The impact of food price shocks in Uganda: first-order effects versus general-equilibrium consequences

Van Campenhout, Bjorn,; Karl, Pauw,; Nicholas, Minot,

doi:10.1093/erae/jby013

The impact of food price shocks in Uganda: first-order effects versus general-equilibrium consequences

Van Campenhout, Bjorn, ;Karl, Pauw,;Nicholas, Minot, 2018-12-01 00:00:00 Abstract For developing countries, whose governments are faced with volatile world food prices, the appropriate policy response hinges on who are the likely winners and losers. Therefore, it is necessary to predict the impact of higher commodity prices on different subgroups of society. We compare the results of a method that is popular with policy makers because of its parsimony and ease of interpretation with the results of a more complex and data-intensive general-equilibrium model. Using historical prices between 2008 and 2011 for Uganda, we find that both methods predict high prices benefit poor rural farmers, but more so if a more elaborate model is used. 1. Introduction The world food price crisis of 2007–2008 has been an eye opener for many governments in developing countries. In response to the ensuing social unrest and political instability, governments have been scrambling to manage the local impact of global food price surges. However, higher food prices may also signal new economic opportunities. To guide the choice of the appropriate policy instruments, governments must determine who are the likely winners and losers of food price surges. This is essentially an empirical question that depends on production and consumption patterns among households as well as the pass-through of global commodity prices (Benson, Mugarura and Wanda, 2008). Estimates of who wins and who loses also depend on the complexity of the models used, which in part is influenced by the availability of data and analytical capacity. At one extreme, the impact of a price increase of a commodity at the household level can be estimated using a first-order approximation of the welfare impact.1 The first-order welfare impact of a price change is the impact without taking into account any response by economic agents to price changes. As such, it can be considered the ‘short-term’ or ‘immediate’ welfare impact. Modest data requirements and a simple formula with a straightforward interpretation make this the workhorse model among many government agencies within developing countries (Headey, 2016). At the other extreme, the impact of a price increase of a commodity is assessed through full-scale computable general-equilibrium (CGE) models. These models trace the impact of an exogenous price shock through the entire economy, incorporating adjustments in consumption and production decisions of households, secondary effects through the labour market, capital and land market consequences, and so on. These models are much more data intensive and require specialised software and advanced modelling skills (Benson et al., 2013). The empirical literature suggests that studies of food price hikes that rely on more elaborate models arrive at more positive (or less negative) welfare conclusions than studies that only consider first-order effects (Headey and Martin, 2016). A recent multi-country assessment of first-order welfare impacts of food price hikes during the second half of 2010 by Ivanic, Martin and Zaman (2012) finds poverty increased by an average 1.1 per cent in low-income countries. For Uganda, Benson, Mugarura and Wanda (2008) evaluate the immediate welfare impact of the 2008 food price crisis and conclude that the impact was likely to be negative, but small. A similar study by Simler (2010) finds the food price crisis increased poverty in Uganda by 2.6 percentage points. A recent example of a study that goes beyond first-order impacts is Jacoby (2016), who finds that in India, rural households across the income spectrum benefit from higher agricultural prices, mainly due to wage effects. For Uganda, Matovu and Twimukye (2009) use a CGE model to look at the effects of rising cereals prices. Their simulated price increases lead to large improvements in well-being. Boysen and Matthews (2012) run an integrated CGE–micro simulation model to analyse the 2006–2008 increase in commodity prices in Uganda. Their results suggest increased commodity prices reduced national poverty by about 2.7 percentage points. Inspired by Robinson et al. (1999), in this paper, we juxtapose the kind of stylised models prevalent in policy environments in developing countries with an ideal situation where economy-wide effects are incorporated in the analysis. In particular, we present the results of both a first-order analysis and a long-run general-equilibrium analysis based on the same case study: historical price movements of a set of key commodities in Uganda over the period 2008–2011. This article contributes to the study the effect of food price changes on welfare in several ways. First, to our knowledge, this is the first study that runs both a first-order effects analysis and a complete CGE analysis on the same case.2 While previous studies suggest substantially more positive welfare effects from price increases when long-run general-equilibrium models are used than when a first-order approach is taken, it is still unclear how much of this difference should be attributed to the difference in method and how much to the fact that different case studies are used. In addition, instead of looking at a hypothetical price increase