wePS Z30- POLICY RESEARCH WORKING PAPER 23 02 Why Liberalization Alone Policiesto osteraccurulation of the assets needed for Has Not Improved agricultural production Agricultural Productivity (including draft animals and implements) and to provide in Zam bia complementary public goods (education, credit, and good agricultural extension services) The Role of Asset Ownership could greatiy help reduce and Working Capital Constraints poverty and improve productivity in Zambia Klaus Deininger Pedro Olinto The World Bank Development Research Group Rural Development March 2000 POLICY RESEARCH WORKING PAPER 2302 Summary findings Deininger and Olinto use a large panel data set from Second, the authors reject the hypothesis that the Zambia to examine factors that could explain the application of fertilizer is unprofitable because of high relatively lackluster performance of the country's input prices. Rather, fertilizer use appears to have agricultural sector after liberalization. Zambia's declined because of constraints on supplies, which liberalization significantly opened the economy but failed government intervention exacerbated instead of to alter the structure of production or help realize alleviating. (Extending the use of fertilizer to the many efficiency gains. They reach two main conclusions. producers not currently using it would be profitable, but First, not owning productive assets (in Zambia, draft increasing the amount applied by the few producers who animals and implements) limits improvements in now have access to it would not be.) agricultural productivity and household welfare. Owning Policies to foster accumulation of the assets needed for oxen increases income directly, allows farmers to till agricultural production (including draft animals and their fields efficiently when rain is delayed, increases the implements) and to provide complementary public goods area cultivated, and improves access to credit and (education, credit, and good agricultural extension fertilizer markets. services) could greatly help reduce poverty and improve productivity. This paper - a product of Rural Development, Development Research Group - is part of a larger effort in the group to analyze determinants of rural growth and market participation. Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Maria Fernandez, room MC3-508, telephone 202-473-3766, fax 202-522-1151, email address mfernandez2@worldbank.org. Policy Research Working Papers are also posted on the Web at www.worldbank.org/research/workingpapers. The authors may be contacted at kdeininger@worldbank.org or polinto@Hworldbank.org. March 2000. (22 pages) The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas abont 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 autho,s and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. 7hey do not necessarily represent the view of the World Bank, its Executive Directors, or the countries they represent. Produced by the Policy Research Dissemination Center Why Liberalization Alone Has Not Improved Agricultural Productivity in Zambia The Role of Asset Ownership and Working Capital Constraints Klaus Deininger and Pedro Olinto World Bank Washington, DC JEL codes Q12, 013, 015 Address for correspondence: Klaus Deininger World Bank 1818 H Street, NW Washington, DC 20433 Tel 202 473 0430 Fax 202 522 1150 Kdeininger@worldbank.org 1. INTRODUCTION In the early i990s, Zambia initiated an ambitious program of liberalization that significantly opened the economy, shifting from a highly regulated and centralized to a more market-based and liberal economic paradigm. While these changes have profoundly affected the economic environment for the agricultural sector, they have up to now failed to alter the structure of production and help realize the efficiency gains that were expected to be associated with economic liberalization. The rural economy continues to be highly dependent on maize, producers' participation in markets for input as well as output has hardly increased, credit remains out of the reach of the ordinary producer, and no appreciable change in investment can be discerned. This paper examines whether this disappointing performance is due to the fact that reforms undermined the profitability of agricultural production or due to other factors, such as market imperfections and structural limitations at the household level. This is important for policy advice because, depending on which of these two possibilities is correct, different forms of policy intervention would be appropriate - either price measures to improve the profitability of farming or non-price policies aiming facilitating increased private sector involvement combined with better provision of public goods. We analyze this issue using data from a two-year panel of about 5000 rural households. Panel data econometric methods allow us to overcome the biases inherent in cross sectional analysis and, from estimation of a production function plus associated demand equation for land, fertilizer, and credit, make inferences regarding the productivity of different factors. Our main results are as follows: * Fertilizer has a significant output-increasing effect. While extending fertilizer use to the large number of producers who are currently not using this input would be highly profitable economically, increasing the amount actually applied by the small number of producers with access to this input is not profitable. * Cattle ownership increases income directly, acts as an "insurance" by allowing producers to till their fields in a more timely fashion in times of delayed rain, increases area of land cultivated, and improves access to credit and fertilizer markets. This points towards the persistence of significant imperfections in markets for rural labor, credit, and draught animals, but at the same time suggests that policies to increase cattle ownership in rural areas could have high payoffs. - We find, somewhat surprisingly, that credit has a direct productivity increasing effect - most likely through supervision that is associated with it. This notwithstanding, the main impact of credit is through increases in cultivated area and -as in the case of fertilizer- major benefits are likely to be realized by providing access to producers who do currently not have access to any credit. * In terms of total factor productivity, female headed households are as productive as male headed households. They are, however, disadvantaged with respect to credit access and, partly as a consequence, use less land and fertilizer. This could point towards labor market imperfections or higher risk aversion on the part of these households. * Supply of extension has a positive, though insignificant, impact on total factor productivity and no significant impact on demand for cultivated area. Improving the quality of extension and gearing it more to be in line with Zambia's relative factor endowment would thus be a necessary precondition for the expansion of such services to be justified economically. * Once other inputs are accounted for, education does not affect productivity or the amount of area cultivated. However, education enables farmers to overcome market imperfections, as reflected in the fact that more educated farmers demand higher amounts of fertilizer and credit per hectare, tend to be more integrated into output markets. Section two describes the sectoral background and, in this context, posits the key question to be investigated and the relationship to earlier literature. A detailed discussion of the conceptual model and estimation issues follows in section three. Section four discusses data issues and empirical results. Section five concludes by reviewing implications for policy and further research. 