects. Robust standard errors clustered by country in brackets. *** p ≤ 0.01 ,** ≤ 0.05, * p ≤ 0.1 6 Table 10: Estimates by region Panel A ln simple tariffs LAC Africa Asia-Paci�c ln rainjt 0.142 -0.058 -0.075 [0.095] [0.127] [0.142] ln rainjt ∗ agrici 0.345*** -0.030 0.303** [0.076] [0.81] [0.115] Observations 15,151 8,248 13,512 Number of countries 22 18 23 Number of products 65 65 65 R-squared 0.528 0.611 0.589 Panel B ln weighted tariffs LAC Africa Asia-Paci�c ln rainjt 0.256* -0.125 -0.113 [0.134] [0.217] [0.183] ln rainjt ∗ agrici 0.398*** 0.274 0.388** [0.128] [0.316] [0.155] Observations 14,969 8,069 13,458 Number of countries 22 18 23 Number of products 65 65 65 R-squared 0.331 0.398 0.448 The estimation method is OLS. Each model includes product, country and year �xed effects. Robust standard errors clustered by country in brackets. *** p ≤ 0.01 ,** p ≤ 0.05, * p ≤ 0.1 Table 11: The role of irrigation Dep. variable ln simple tariffs ln weighted tariffs (1) (2) (3) (4) ln rainjt 0.048 0.056 0.013 0.024 [0.074] [0.074] [0.096] [0.096] ln rainjt ∗ agrici 0.243*** 0.132** 0.447*** 0.291*** [0.048] [0.051] [0.074] [0.069] ln rainjt ∗ agrici ∗ irrigj 0.004*** 0.006*** [0.001] [0.001] Observations 18,812 18,812 18,683 18,683 Number of countries 32 32 32 32 Number of products 65 65 65 65 R-squared 0.586 0.588 0.422 0.423 The estimation method is OLS. Each model includes product, country and year �xed effects. Robust standard errors clustered by country in brackets. *** p ≤ 0.01 ,** p ≤ 0.05, * p ≤ 0.1 7 Table A.1 : Variables and data sources Variable De�nition Source Rainfall Population-weighted average rainfall by Kenji Matsuura and Cort Willmott country-year in 100s mm. (2007) and Dell et al. (2012) Irrigation Proportion of agricultural land that is World Bank WDI irrigated by country (average of non- missing country-year observations for the period 2001-2009). Simple Tariffs Simple average of bilateral applied im- UNCTAD TRAINS port tariffs by country-product-year, SITC classi�cation. Weighted Tariffs Weighted average of bilateral applied UNCTAD TRAINS and UN COM- import tariffs by country-product-year, TRADE based on the share of each source coun- try in total country-product imports in the base year, SITC classi�cation. Imports Bilateral imports value by country- UN COMTRADE product in the base year, SITC classi�- cation. 8 Table A.2 : Sample period by country Country Period Country Period Algeria 1993, 1997-1998, 2001-2003, 2005-2006 Lesotho 2001, 2004-2006 Argentina 1992-1993, 1995-2006 Macedonia 2001, 2004-2006 Australia 1991, 1993, 1996-2006 Malawi 1994, 1996-1998, 2001, 2006 Bangladesh 1989, 1994,1999-2000, 2002-2006 Malasyia 1988, 1991-1993, 1996-1997, 2001-2003, 2005-2006 Belize 1996, 1999, 2001-2003,2006 Mali 1995, 2001-2006 Benin 2001-2006 Mexico 1991, 1995, 1997-2006 Bolivia 1993-2002, 2004-2006 Morocco 1993, 1997, 2000-2003, 2005-2006 Botswana 2001, 2004-2006 Mozambique 1994, 1997, 2001-2003, 2005-2006 Brazil 1989-2006 Myanmar 2001-2006 Brunei 1992, 2001-2006 Nepal 1993, 1998-2000, 2002-2006 Burkina Faso 1993, 2001-2006 New Zealand 1992-1993, 1996-2000, 2002-2006 Cambodia 2001-2003, 2005 Nicaragua 1995, 1998-2002, 2004-2005 Canada 1989, 1993, 1995-2006 Niger 2001-2006 Chile 1992-1995, 1997-2002, 2004-2006 Nigeria 1988-1990, 1992, 1995-2002, 2005-2006 China 1992-1994, 1996-2001, 2003-2006 Norway 1988, 1993, 1995-1996, 1998, 2000-2003, 2006 Colombia 1991-1992, 1994-1997, 1999-2002, 2004-2006 Pakistan 1995, 1998, 2001-2006 Costa Rica 1995, 1999-2005 Papua New Guinea 1997, 2002-2006 Croatia 2001, 2004-2006 Paraguay 1991, 1994-2006 Cuba 1993, 1997, 2002-2006 Peru 1993, 1995, 1997-2000, 2004-2006 Dominican Republic 1997, 2000-2006 Philippines 1988-1990, 1992-1995, 1998-2006 Ecuador 1993-1999, 2002, 2004-2006 Saudi Arabia 1994, 1999-2000, 2003-2006 El Salvador 1995, 1997-1998, 2000-2002, 2004-2006 Senegal 2001-2006 European Union 1988-2006 South Africa 1990-1991, 1993, 1996-1997, 1999, 2001, 2004-2006 Guatemala 1995, 1997-1998, 2000-2002, 2004-2005 Sri Lanka 1990, 1993-1994, 1997, 2000-2001, 2004-2006 Guyana 1996, 1999-2003, 2006 Switzerland 1990, 1993, 1995-2006 Honduras 1995, 1999-2002, 2004-2005 Taiwan, China 1989, 1992, 1996, 1999-2003, 2005-2006 Indonesia 1989-1990, 1993, 1995-1996, 1999-2006 Togo 2001-2006 Israel 1993, 2004-2006 Trinidad & Tobago 1991-1992, 1996, 1999, 2001-2003, 2006 Jamaica 1996, 2000-2003, 2006 Tunisia 1990, 1992, 1995, 1998, 2002-2006 Japan 1988-2006 Uganda 1994, 2000-2006 Jordan 2000-2003, 2005-2006 United States 1989-2006 Kenya 