of one particular commodity, we try as much as possible to base both analyses on actual, historical price movements of a range of commodities. As such, our simulations also serve as an impact evaluation of price movements, but controlling for other external shocks that may have affected socio-economic outcomes during the same period. This study also adds to the available evidence of the welfare impact of food prices in Uganda, where this is a politically contentious issue (Bellemare, 2015). Finally, by comparing the pragmatic first-order approach to a comprehensive, economy-wide analysis, we also aim to provide an additional explanation for the often opposing views held on the costs and benefits of higher commodity prices (Swinnen and Squicciarini, 2012). The remainder of this article is organised as follows. The next section presents the first-order analysis, including sub-sections on the methodology, data and results. Section 3 presents the long-run general-equilibrium analysis, including descriptions of the social accounting matrix (SAM), behavioural assumptions underlying the CGE model, scenarios we simulate, closure rules and results. The fourth section concludes. 2. First-order impact of food prices on household welfare 2.1. A welfare compensation index Farm households in developing countries characterised by semi-substance agriculture typically participate in the market as both buyer and seller for a range of commodities. Household surveys show that, in many developing countries, many rural households (often a majority) are net buyers of major crops, even staple food crops (Weber et al., 1988). In this context, Deaton (1989) proposed a simple first-order approximation of the welfare impact of price changes in a study of rice farmers in Thailand, who both produce for the market and consume from the market, using data from a standard household survey: Δyy=(rpy−qpy)Δpp (1) where y is a measure of household welfare, p is the price of the commodity, r is the quantity of sales and q is the quantity of purchases. The expression in parentheses is called the net benefit ratio (NBR) of a commodity and can be considered the short-term elasticity of welfare with respect to the commodity price. If the household is a net seller of the commodity, the NBR is positive, and price increases are associated with higher welfare. If the household is a net buyer, the NBR is negative and higher prices mean lower welfare. Deaton’s expression has been extended to include second-order terms (accounting for household response to price changes), to distinguish between changes in producer and consumer prices and to measure the welfare effect of the price changes in multiple commodities (Minot and Goletti, 1998; Boysen and Matthews, 2012; Dawe and Maltsoglou, 2014; Tiberti and Tiberti, 2016). In this study, we extend Deaton’s expression to multiple commodities and allow differences in the proportional changes in producer and consumer prices, as shown in the following equation3: Δyy=∑iriwiyΔwiwi−∑jqjpjyΔpjpj (2) where ri is the quantity of commodity i sold, wi is its sale price and pj is now the price of purchased good j. The first term on the right side sums the share of income from each commodity (riwi/y) multiplied by the corresponding proportional change in producer prices (⁠ Δwi/wi ⁠). The second term on the right side sums the share of the budget spent on each commodity consumed (qjpj/y) multiplied by the corresponding proportional change in consumer price (⁠ Δpj/pj ⁠). Equation (2) is calculated for each household and then aggregated to some level. For instance, one can simply take the average to get the overall first-order proportional impact on welfare. However, it is often more instructive to compare the average impact of different groups of households that have different consumption and production patterns, such as urban and rural households or farm and non-farm households (Benson et al., 2013). Most studies using Deaton’s method adopt a before-and-after approach, where price changes are calculated as the difference between price levels of all commodities sold and bought at two distinct points in time (⁠ Δwi=wi,t=1-wi,t=0 and Δpj=pj,t=1-pj,t=0 ⁠, where wi,t=0 and pj,t=0 are, respectively, the producer and consumer prices before the price change and wi,t=1 and pj,t=1 are the producer and consumer prices after the price change). These studies are very sensitive to the choice of these points in time. Often, the choice of points is arbitrary, is determined by data availability, or represents extreme situations (e.g. the lowest versus the highest prices over a certain period). In general, however, price data are much more readily available than data on quantities purchased and sold, which are typically derived from household surveys. The availability of time series data on prices allows one to define price changes over a range of periods, fixing baseline price levels at one point in time and calculating the difference with various end points. Evaluating equation (2) at different end points in time allows one to create a ‘welfare compensation index’. As noted above, a standard Deaton first-order welfare analysis, as represented by equation (1) above, quantities sold in the market (r) and quantities purchased through the market (q) are assumed to be fixed and measured before the price change happens. This measure does not take into account changes in household behaviour as a result of the price change, even though it is very likely that households will