2. SECTOR BACKGROUND AND DATA SOURCES Despite an abundant land resource (75 million hectares, 55% of which are suitable for agricultural production) favorable agro-climatic endowment, and low population density (a total of 9 million people out of which slightly less than 6 million live in rural areas), Zambia has long neglected its rural population and agricultural sector. At the macro level, almost exclusive reliance on copper for exports, industrial protection, overvalued exchange rates, and unsustainable fiscal policies all contributed to a long-term decline of the productive infrastructure in a country that had, in the early 1960s, a per capita GDP higher than Korea. These macro-economic policies were complemented by unsustainable government interventions at the sectoral level (Jansen 1991). Producer subsidies for fertilizer and maize led to the extension of maize cultivation into unsuitable areas which increased vulnerability to drought, distorted factor prices, and biased the direction of research away from high value export crops to staples with low profitability. Although Zambia does not have a strong comparative advantage in bulk commodities, more than 50% of research and extension resources were regularly devoted to maize, mainly because of price distortions introduced by government policies (World Bank, 1992). As a result, Zambia lost ground even in production of commodities (e.g. tobacco) where it had once been a leading producer in the region (Keyser 1996). In addition, consumer subsidies posed a huge fiscal burden. Between 1971 and 1988 the average subsidy to consumers amounted to about 70% of the price of maize meal maize related subsidies accounted for almost 20% of total government spending in the 1980s (McPherson, 1995) Pan- territorial and pan-seasonal pricing, together with state monopoly trading, distorted incentives for private sector storage and helped to establish an inefficient marketing and processing structure. Credit subsidies biased investment decisions, discouraged savings mobilization, and undermined the viability of an independent financial system - real interest rates were negative throughout the 1981-93 period, with a peak of -137% in 1992 (World Bank 1992). Following a number of aborted attempts at reform (e.g. Jansen and Ruvumo 1992; McPherson 1995), the government that took power in 1991 initiated a series of far-reaching macro-economic and sectoral reforms including the elimination of subsidies, privatization, and greater decentralization. From our perspective, two issues are of relevance. First, the reductions in maize prices together with the increases in fertilizer prices caused by the elimination of subsidies unambiguously worsened the terms of trade for the agricultural sector. Second, withdrawal of government procurement and distribution agents was supposed to open up space for the private sector to exploit profit opportunities - but this was countervailed by continued ad-hoc involvement by the public sector which greatly undermined predictability for the public sector. Conceptual models predict that these reforms should lead to a major shift of producers out of maize (Holden, Taylor, and Hampton 1999) and emergence of a private trading sector that would realize significant welfare gains (Mwanaumo, Masters, and Preckel 1997). However, little of this has materialized, owing mainly to the fact that adjustment was associated with reductions in input use rather than changes in the quantity or mix of output. On the output side, between 1990 and 1996 Zambia's rural households continued to produce mainly for subsistence (only about 40% sold in the market, virtually unchanged from 1991), rural incomes remained backward and low (about one third urban households), and no diversification out of maize took place in response to changed price signals - the share of area planted to maize in 1996 and 1997 was higher than it had been in 1990. Changes in input prices had not only a marked effect on application of inputs; use of hybrid seed and fertilizer use was reduced by one half and one third, respectively, they were also associated with a reduction in farmers' capital stock. Partly due to a drought in 1992 from which the sector has yet to recover, the percentage of producers owning livestock decreased by almost one third (from 18% to 12%) between 1992 and 1996. It appears that policy reforms have contributed to stagnation or even regression, instead of helping Zambia's agricultural sector realize the strong regional growth linkages that have been demonstrated in the literature (Hazell and Hojjati 1995). The question whether this disappointing performance is due to the fact that reforms undermined the profitability of agricultural production or due to other factors, such as market imperfections and structural limitations at the household level, that prevent a more vigorous supply response is of considerable policy relevance. We aim to provide and answer by focusing on two aspects. First, we are interested whether producers apply variable inputs in a profit-maximizing fashion. Second, we aim to identify structural determinants of factors demand that may underlie such inefficient allocation, if it exists. In doing so we examine variables at the household level (education, female headship, asset endowments) and in the broader market environment (extension, access to infrastructure). While none of these issues are new, our approach improves on previous literature by using a larger sample and applying panel data econometric techniques. A number of contributions (Hatting et al. 1998, Savadogo et al. 1995, etc.) have demonstrated that- farm level data can provide useful insights to examine these issues. However, these studies are characterized by small samples (122 and 150 households, respectively), a restricted set of variables (e.g. no information on household education, non-agricultural assets, or soil quality is available), and reliance on cross-sectional analysis. This limits not only the ability to generalize from the results obtained and make broader inferences but, more importantly, may also bias the results significantly. The reason is that there is likely to be a strong correlation between the large number of unobserved household and farm characteristics and input use. As, to take just one example, more educated farmers would tend to use inputs more intensively, the coefficients on observed inputs in a traditional production function will be biased upwards. Thus, the impact of purchased inputs would be overstated and cross-sectional evidence alone would appear to suggest that farmers "'underuse" these inputs. To overcome these limitations of cross-sectional analysis, we apply panel data techniques. Under the condition that farmers' unobserved characteristics are time-invariant, use of panel data techniques allows to overcome these limitations that have plagued the earlier literature and are likely to have led to an upward bias on the coefficients for purchased inputs. Instead of making it appear these inputs are more effective and farmers inefficient in not applying the optimal quantities, they allow to retrieve structural parameters regarding the impact