1994, 2000-2001, 2004-2006 Uruguay 1992, 1995-2002, 2004-2006 Korea, Republic of 1988-1990, 1992, 1995-1996, 1999, 2002, 2004, 2006 Venezuela 1992, 1995, 1997-2000, 2002, 2004-2006 Laos 2000-2001, 2004-2006 Vietnam 1994, 1999, 2001-2006 Lebanon 1999-2002, 2004-2006 Zimbabwe 1996-1999, 2001-2003 9 Table A.3 : SITC Classi�cation SITC Agric Agric2 Description 00 1 1 Live animals other than animals of division 03 01 1 1 Meat and meat preparations 02 1 1 Dairy products and birds eggs 04 1 1 Cereals and cereal preparations 05 1 1 Vegetables and fruit 06 1 1 Sugars, sugar preparations and honey 07 1 1 Coffee, tea, cocoa, spices, and manufactures thereof 22 1 1 Oil-seeds and oleaginous fruits 08 0 1 Feeding stuff for animals 09 0 1 Miscellaneous edible products and preparations 11 0 1 Beverages 12 0 1 Tobacco and tobacco manufactures 29 0 1 Crude animal and vegetable materials, n.e.s. 03 0 0 Fish, crustaceans, molluscs and aquatic invertebrates, and preparations thereof 21 0 0 Hides, skins and furskins, raw 23 0 0 Crude rubber (including synthetic and reclaimed) 24 0 0 Cork and wood 25 0 0 Pulp and waste paper 26 0 0 Textile �bres (other than wool tops and other combed wool) and their wastes 27 0 0 Crude fertilizers, other than those of Division 56, and crude minerals 28 0 0 Metalliferous ores and metal scrap 32 0 0 Coal, coke and briquettes 33 0 0 Petroleum, petroleum products and related materials 34 0 0 Gas, natural and manufactured 35 0 0 Electric current 41 0 0 Animal oils and fats 42 0 0 Fixed vegetable fats and oils, crude, re�ned or fractionated 43 0 0 Animal or vegetable fats and oils, processed; waxes and mixtures 51 0 0 Organic chemicals 52 0 0 Inorganic chemicals 53 0 0 Dyeing, tanning and colouring materials 54 0 0 Medicinal and pharmaceutical products 55 0 0 Essential oils and resinoids and perfume materials; toilet, polishing and cleansing preparations 56 0 0 Fertilizers (other than those of group 272) 57 0 0 Plastics in primary forms 58 0 0 Plastics in non-primary forms 59 0 0 Chemical materials and products, n.e.s. 61 0 0 Leather, leather manufactures, n.e.s., and dressed furskins 62 0 0 Rubber manufactures, n.e.s. 63 0 0 Cork and wood manufactures (excluding furniture) 64 0 0 Paper, paperboard and articles of paper pulp, of paper or of paperboard 65 0 0 Textile yarn, fabrics, made-up articles, n.e.s., and related products 66 0 0 Non-metallic mineral manufactures, n.e.s. 67 0 0 Iron and steel 68 0 0 Non-ferrous metals 69 0 0 Manufactures of metals, n.e.s. 71 0 0 Power-generating machinery and equipment 72 0 0 Machinery specialized for particular industries 73 0 0 Metalworking machinery 74 0 0 General industrial machinery and equipment, n.e.s., and machine parts, n.e.s. 75 0 0 Office machines and automatic data-processing machines 76 0 0 Telecommunications and sound-recording and reproducing apparatus and equipment 77 0 0 Electrical machinery, apparatus and appliances, n.e.s., and electrical parts thereof 78 0 0 Road vehicles (including air-cushion vehicles) 79 0 0 Other transport equipment 81 0 0 Prefabricated buildings; sanitary, plumbing, heating and lighting �xtures and �ttings, n.e.s. 82 0 0 Furniture and parts thereof; bedding, mattresses, mattress supports, cushions and similar stuff 83 0 0 Travel goods, handbags and similar containers 84 0 0 Articles of apparel and clothing accessories 85 0 0 Footwear 87 0 0 Professional, scienti�c and controlling instruments and apparatus, n.e.s. 88 0 0 Photographic apparatus, equipment and supplies and optical goods, n.e.s.; watches and clocks 89 0 0 Miscellaneous manufactured articles, n.e.s. 96 0 0 Coin (other than gold coin), not being legal tender 97 0 0 Gold, non-monetary (excluding gold ores and concentrates) Note: Agricultural products are de�ned based on Standard International Trade Classi�cation, Revision 4; United Nations Statistics Division 10 Figure 1: Rainfall and droughts Figure 2: Average bound and applied import tariffs on agricultural goods 120 Bound tariff Applied tariff in1998 100 80 60 40 20 0 Uruguay Colombia Nicaragua Republic of Korea Paraguay Philippines Ecuador Argentina Guatemala Indonesia Panama Vanezuela El Salvador Costa Rica India Morocco Tunisia Brazil Mexico Pakistan Note: Bound tariffs are MFN rates based on final URAA implementation. Applied tariffs represent annual average. Data on average applied import tariffs refer to 1998, with the exception of Costa Rica (1995), Republic of Korea (1996), Indonesia (1996), India (1997) and Morocco (1997). Source: Economic Research Service, USDA.