buy less if a commodity becomes more expensive and sell more if the price increases. It is an approximation of reality that is most appropriate in the very short run, for small price changes, or for commodities with inelastic supply and demand. Furthermore, it neglects changes in factor markets and other indirect effects of the price change (Ferreira et al., 2013). As one may argue that adding second-order effects to the short-run analysis would make our application more realistic, we want to reiterate that we want to compare results obtained from a simple method that is very popular among policy makers because of its ease of use and light data requirements to a more elaborate empirical assessment of the impact of price changes. While second-order estimates of the welfare impact are relatively easy to calculate, they do rely on estimates of the elasticity of supply and demand for the relevant commodities. In addition, particularly when working with staple crops for which supply and demand are usually assumed to be inelastic (around −0.3 and 0.3), our experience suggests that adding second-order effects does not make much difference (e.g. Minot and Dewina, 2015). The first-order method also abstracts from several other real-life features. For instance, it is assumed that all households face the same changes in consumer and producer prices. However, as already mentioned in footnote 3, the existence of marketing margins drives a wedge between consumer and producer prices. In addition, the dates for prices at which commodities are bought and sold are assumed to be the same, while in reality, farmers often sell at one point in time and buy at another (Stephens and Barrett, 2010). As such, for groups of farmers that are particularly affected by intertemporal and spatial price dispersion, such as farmers living in remote areas that sell most of the harvest immediately after the harvest and are forced to turn to the market in the lean season, the metric in equations (1) and (2) may provide a poor description of what actually happens to their welfare (Dawe and Maltsoglou, 2014; Minot and Dewina, 2015; Van Campenhout, Lecoutere and D’Exelle, 2015). The extent to which spatial and intertemportal price dispersion can be accounted for depends on data availability. In this study, we have access to prices collected in different locations, allowing for some degree of spatial price dispersion. 2.2. Data We use two data sources for our application. For producer prices (wi) and consumer prices (pj), we use data from FIT-Uganda, a business development consultancy that runs the Infotrade market information service. FIT-Uganda collects prices for 44 commodities from 23 towns and cities throughout Uganda three times a week. For most commodities, both wholesale (which we use to proxy producer prices) and retail prices (which we use as consumer prices) are available.4 We include 10 of the most important agricultural commodities in Uganda, and aggregate the series to a monthly frequency. For the quantities produced (ri) and consumed (qj), we use the 2009/10 wave of the Uganda National Panel Survey, which has a detailed agricultural module. Given the importance of prices in a first-order analysis, it may be useful to start by describing the price data we will use. The data have three dimensions: time, location and commodity. Figure 1 graphs the time dimension (2008–2011), averaging over space and commodities. Although the average price level during the analysis period is 1,076 Ugandan shillings (UGX) for wholesale prices and UGX 1,256 for retail prices, such an average was clearly not representative anymore in 2011. World prices for staple foods were on the rise between 2006 and 2008 and accelerated sharply in the beginning of 2008. While food prices dropped significantly by the end of 2008, they rose again in 2009 and remain high by historical standards in both international and local markets (Headey, Malaiyandi and Fan, 2010). Thus, the 2011 spike happened on top of historically high prices. Another interesting feature is the difference between wholesale and retail prices. Although prices at both the wholesale and the retail level rose sharply at the beginning of 2011, wholesale prices retracted more than retail prices. In fact, around May 2011, retail prices and wholesale prices seemed to move in opposite directions. A widening gap between wholesale and retail prices is likely to hurt farmers who participate in the market, as their revenue from sales decreases whereas the cost of purchases increases. Fig. 1. View largeDownload slide Average prices of three commodities over time. Source: Authors’ calculations based on FIT Uganda. Note: The figure shows the average over commodities (matooke, maize and groundnuts) and five markets. Fig. 1. View largeDownload slide Average prices of three commodities over time. Source: Authors’ calculations based on FIT Uganda. Note: The figure shows the average over commodities (matooke, maize and groundnuts) and five markets. Figure 2a shows the evolution of prices of three unprocessed staple foods over time (averaged over the different market locations). Since these are all staple foods, one would expect these prices to move closely together as substitution in consumption keeps the prices from diverging too much. However, there were some episodes when prices for different staple crops moved in opposite directions. For instance, from June 2009 to November 2009, maize became cheaper whereas cassava became more expensive. This is likely to affect the terms of trade of households and can be detrimental to households that sold maize and bought cassava with the revenues. Figure 2b shows the evolution over time of higher value or lightly processed commodities. Also here, there are marked differences. The price of groundnuts increased significantly over the period, whereas the price of maize flour increased only moderately. Fig. 2. View largeDownload slide Price evolution of selected crops. Source: Authors’ calculations based on FIT Uganda. Note: Figures show the average of wholesale and retail prices and average across five markets. Fig. 2. View largeDownload slide Price evolution of selected crops. Source: Authors’ calculations based on FIT Uganda. Note: Figures show the average of wholesale and retail prices and average across five markets. 