of specific inputs in the production function. Furthermore, use of instrumental variable estimators allows us to estimate the impact of time invariant productivity shifters such as producers' endowments of physical assets, human capital, and their access to public services. Both the structural parameters for specific inputs and the productivity shifters enable us to derive more precise policy recommendations. Instead of relying on a survey that was specifically fielded for this purpose and the small sample that is generally associated with such efforts, we use data on farm production from a panel of 4853 farm households from the Central Statistical Office's Post Harvest Surveys for 1993/94 and1994/95. The data were collected sufficiently long after initiation of the reforms for them to have shown at least some impact. While the time period covered is rather short, it coincides with significant variability in the economic and agro-climatic environment such as the final closure of official credit institutions, and a drought in 1994/95. As a result, there is sufficient "within" household variation in input use for panel data methods to be meaningful and allowing us to estimate a production function and input demand functions for labor, fertilizer, and credit. Before describing the data in detail, we discuss the conceptual model and estimation issues. 3. MODEL AND ESTIMATION ISSUES Conceptual Framework Consider a household i that at time t may undertake production employing a Cobb-Douglas technology given by: (I) ln(Y,,) = ln(A,,) + PI ln(T1,) + 321n(F,,+ 1), where Ti, is the total area cultivated, F1, is the total amount of fertilizer applied,Aj, is an index that measures the total factor productivity (TFP) achieved by household i at time t, and 1I and P32 are technology parameters assumed to be constants across households and time. For simplicity we assume that other inputs--as labor effort, seeds, mechanical and animal power--are employed in fixed proportions to the cultivated area, i.e. T,, represents the area of prepared and planted land! We assume that the TFP index Ai, is detennined by household/farm i's observed and unobserved characteristics at time t, specifying the following log-linear equation forAj,: (2) ln(A,,) = ao + cl'X;, + a2'Z, + rli + 1ij,. In (2), Xi, and Z, are vectors of observed household/farm specific time-variant and invariant characteristics, respectively, which will be described below. The error term in (2) has two components: (i) r1i is a household/farm specific time-invariant effect, which is known to the household at the time production decisions are made, but is unknown to the econometrician; and (ii) the stochastic exogenous shock ui,, which is assumed to be independent and identically distributed (iid) across households and time, and is observed neither by the households nor the econometrician. For tractability we assume that the decision making household's conditional expectation of exp(Q.j,) given Xi,, Z;, Tj,, Fj, equals one. Thus, household i chooses T,, and (F,,+ 1) in order to maximize expected profits at year t which are given by: (3) Pi, = E[Yj, I Xj,,Zj,Tj,,Fj,j - r,T,, - q,(F,,+1) + q, where r, is the market rental rate of prepared land, and q, is the market price (plus transportation and application costs) of a unit of fertilizer. 'This specification captures the fact that fertilizer is not an essential input, so that Fj,=O does not imply in Yi,=O. The first order necessary conditions (FONCs) for maximum profit are respectively given by: (4.a) P7. - r, 0, (P7- r,)T,, = 0, and TI, > 0; (4.b) P''- q, 0, (P1;+ - q,)(F,,+l) = 0, F,, > 0. The solution to system (4) yields the reduced form demand equations for fertilizer and cultivated area, which are given by: (5.a) lnT,, = yo(w,q,,r,;P)+ y1l(f)(ao + ac'X;, + a2'Zi + ale); (5.b) In(Fj,+l)= yo(w,,q,,r,;P)+ y(730Xco + al'Xi, + a2'Zi + Tj), where Yk, 4k,Yk, for k = 0,1 are functions of the structural parameters in P and the exogenously given prices w,, q,. Thus, under the above specification, the resulting reduced form demand equations are linear in the household/farm specific effects m., and moreover, rj has a time- invariant coefficient that can be eliminated by using fixed effects estimation techniques. Econometric Model and Estimation Issues The Production Function Equations (1) and (2) suggest the estimation of the following regression equation: (6) ln(Y,,) = ao + a'I'Xil + a2'Zi + 31 ln(T,,) + f321n(F1,+ 1) + vi, where vi, is a random disturbance with two components: i.e., v;, = Ili + ei,, where 1mi is defined above, and Si, is a idiosyncratic household/farm shock in productivity which might be observable by the household but not by the econometrician. To estimate the structural parameters in (6), we must confront the problem of endogeneity of cultivated area ln(T,r), fertilizer use ln(Fj,+ l,), and asset endowments (Z), a common feature in the estimation of production functions with household and farm level data. As seen in the reduced form equations (4.a)-(4.b), both ln(T,r) and ln(Fi,+ lU,) are functions of unobserved time-invariant household/farm characteristic m, which is a component of vi,, the error term in (6). Therefore, In(T,,) and ln(Fi,+ I j,) are clearly correlated with the disturbance v;,, which implies that the OLS estimator of the parameters in (6) will be biased and inconsistent. As discussed in Mundlak (1996), panel-data can provide a wealth of Instrumental Variable (IV) estimators that tackies this common identification problem in the estimation of production functions via a primal approach. In what follows we discuss three IV estimators: (i) The within (or fixed-effects) estimator, (ii) the two-stage least squares estimator of Hausmann and Taylor (1981), hereafter the HT estimator, and the Amemyia and MaCurdy (1989), hereafter the AM estimator. These are discussed below. The Within Estimator: To simplify notation, rewrite equation (6) as: Yi, = 11'Xj, + 712'Zi + +1,i i= 1,..,N; t = I,..,T, where xi, is a K x I vector collecting all the time-variant explanatory variables in equation (7),z, is a G x 1 vector collecting all the time-invariant explanatory variables, and it1 and it2 are conformably dimensioned parameter vectors. We assume that the disturbances si, are iid N(0,a62) and the individual effects m1 are iid N(O,u,,2). The time and household-variant component 6,, are assumed to be orthogonal to both the explanatory variables and the individual effects, whilerli may be correlated with parts of x and z. Combining all NT observations we can write (8) as: y= xn +zn4+-V7 + where y and are NT x 1, x is NT x K, z is NT x G, and V is an NT x N matrix of individual- specific dummy variables. Now define the matrix Pv = V(V'V)-'V as the projection onto the column space of V. Thus, Pv is a matrix that transforms a vector of observation into a vector of individual means across time: i.e., Pvy,, = (=I_)zyi, yi. Then, Qv = I - Pv is defined as the projection onto the null space of V, i.e., Q, produces a vector of deviations from individual means: i.e., Pvy,= yj, -yp. The within estimator is computed by projecting (9) onto the null space of V and performing least- saquares. Because Q^z = 0, and Q,n = 0, only 7c1 can be estimated. Therefore, the within estimator of Rit is: (10) 2hw = (x'Q^x)Yx'QVY, and is consistent whether or not the explanatory variables are correlated with the individual effects 77i. The problem with the within estimator is that it does not allow for the estimation of the vector 7it2. In what follows we explore two estimators that will be consistent under some mild orthogonality assumptions. 2SLS Estimators: As explained in Hausmann and Taylor (1981), a more efficient IV estimator can be computed if we are willing to assume that some explanatory variables in x and z are orthogonal to the individual effects 17k. Consider the partition ofx and z given by x = (xi, x2) and z = (Zi, z2), such that xi, and zi are orthogonal to i7r, but x2 and Z2 are not. The HT 2sls estimator is therefore given by: i, 2 [(X, Z)'Q-1/2pA Q-12(X, Z)]-l (X, Z)'Q"2PA Q*1/2y where Q&/I2 = Q + OPK, is a weighting matrix suclh that 0 = Cy6(2 + Tal2'', PA = A(A 'A) A iS the projection onto the column space of A. For the HT estimator, A is a matrix of instruments given by: A = (Q xi, Q,x2, P,Xi, Zl) Thus, each variable in xi provides two instruments (Q,x1 and P, xj), while the variables inx2 and Zi provide one instrument each (QX2 and zi). The order condition for identification gives the result that the number of columns in xl must be at least as large as the number of columns in Z2. Amemyia and MaCurdy (I 986) suggests the following set of alternative instruments: A = (Qv xi, Q, x2, x *, zi) where xl* is a NTx TK matrix where each column contains values ofxl, for a single time period. The AM estimator uses each of the xl variables as (T+ I). Its order condition for existence is that Ttimes the number of columns in xl must be greater than or equal to the number of columns inzl. As shown by Amemyia and MaCurdy (1986), consistency of the AM estimator will depend on a stronger exogeneity assumption than the HT estimator. While the HT estimator requires only the means of the xi variables to be orthogonal to the individual effects, the AM estimator requires orthogonality at each point in time. Nevertheless, asAmemyia and MaCurdy (1986) and Breusch, Mizon and Schmidt (1989) point out, orthogonality between P, xl and the individual effects n is likely to result from the orthogonality between x, and n. Input demand equations The IV methods described above are also employed to estimate the parameters of the reduced form of a demand for cultivated land equation. However, because fertilizer demand is censored at zero, a Tobit specification is required. We use two types of estimators for these censored equations: (i) Bo Honore's (1992) trimmed least squares estimator (TLS) for fixed-effects panel data models, and a simulated maximum likelihood (SML) estimator which assumes that the 2 individual-specific random effect are orthogonal to the explanatory variables. 2 For a description of SML estimators for panel data Tobit type models, see Gourieroux and Monfort, 1993. As indicated in equations (5.a) and (5.b), we cannlot identify the own- and cross-price elasticities of demand since the functions yo(w,,q,,r,;f3) and y0(w,,q,,r,;f3) do not vary across cross-sectional units. Therefore we focus on estimating the effect of the houselhold specific variablesXi, and Zi on demand for inputs so that we can compute their overall expected effect on farm output. For instance, the total impact of ox ownership on output is given by: dY _ DY DY aT aY a(F + 1) DT a(F + 1) dXj DX, ATDX, D(F+1) DXi aX 2 aXj where estimates of ox;, Pi, and P32, are obtained tlhrough the estimation of the production function, and estimates of aT and (F ) are obtained through the estimation of the input demand DxJ Dx.j equations (5.a) and (5.b). Note however, that the estimates of a and (F ) are themselves axJ ax. functions of the production parameters P3i, and the total factor productivity parameters (I ,. Variahles included in tlte estimation Time-varying explanatory variables included in the production function are cultivated area, fertilizer used, household population, the value of cattle owned by the household, the stock of draught animals (oxen) and farm equipment, the amount of credit received, and a "weather shock" variable. "Weather shocks" are defined as the percentage deviation of precipitation from its 30-year average during the planting season, two growing periods, and one harvesting season. The implicit assumption that too much rain can be as harmful as too little of it seems to be quite realistic under Zambian conditions. Higher order terms (i.e. the weather shock variable squared, cubed and raised to the forth power) are introduced to account for possible non-linearities. Animal-owners' ability to perform farming operations in a more timely manner is likely to be of particular relevance under abnormal climatic conditions. To account for this possibility, we interact the weather shock with the value of the farm's stock of draught animals and farm equipment. Of the time varying variables, -household population, the non-interacted weather variables, the time trend and district dummies-time trend interactions are assumed to be exogenous while the remainder are assumed to be endogenous (i.e., correlated with unobserved individual effects). Time-invariant explanatory variables are the sex of the household head, access to extension, primary and secondary education dummies and sixty district dummies. Extension and the education dummies are assumed to be endogenous while the remaining time-invariant explanatory variables are assumed exogenous. Unfortunately, input price data are not available for the first year of the data and only selectively in the second year. Adding 60 district dummies and their interaction with time (not reported) will avoid bias due to omission of price variables. 4. RESULTS 4.1 Descriptive statistics Before discussing econometric estimation results it is of interest to describe intertemporal and inter-regional variation in the data using descriptive statistics. As illustrated in table 1, rural producer households in Zambia had on average 6.2 members and cultivated 1.9 hectares. The the mean age of the household head was 45 and about 20% of households were female headed. 56% and 15% have completed primary and secondary education, respectively, and 24% of households had access to extension. Both value of total production and area cultivated show significant variation across regions, with output value ranging from K 407,000 in Central to 116,000 in Western region (corresponding to mean areas of 2.82 and 1.44 hectares). Looking at changes over time, one notes that output value changed little over time even though area cultivated decreased and market participation, measured as the percentage of output (in value terms) sold decreased for almost all provinces. The share of producers with access to credit decreased significantly, from 19.2% to 9.8%, in addition to a significant drop in the mean amount received. This decrease in access to financing seems to have precipitated a reduction both in the share of producers having access to fertilizer (from 35% to 27%) and a halving of the amount spent on fertilizer by those who use it. To examine the implications of this for production in more detail we turn to econometric analysis. 