2.3. Welfare and poverty impact of price changes Figures 3a plots the percentage change in welfare, proxied by consumption expenditure per capita, as a result of actual price changes between June 2008 and December 2011. In particular, we calculate the metric defined in equation (2) and then take the average over all households. Since we have monthly data for prices between 2008 and 2011, we can calculate equation (2) evaluated at prices in each month over this period relative to some baseline price level, hence the results are plotted as a time series.5 In this way, we do not have to choose an endpoint at which to evaluate the price evolutions, but we can investigate what would have been the outcome if, for instance, the government of Uganda decided to evaluate the impact at a particular point in time. Fig. 3. View largeDownload slide Welfare impact index and short-run poverty. Source: Authors’ calculations based on FIT Uganda and UNPS 2009/10. Fig. 3. View largeDownload slide Welfare impact index and short-run poverty. Source: Authors’ calculations based on FIT Uganda and UNPS 2009/10. We see that if the government would evaluate the evolution of prices anywhere between September 2008 and June 2009, the first-order impact would have yielded a positive welfare impact. However, effects are generally less than 1 per cent of welfare. If wi,t=1 and pj,t=1 are fixed after June 2009, welfare changes turn negative again, and become larger over time, reaching about 2 per cent for the comparison between the baseline period and September 2010. The year 2011 is volatile with a welfare increase of more than 2.5 per cent comparing the baseline with March 2011 and a welfare loss of 2.5 per cent comparing the baseline with December 2011. The evolution of the welfare impact over time is in line with the movement of the price series in Figures 1 and 2. Episodes where prices are higher than at baseline, such as around March and April 2011, correspond to an increase in welfare, suggesting many farmers are net sellers of the affected commodities. Note also that at the end of the sample, wholesale prices decrease while retail prices level out. At the same time, the price of cassava, a crop bought by the poor, shoots up, while the price of maize, the main income earner for smallholders, levels out. Together, this explains the sharp reduction in first-order welfare impact between the baseline and December 2011.6 To calculate the impact of price changes on poverty, we use the first-order welfare impact to calculate real per capita consumption expenditure for each household at different sets of prices. In particular, for each household, we add (∑iriΔwi-∑jqjΔpj) to per capita consumption expenditure at baseline price levels, and compare this to the official poverty line in Uganda. Figure 3b plots the poverty headcount, defined as the percentage of people that fall below the poverty line, for each monthly set of prices. In the reference period, headcount poverty stands at 23.1 per cent, represented by the horizontal line. The correspondence between the welfare changes depicted in Figure 3a and the poverty rates seem reasonable, with positive first-order welfare impacts corresponding to poverty reductions, and vice versa. The exception seems to be when the price situation around the beginning of 2011 is considered the endpoint of a first-order analysis. A sharp increase in welfare seems to coincide with an equally sharp increase in poverty. This suggests that the price pattern in the beginning of 2011 benefited richer farmers, while at the same time hurting farmers that were just above the poverty threshold. In other words, there is considerable heterogeneity in the price impact of price changes on individual farmer’s well-being and position relative to the poverty line. We can explore some of this heterogeneity a bit further by looking at the first-order effects of price changes for particular groups of people. Because urban households are generally net buyers and would be more adversely affected by higher food prices, we first divided the sample into rural and urban households and consider the evolution of welfare separately (see Figure 4a). However, since there is a great deal of agricultural activity in peri-urban areas in Uganda (and to facilitate comparison to the CGE results below that use similar groups of representative households), we also distinguish rural farmers and urban non-farming households who reside either in secondary towns/cities or in Kampala metropolitan area. Results are in Figure 4b. Fig. 4. View largeDownload slide Welfare impact for different types of households. Source: Authors’ calculations based on FIT Uganda and UNPS 