4.2 Econmetric results Results from the production function estimation (table 2) point to a number of interesting results. First, availability offamily labor is an important determinant of output, indicating that labor markets in rural Zambia are thin or non-existent. The elasticity of output with respect to household population is between 0.07 and 0.12, suggesting that even once other factors such as area, fertilizer use, and ownership of oxen are controlled for, larger households tend to be more productive than small households. The rationale for such a relationship is that, as one would expect in an environment where land is relatively abundant, larger households would have less difficulty to muster the necessary labor to complete critical tasks during seasons of peak demand (e.g. weeding) when spot labor markets dry up. Second, the data also allow ascertaining the impact offertilizer on output without the biases intoduced by unobserved land quality and farmer skills which commonly plague cross-sectional analysis. This is of interest in view of the heated policy discussion on whether or not use of fertilizer will still be profitable at the increased post-liberalization prices for this input. We find that the output elasticity of fertilizer varies between 0.07 and 0.08. To interpret this figure, we have to distinguish between farmers who apply fertilizer to their fields and those who do not. For the first group, the marginal benefit from applying additional amounts of fertilizer is slightly lower than the marginal cost, pointing towards consistency with profit-maximizing behavior. This strongly supports the hypothesis that, even at the increased post-liberalization prices, fertilizer use is economically profitable and that farmers with access to fertilizer therefore apply more or less the "optimum" amount and do not appear to be quantity-rationed. By contrast, producers who are rationed out of the fertilizer market, seem to forgo significant productivity and welfare benefits. Estimates suggest that providing producers without fertilizer access with the average amount of fertilizer applied by users (170kg) would increase their income by about US$85, generating a 70% return on the investment of US$ 503 This high return points towards non-price rationing, whereby access rather than price appears to be the relevant problem. The policy implication emerging from this is that, instead of aiming to curb "speculative excess" by traders through continuing government involvement in fertilizer procurement, encouraging private sector entry would be expected to result in increased and more optimal use of this input. Descriptive statistics, which indicate that that many farmers close to the line of rail bought more than double the amount of fertilizer they actually used, support this notion and furthermore suggest that government subsidies constituted a de-facto transfer to the rich who have advantageous access to infrastructure (Deininger et al. 1998). Third, to identify the impact of time-varying factors, the instrumental variable techniques described earlier also allow us to identify the effect of time-invariant or slowly changing conditions such as household headslhip, distance to infrastructure, education, and access to extension services. We find that female-headed households are as productive as their male counterparts, i.e. the dummy is negative but not significant. While the state of rural infrastructure is often viewed as a major obstacle to expansion as well as diversification of agricultural production, our estimates suggest that a producer's distance to markets does not affect total factor productivity, altlhough it has an impact on factor use. Similar conclusions hold for education (separated into primary and secondary) which does not seem to increase productivity, but, as can be seen from the fertilizer demand function and credit use equation (tables 3 and 4), affects input use in what can still be considered an environment with numerous market imperfections. One explanation might be that agricultural experience, rather than formal education, is of greater relevance for increasing total factor productivity. The share of individuals in any village who have access to extension (excluding the producer under concern) has a positive but statistically insignificant impact on total factor productivity. This would support the claim that the government's system has not yet adjusted to fully incorporate the realities and requirements of a liberalized environment. Fourth, the results of the production function estimation allow to make inferences on the impact of credit access. They indicate that there is a positive and statistically significant relationship between the amount of credit received by a farm household and total factor productivity. Thus, an extra US $100 (100,000 K) of credit received by an average farm household, is expected to increase output by between 2.6 and 4%, even when other inputs are held constant. Moreover, we note that increasing access to credit from zero to the mean value for the sample of credit users (K 150,000) would, according to the estimates, augment output of non-borrowers by 3% due just to this increase in total factor productivity. This is surprising, as one would normally expect any impact of credit pre-planting season rain is below the long-term average) possession of animals would allow to complete soil preparationi area faster once soil moisture has reached workable levels, thus minimizing the yield loss incurred. Obviously, owners would tend to till their own fields before renting out to others, implying that farmers who own draught animals are -at least to some extent- less vulnierable to climatic shocks. This is of particular interest since a similar interaction term between the number of household members and rainfall (suggesting that larger households have an advantage in mobilizing family labor quickly) is insignificant. In addition to its impact on total factor productivity, ownership of productive assets (in this case draught animals) is also a key determinant of expected land demand, which in turn has a large impact on production.5 The estimates imply that by exogenously receiving one pair of oxen The fertilizer price is the real cost ex depot. This is needed because the survey contains too little detail on modalities of payment for prices to be useful. Adding transport costs for a distance of 500 km (at a cost of US $ 0.1 per t-km) reduces the return to about 45%, which should still be sutlicient to finance the input on credit. 4 Under the assumptions discussed in the appendix the iV estimators are not biased by the likely correlation between unobserved managerial skills and credit access. 9 Recall from the results of the production function regression that land has a production elasticity of between 0.6 and 0.7. valued at K310,000, non-ox owners would be able to increase their area of cultivation by approximately 25 percent, which in turn represents an expected increase in income of approximately 18 percent (or K45,000, given that the average non-ox-owner generates approximately K250,000 of farm output). If we add the 4-12 percent increase in total factor productivity computed above, we conclude that non-ox-owners would be able to pay back the pair of oxen plus interest in approximately 4 to 7 years (given a total increase in income of approximately K40,000 to K80,00.0 per year). The estimated effect of draught animal ownership oni producers' propensity to apply fertilizer is ambiguous (table 3). Wilie large and statistically significant under the SML model, the impact is small and insignificant under the TLS model. Since the SML estimate is only consistent under the assumption of orthogonality between the farm-specific random effect and the explanatory variables, whereas the TLS estimate is consistent even when this orthogonality condition does not hold, we conclude that there is no significant direct impact of draught animals ownership on fertilizer demand. To Anotlher way in which ownership of draught animals may affect farm output is through its impact oni fertilizer use and area cultivated via increased credit access. That is, because draught animals may serve as collateral in credit transactions, draught animals owners may have better access to credit than non-owners. To test this hypothesis, we estimate a credit access random-effects Probit model. Results, presented in the last column of table 4, suggest that, indeed, draught animal ownership has a significant impact on credit access' Taking all of these effects together ox ownership would increase household income by about 22% or K 56,000 per year. Even with a real interest rate of 15%, the capitalized value of this benefit would be greater tlhan the cost of the investment, suggesting that even in a high-interest rate environment, and neglecting indirect benefits, acquisition of oxen would be a profitable investment and that the issues involved should be explored in more detail.7 5. CONCLUSION In this paper we explored whether structural factors or lack of profitability are at the root of the apparently very limited supply response by rural producers in Zambia. Using data that were ' We opt for a random-effects Probit model instead of a fixed-effects Tobit model because we cannot determined ex ante whether observed credit use is determined by supply, in which case use would indicate access, or demand. If farmers who own more draught animals demand more credit but are also less likely to be credit rationed, the fixed-effects Tobit model, would likely yield an upward- biased estimate of the impact of oxen ownership on access to credit. The reason is that it departs from the implicit assumption that all producers are credit rationed. Such bias would be smaller under a random-effects specification since poor farmer with zero loans are less likely to implicitly demand zero credit. The purpose of this example is to highlight the implication from regression results rather than to provide an in-depth economic calculation. We therefore implicitly assume that the cost of feeding the animal is approximately equal to the indirect benefits such as manure, milk, beet; insemination services, etc. from the animal that are not captured in our output measure. collected 3-4 years after a comprehensive set of reforms had been initiated, one observes only a very limited responses from the productive sector. Three main conclusions emerge. First, using panel data methods we can clearly reject the hypothesis that farmers reduced use of purchased inputs because this is no longer profitable - and that therefore government should intervene to improve the functioning of markets and reduce input prices faced by producers. By contrast, the fact that use of purchased inputs is highly profitable points to the importance of non- price factors. Policies that would increase private sector activity in this industry and remove producers' non-price rationing through provision of public goods that increase input demand could have significant productivity benefits. This is consistent with the finding emerging from a large number of participatory assessments according to which unreliable delivery-which is clearly related to Government intervention in physical input distribution- is a more serious problem for producers than the non-affordability of agricultural inputs (Milimo 1995). Given the high returns from increasing access to fertilizer, and the failure of public sector involvement to guarantee such access in the 1992-95 post-liberalization period, there appears little justification for continued Govemment involvemenit in physical distribution of inputs. Second, under Zambian conditions of land abundance, ownership of productive assets is a key constraint to enhanced productivity, increases in cultivated area, greater use of credit and purchased inputs, and better adaptation to the vagaries of climate. Regression coefficients suggest that purchase of a pair of oxen is a profitable venture even at high real interest rates. Greater focus on factors related to accumulation of productive assets (such as lack of knowledge on management practices, threat of infectious diseases, and non-availability of financial infrastructure to obtain longer-term capital) may be a more promising focus for government policy than continued concern about -and self-defeating intervention to ensure- the proper functioning of input markets. The fact that lack of long-term credit facilities to allow acquisition of productive assets such as cattle is one of the key concerns emerging from participatory assessments (Tembo et. al 1995; Francis et al. 1997) supports our analysis and suggests that more detailed study of mechanisms addressing this constraint could have a high payoff. Finally, attempts to increase agricultural productivity and rural well-being through provision of public services have thus far largely been unsuccessful. Education has a very significant impact on input demand but not on agricultural productivity andthe impact of extension is insignificant. Our interpretation is that the constraints facing rural producers in Zambia are still related more to market access and the ability to obtain the necessary inputs in a highly volatile economic environment rather than the application of more productive technology. At the same time, this calls for a thorough assessment of the technology available and the messages that are being disseminated through public technical assistance. Unless these can be adapted more to the prevailing economic environment and the incentive structure facing producers, greater public spending on teclnical assistance may be difficult to justify. From a methodological point of view, our results suggest that farm-level panel information can provide policy-relevant insights that avoid the difficulties inherent in the interpretation of cross- sectional data. While the panel underlying our study was collected more by accident than by design, greater emphasis on collection of panel data, adoption of the necessary protocols to ensure data quality, and inclusion of other household characteristics (e.g. asset endowments) is likely to increase the usefulness of such data to yield policy relevant conclusions. From a substantive point of view, our analysis supports the notion that without "re-emphasizing" the role of agriculture, it will be difficult for African countries to lay the basis for sustained growth and prevent a reversal of policy that would undo much of the progress and potential from recent liberalization. It is now recognized that, eYen though Africa faces specific challenges in the field of health, most of the stagnation of the rural sector has been policy-induced, that the gap between potential and actual output is greatest in Africa, and that demand is unlikely to become a constraint any time soon (McPherson 1999). This reinforces the importance of providing public goods, in addition to price policies, to bring about a sustained agricultural supply response (Schiff and Montenegro 1997). The fact that use of purchased inputs is way below the optimum, that credit constraints prevent investment in productivity-enhancing technology such as livestock, and that public services are not (yet) effective in delivering the needed technology, suggests that much remains to be done to address non-price