2009/10. Fig. 4. View largeDownload slide Welfare impact for different types of households. Source: Authors’ calculations based on FIT Uganda and UNPS 2009/10. Figure 4a shows that, for most of the period, recorded price levels have benefited, or at least not hurt, the rural population as judged by the first-order effects and relative to average price levels of 2008. It seems that, in general, the higher prices observed between 2008 and 2011 were beneficial to rural households, suggesting many of them are net sellers in the market for the commodities that saw the highest price hikes and net buyers for commodities for which prices increased less dramatically.7 Especially, the price hike in the beginning of 2011 resulted in large welfare impacts of over 5 per cent of household welfare. Only in mid-2010, when prices were lowest, we find a negative first-order impact for the rural population, but the magnitude is small. As expected, urban households were adversely affected by high prices, as they are generally less likely to sell on the market. Losses are substantial, reaching well over 10 per cent if December 2011 is taken as the endpoint. Figure 4b shows that these losses become more important for urban dwellers that are not involved in farming. 3. General-equilibrium impact of food price shocks on the Ugandan economy 3.1. Model and data CGE models are particularly suited to the analysis of shocks with complex economy-wide implications, including direct (first-order) and indirect (second-order) effects that become particularly relevant under a longer time frame. First, they account for behavioural responses – primarily to relative price changes – and interaction between producers, consumers and government within a whole-economy framework that includes markets for labour, capital and commodities. Second, through simulation, the impact of complex policy or economic shocks can be better understood and decomposed, e.g. by sector/product or household type, or in the short or longer run, with the latter characterised by assumptions of greater flexibility in producers’ or consumers’ responses to changing market conditions, particularly relative price shocks. Third, by tracking changes in prices, outputs and household incomes, these models – particularly more disaggregated ones – provide a theoretically consistent framework for analysing the welfare effects of policies and external shocks. For this study, we calibrate the International Food Policy Research Institute (IFPRI) standard CGE model (Löfgren, Harris and Robinson, 2002) to a 2007 SAM for Uganda developed by Thurlow (2012). SAM accounts define the economic agents and markets included in the CGE model, such as productive activities, commodity markets, factor markets, current accounts of domestic institutions (i.e. households, government and incorporated business enterprises), a capital account that captures savings and investments and a rest-of-the-world account. The SAM includes 50 activities and 50 commodity accounts, including 21 subsectors in the agriculture, forestry and fishing sectors as well as five agro-processing sectors. There are also six types of production factors (family farm labour, unskilled labour, skilled labour, capital, cattle stock and land) and five representative household groups (rural farm, rural non-farm, urban farm, urban non-farm and Kampala households). 3.2. Model behavioural assumptions While SAM entries record transactions that took place between economic agents during a particular year, the CGE model itself is defined by behavioural relationships. Economic relationships are modelled as a mix of nonlinear and linear equations that govern how economic agents in the model respond to changes in market conditions brought about by exogenous policy or economic shocks. A detailed model description appears in Löfgren, Harris and Robinson (2002); below we focus briefly on some of the key features of the model. In the CGE model, we assume that producers (activities) maximise profits when combining intermediate inputs with primary factors of production such as land, labour and capital. Production is specified using nested constant elasticity of substitution (CES) functions. These reflect sector-specific technologies and allow for imperfect substitution between factors of production. Factor incomes are distributed to their owners. Income from self-employed farm labour, land and cattle stock accrues to farm households, while income from the remaining factors accrues to non-farm and urban households. Agricultural land is an interesting case worth highlighting. We assume that all returns to land accrue to those who actively farm the land, irrespective of whether they own or rent the land. In cases where land is rented from land owners, who may or may not be farmers themselves, rental income is channelled via the capital account in the same way any investment income is channelled. Households also earn income from government transfers or in the form of remittances from other households or from abroad. Households save and pay taxes and use the balance of income for consumption expenditure. The SAM was constructed on the assumption that all consumption is marketed consumption, i.e. home consumption is valued at market prices. Consumption demand is governed by a linear expenditure system (LES) which assumes that a certain portion of disposable income is allocated to subsistence consumption that is unresponsive to price changes. During calibration of the model, we arbitrarily