related constraints to agricultural production, generated and disseminate technology, and thus help rural producers make better use of the resources at their disposal. References: Alwang,-Jeffrey; Siegel,-Paul-B.; Jorgensen,-Steen-L. 1996, Seeking Guidelines for Poverty Reduction in Rural Zambia, World-Development, 24(11) 1711-23. Amemyia, T. and T.E. MacCurdy, (1986), "Instrumental variables estimation of an error components model", Econometrica, 54, 869-880 Breusch, T.S., G.E. Mizon and P. Schmidt, (1989), "Efficient estimation using panel data", Econometrica, 57, 695-700. Deininger, Klaus, Pedro Olinto, Mukata Wamulume, and Dennis Chiwele, "Agricultural sector performance in a post-reform environment: Implications for second generation reforms in Zambia, World Bank, mimeo. Francis Paul A et al. 1997, Listening to farmers: Participatory assessment of policy reform in Zambia's agriculture sector, World Bank Technical Paper 375, Washington, D.C. Gourieroux, C. and A. Monfort (1993). "Simulation-Based Inference: A Survey with Special Reference to Panel Data Models," Journal of Econometrics, 59:5-33. Hattink,-Wolter; Heerink,-Nico; Thijssen,-Geert 1998, Supply Response of Cocoa in Ghana: A Farm-Level Profit Function Analysis, Journal-of-African-Economies;7(3), October 1998, pages 424-44. Hausmann, J. A. and W. Taylor, (1981) "Panel data and unobservable individual effects", Econometrica, 49, 1377-1399 Hazell, Peter B. R.; Hojjati, Behjat 1995, Farm/Non farm Growth Linkages in Zambia, Journal of African Economies;4(3): 1995, 406-435 Holden, Stein T.; Taylor, J. Edward; Hampton, Stephen, 1999, Structural Adjustment and Market Imperfections: A Stylized Village Economy Wide Model with Non-separable Farm Households, Environment and Development Economics; 4(1): 1999, 69-87. Honore, B.E. (1992). "Trimmed LAD and Least Squares Estimation of Truncated and Censored Regression Models with Fixed Effects," Econometrica 60:533-565. Jansen, Doris J and Rukovo, Andrew 1992, Agriculture and the Policy Environment: Zambia and Zimbabwe, Technical Paper # 74 for the research programm on Developing Country Agriculture and International Economic Trends, OECD Development Center, Geneva 1992 Jansen, Doris J. (1991), The Political Economy of Agricultural Pricing Policy: Volume 3: Africa and the Mediterranean: Zambia, in: Krueger, Anne O.;Schiff, Maurice; Valdes, Alberto, eds. The political economy of agricultural pricing policy. Volume 3. Africa and the Mediterranean., Johns Hopkins University Press, Baltimore and London: 268 327. Keyser, J. 1996, Zambia's Comparative Advantage, Report prepared for the World Bank, Lusaka McPherson, M.F. 1995, The Sequencing of Economic Reforms: Lessons from Zambia, Harvard Institute for International Development, Discussion Papers 516, 1995 McPherson, M.F 1999, Sustaining Trade and Exchange Rate Reform in Africa: Lessons for Macroeconomic Management, HIID Development Discussion Papers 713, Cambridge MA Milimo, John T. 1995, An Analysis of Qualitative Information on Agriculture from beneficiary assessments, participatory poverty assessment and other studies which used qualitative research methods, The Participatory Assessment Group, Lusaka mimeo Mundlak, Y., (1996), "Production Function Estimation: Reviving the Primal", Econometrica, 64,431-438. Mwanaumo, Anthony; Masters, William A.; Preckel, Paul V. 1997, A Spatial Analysis of Maize Marketing Policy Reforms in Zambia, American Journal of Agricultural Economics;79(2), May 1997, 514- 23. Rosenzweig, Mark R. 1996, When Investing in Education Matters and When It Does Not, Challenge 39(2) 22-29 Savadogo,-Kimseyinga; Reardon,-Thomas; Pietola,-Kyosti 1995, Mechanization and Agricultural Supply Response in the Sahel: A Farm-Level Profit Function Analysis, Journal-of-African- Economies;4(3), December 1995, pages 336-77. Schiff,-Maurice; Montenegro,-Claudio-E. Aggregate Agricultural Supply Response in Developing Countries: A Survey of Selected Issues, Economic-Development-and-Cultural-Change;45(2):393- 410. Tembo, S, Nkomeshy, A., Ntalasha, H., Kalumbi, J. 1995, Pre-ASIP Study of the Status of smallholder farming and agricultural services in Sinazongwe and Zimba, Southern Province of Zambia, Institute for African Studies, University of Zambia, Working Paper 2, Lusaka World Bank 1992, Zambia Agriculture Sector Strategy: Issues and Options, Washington DC, mimeo Table 1: Descriptive statistics Total Central Copperbelt Eastern Luapula Lusaka Northern Northwest South West Household size 6.19 7.37 5.99 5.93 5.49 6.98 5.93 5.55 8.71 5.58 Female headed 20.52% 16.67% 18.43% 21.35% 21.22% 16.20% 19.47% 18.31% 12.07% 31.22% Access to extension 23.77% 14.08% 22.44% 30.76% 13.43% 24.22% 29.18% 28.34% 27.24% 15.66% Primary education 56.46% 58.11% 50.00% 57.20% 60.10% 47.22% 55.87% 50.00% 66.84% 53.01% Secondary education 14.84% 22.81% 26.28% 8.52% 13.92% 16.67% 17.47% 10.83% 16.71% 11.06% Age 45.44 45.72 44,42 45.40 44.45 45.44 44.34 45.29 46.09 48.31 Area cultivated (hectares) 1.89 2.82 1.58 2.24 1.18 1.72 1.60 1.20 3.30 1.44 Value of production (1000 K) 231.44 407.08 224.36 257.93 139.38 157.37 230.35 171.65 358.46 116.18 Value of sales 90.12 212.61 101.09 92.94 34.78 47.80 92.01 68,27 123.51 37.50 Share of sales 38.94% 52.23% 45.06% 36.03% 24.95% 30.37% 39.94% 39.77% 34.46% 32.27% Posession of caUtle 19.14% 27.85% 9.49% 27.67% 1.10% 21.30% 8.42% 6.37% 60.88% 19.84% Value of cattle (1000 K) 152.62 240.77 69.38 162.71 3.82 172.22 33.43 42.25 712.81 172.64 Ownership of draught animals 14.86% 26.75% 8.03% 22.92% 0.17% 18.52% 2.32% 1,43% 57.56% 12.20% Access to credit 14.53% 21.71% 4.56% 24.58% 9,85% 23.61% 13.51% 5.25% 16.05% 5.45% Amount of credit received (1000 K) 160.75 210.88 165.64 134,23 132.56 81.47 136.63 70.09 281.45 252.27 Use fertilizer 31.21% 61.51% 25.73% 27.42% 17.74% 75.00% 35.58% 19.75% 48.28% 10.24% Value of fertilizer used 80.36 96.42 106.18 108.48 45.92 51.49 58.74 46.25 79.91 118.00 No of observations 9,706 912 548 2,042 1,178 216 2,198 628 754 1,230 1994 Area cultivated 2.01 3.07 1.68 2.34 1.28 1.77 1.71 1.07 3.77 1.53 Value of production (1000 K) 227.93 386.76 208,32 247.76 123.51 184.61 243.52 166.15 376.35 106.27 Value of sales 95.32 207.41 99.35 84.49 32.76 62.89 114.62 73.12 149.09 37.87 Share of sales 41.82% 53.63% 47.69% 34.10% 26.53% 34.07% 47.07% 44.01% 39.62% 35.64% Ownership of cattle 18.61% 27.85% 9.49% 25.47% 1.53% 21.30% 8.55% 5.73% 59.42% 19.84% Value of cattle (1000 K) 127.01 228.55 64.16 141.74 4.79 169.44 28.39 32.10 505.78 157.37 Ownership of draught animals 15.66% 27.19% 8.76% 24.19% 0.34% 18.52% 2.55% 1.59% 58.36% 14.63% Access to credit 19.23% 26.54% 5.47% 29.29% 14.43% 32.41% 21.02% 9.87% 19.89% 6.67% Amount of credit received (1000 K) 31.97 59.29 10.04 41.86 14.02 26.93 30.35 6.87 62.47 20.19 Use fertilizer 34.82% 65.57% 22.26% 31.54% 22.75% 87.04% 40.49% 26.75% 48.28% 11.22% Value of fertilizer used 97.46 112.94 192.83 131.77 51.09 58.63 73.30 53.31 94.41 144.73 1995 Area cultivated 1.76 2.58 1.48 2.14 1.07 1.67 1.48 1.33 2.82 1.35 Value of production (1000 K) 234.96 427.39 240.40 268.10 155.26 130.12 217.19 177.16 340.57 126.10 Value of sales 84.93 217.81 102.83 101.39 36.80 32.71 69.39 63.42 97.93 37.12 Share of sales 36.14% 50.96% 42,78% 37.82% 23.70% 25.14% 31.95% 35.80% 28.75% 29.44% Ownership of cattle 19.68% 27.85% 9.49% 29.87% 0.68% 21.30% 8.28% 7.01% 62.33% 19.84% Value of cattle (1000 K) 178,24 252.98 74.60 183.67 2.85 175.00 38.47 52.39 919.84 187.92 Access to credit 9.83% 16.89% 3.65% 19.88% 5.26% 14.81% 6.01% 0.64% 12.20% 4.23% Amount of credit received (1000 K) 10.31 22.59 3.55 16.90 8.46 8.08 4.60 0.35 19.51 5.11 Ownership of draught animals 14.05% 26.32% 7.30% 21.65% 0.00% 18.52% 2.09% 1.27% 56.76% 9.76% Use fertilizer 27.59% 67.46% 29.20% 23.31% 12.73% 62.96% 30.66% 12.74% 48.28% 9.27% Value of fertilizer used 58.88 77.57 40.11 77.06 36.77 41.71 39.70 31.44 65.57 85.07 Table 2: Summary of regression results Production function equation Land demand equation Within estimator IV Estimator Within estimator IV Estimator Coeff. t-value Coeff. t-value Coeff. t-value Coeff. t-value Ln(area) 0.6292 46.8726 0.7157 37.322 Ln(HH population) 0.1238 5.6355 0.0908 3.857 0.247 12.6558 0.4185 7.6 Value of Cattle (1000 K) 0.0001 9.189 0.0001 8.018 0.00001 1.6613 0.0000 1.208 Credit Received (1000 K) 0.0003 3.1614 0.0004 4.122 0.001 15.2649 0.0016 5.407 Value of Oxen & implements (1000 K) 0.0004 4.2267 0.0001 2.455 0.0002 2.5575 0.0008 8.171 ln(Fert+1) 0.0723 17.1508 0.0755 5.684 Dummy for 95 -0.7321 -2.7064 0.337 2.841 -0.125 -5.982 -0.1683 -2.651 Rainfall deviation I (pre-planting period) 0.114 4.7054 0.0198 1.993 0.0072 6.0029 0.0079 2.715 Rainfall deviation I squared -0.0047 -4.373 -0.0005 -1.071 -0.0001 -3.8391 -0.0002 -3.056 Rainfall deviation I cubic -0.0001 -4.4121 0.00002 -1.267 Rainfall deviation I fourth power 0.00002 4.4256 0.00001 1.263 Rainfall deviation 11 (growing period) 0.0056 6.2458 0.0019 3.876 Rainfall deviation Illi (harvest period) -0.0092 -3.687 0.0005 0.511 Rainfall deviation I x Fertilizer -0.0008 -6.1354 -0.0004 -1.928 Rainfall deviation I x Oxen -0.00002 -2.0854 -0.00004 -3.001 -0.00004 -6.7952 0.0000 -0.821 Rainfall deviation I x Oxen squared 0.00001 5.5469 0.00003 2.426 Rainfall deviation I x HH pop 0.0001 0.235 -0.001 -1.207 -0.0017 -2.8529 -0.0015 -1.231 Rainfall deviation I x HH pop squared 0.00001 1.9335 0.0001 2.503 Female headship I dummy -0.0691 -0.703 -0.4231 -1.512 Distance to market -0.003 -0.172 -0.0915 -1.725 District Extension % 0.0109 1.456 -0.0223 -1.392 Primary education -0.0463 -0.099 -1.0837 -0.863 Secondary education 0.3193 0.668 0.5431 0.585 Table 3: Demand for Fertilizer / ha. EXPLANATORY VARIABLES SML estimator TLS estimator ESTIMATES STD. ERRORS ESTIMA TES STD. ERRORS Intercept -3.01** 0.22 Time Trend (T) -0.41** 0.12 -0.59** 0.13 Log of HH adult population (POP) 1.01** 0.09 0.54- 0.18 Credit Received (1000 Ks) 0.48** 0.02 0.13* 0.07 Cattle Stock (1000 Ks) 0.05 0.07 0.11 0.10 Draught Animals and Implements (1000 Ks) 1.21** 0.25 0.46 0.77 DM01 -0.01** 0.00 -0.04 0.04 Sex of HH's head (Women = 1) -0.68** 0.13 Distance to Markets -0.22** 0.02 Extension -0.15 0.20 Primary education -0.09** 0.12 Secondary Education 1.53 0.11 3.34** 0.01 Mean Log- Likelihood -1.10792 Number of Observations 9706 9706 - Indicates statistically different from zero at the 10% significance level. Indicates statistcally different from zero at the 5% significance level. Table 4: Credit Use Equation EXPLANATORY VARIABLES SML estimator TLS estimator Random effect probit ESTIMATES STD. ESTIMATES STD. ESTIMATES STD. ERRORS ERRORS ERRORS Intercept -450.65** 36.94 Time Trend (T) -161.94** 13.98 -170.48** 28.97 Log of HH adult population (POP) 113.35** 13.69 112.22** 56.31 0.335365** 0.042109 Cattle Stock (1000 Ks) -37.44** 13.98 -56.11** 29.13 -0.00018** 0.000046 Draught Animals & implement (1000 K) 386.62** 56.84 333.18 236.07 0.000976** 0.00014 Sex of HH's head (Women = 1) 94**-0.37624** 0.065369 Primary education 5.80** 1.54 -0.12828** 0.009673 Secondary Education 64.80** 15.93 0.005 0.057659 Distance to Markets -34.98** 2.94 0.170135* 0.054838 s2 359.20** 19.83 0.6915 0.0485 Mean Log- Likelihood -1.27 Number of Observations 9706 9706 9706 Indicates statistically different from zero at the I0% significance level. Indicates statistically different from zero at the 5% significance level. Policy Research Working Paper Series Contact Title Author Date for paper WPS2277 Determinants of Bulgarian Brady Nina Budina January 2000 N. Budina Bond Prices: An Empirical Tzvetan Mantchev 82045 Assessment WPS2278 Liquidity Constraints and Investment Nina Budina January 2000 N. Budina in Transition Economies: The Case Harry Garretsen 82045 of Bulgaria WPS2279 Broad Roads in a Thin Country: Andr6s G6mez-Lobo January 2000 G. Chenet-Smith Infrastructure Concessions in Chile Sergio Hinojosa 36370 WPS2280 Willingness to Pay for Air Quality Hua Wang January 2000 R. Yazigi Improvements in Sofia, Bulgaria Dale Whittington 37176 WPS2281 External Sustainability: A Stock C6sar Calder6n January 2000 H. Vargas Equilibrium Perspective Norman Loayza 38546 WPS2282 Managing Fiscal Risk in Bulgaria Hana Polackova Brixi January 2000 L. Zlaoui Sergei Shatalov 33100 Leila Zlaoui WPS2283 New Tools and New Tests in Thorsten Beck February 2000 P. Sintim-Aboagye Comparative Political Economy: George Clarke 38526 The Database of Political Institutions Alberto Groff Philip Keefer Patrick Walsh WPS2284 The Use of Asset Management Daniela Klingebiel February 2000 R. Vo Companies in the Resolution of 33722 Banking Crises: Cross-Country Experience WPS2285 Industrial Environmental Performance Susmita Dasgupta February 2000 Y. D'Souza in China: The Impacts of Inspections Beno-it Laplante 31449 Nlandu Mamingi Hua Wang WPS2286 Transparency, Liberalization, Gil Mehrez February 2000 D. Bouvet and Banking Crises Daniel Kaufmann 35818 WPS2287 The Vicious Circles of Control: Raj M. Desai Fsbruary 2000 S. Cox Regional Governments and Insiders itzhak Goldberg 36633 in Privatized Russian Enterprises WPS2288 Ten Years of Transformation: Charles Wyplosz February 2000 iV. Jandu Macroeconomic Lessons 33103 Policy Research Working Paper Series Contact Title Author Date for paper WPS2289 Exchange Rate Overvaluation and Howard J. Shatz February 2000 L.Tabada and Trade Protection: Lessons David G. Tarr 36896 from Experience WPS2290 Decentralization and Corruption: Raymond Fisman February 2000 E. Khine Evidence across Countries Roberta Gatti 37471 WPS2291 Incentives for Pollution Control: Jerome Foulon February 2000 Y. D'Souza Regulation and Public Disclosure Paul Lanoie 31449 Benoit Laplante WPS2292 Dividing the Spoils: Pensions, Ethan B. Kapstein March 2000 P. Sader Privatization, and Reform Branko Milanovic 33902 in Russia's Transition WPS2293 Should Capital Flows be Regulated? Roumeen Islam March 2000 R. Islam A Look at the Issues and Policies 32628 WPS2294 Reforming the Urban Water System Mary M. Shirley March 2000 P. Sintim-Aboagye in Santiago, Chile L. Colin Xu 38526 Ana Maria Zuluaga WPS2295 Resolving Bank Failures in Argentina: Augusto de la Torre March 2000 M. De Loayza Recent Developments and Issues 38902 WPS2296 An Ecological and Historical Yujiro Hayami March 2000 P. Kokila Perspective on Agricultural 33716 Development in Southeast Asia WPS2297 Sources of Ethnic Inequality Dominique van de Walle March 2000 H. Sladovich in Vietnam Dileni Gunewardena 37698 WPS2298 Fiscal Deficits, Monetary Reform, Nina Budina March 2000 N. Budina and Inflation Stabilization in Romania Sweder van Wijnbergen 82045 WPS2299 Household Savings in Transition Cevdet Denizer March 2000 A. Cubukcu Economies Holger C. Wolf 38449 Yvonne Ying WPS2300 Single Mothers in Russia: Household Michael Lokshin March 2000 P. Sader Strategies for Coping with Poverty Kathleen Mullan Harris 33902 Barry Popkin WPS2301 Identifying Welfare Effects from Martin Ravallion March 2000 P. Sader Subjective Questions Michael Lokshin 33902