set the Frisch parameters to −1 for all households, and use income elasticities estimated by Wiebelt et al. (2011), which results in approximately 14 per cent of disposable income being allocated to subsistence consumption in the base. Producers supply their output to national product markets. Transaction costs and consumption taxes separate producer and consumer prices. A constant elasticity of transformation (CET) function allows production to shift imperfectly between domestic and foreign markets depending on the relative prices of exports and domestic products. Similarly, a CES function governs the mix of imported and domestically supplied goods demanded locally, depending on relative prices of imported or domestically produced goods. Elasticity values for CES and CET functions are taken from Wiebelt et al. (2011). In order to facilitate a more nuanced analysis of household poverty, we link our modelled changes in prices and consumption at the representative household group level to corresponding member households in the Uganda National Household Survey (2005/06). A survey-based poverty module, a standard feature of IFPRI CGE models, then computes changes in income poverty measures by applying model-estimated percentage changes in consumption for household groups to consumption reported by individual households in the survey. The initial and new consumption levels are compared against an official poverty line, defined for rural and urban households, and estimates of changes in poverty are obtained. We report poverty changes at national level and for various household subgroups. This allows for a direct comparison with the first-order welfare impact estimated in the previous section. At least one perceived limitation of CGE models is their assumption that information flows freely and markets perform well. This is not completely unrealistic for agriculture and food markets in a country such as Uganda, where there are many producers (sellers) and consumers (buyers) and where markets are reasonable well integrated spatially. Standard CGE models also capture some rigidities in markets: for example, in instances where traders extract significant rents due to lack of information or market access by farmers, these are reflected in high trade and transport margins. As discussed later, we restrict mobility of some factors of production to better represent the functioning of factor markets. 3.3. Modelled simulations The main objective of the CGE simulations is to explore the impact of agricultural, food and fuel price changes over the period 2008–2011 relative to a base case scenario representing the economy of Uganda in 2007. As a small country, Uganda is considered a price taker in world markets. Domestic prices, on the other hand, are endogenously determined and can be influenced by exogenous changes in world prices (via trade linkages), policy shocks that alter domestic demand and supply relationships, or economic shocks that affect costs of production. Whereas our world price shock simulations are based on actual observed price changes over 2008–2011, we cannot exogenously replicate observed domestic price changes, since these are endogenously determined in the model. The four simulations presented in this study (sim1–sim4) are run sequentially, each building on the previous simulation by adding a new dimension to the simulated shock. Simulation results are reported as changes relative to the base case in 2007, but given the simulation design, results can also be compared against previous simulations to identify the marginal change of each additional shock introduced. In the first simulation, sim1, foreign currency-denominated world import and export prices of agricultural and processed food commodities are shocked exogenously, based on observed world price changes during 2008–2011. In particular, we model a 50 per cent increase in world maize prices; a 25 per cent increase in rice and other cereals prices; a 35 per cent increase in processed grain prices (flour); and a 25 per cent increase in processed meat, fish and other food products. The second simulation, sim2, implements the same food price shock as sim1, but also increases world prices of key export products, specifically coffee prices (100 per cent) and tea, cocoa and vanilla prices (300 per cent). This particular simulation, therefore, captures the combined effect of observed changes in international food and cash crop prices for both imported and exported commodities. The third simulation, sim3, models the commodity price shocks of sim2 and adds to this a 6 per cent rise in global fuel prices. Fuel prices have been identified as a major cause of commodity price increases globally (Headey, Malaiyandi and Fan, 2010). Domestically, they cause transport cost increases, which tend to be passed on to consumers. For example, UBOS (2012) attributes inflation of 18.7 per cent in 2011 partly to food, beverage, and clothing price increases, but also to high transport costs (transport fares increased by 17.8 per cent that year), which in turn have been linked to rising fuel prices. World fuel prices spiked in 2008 but subsequently retreated to lower levels again, so that the net increase during 2008–2011 was only 6 per cent. The fourth simulation, sim4, simulates the effect of a 7.6 per cent decline in the availability of agricultural cropland, together with the world price shocks of sim3. Although in the static version of the model we do not account for population changes over time, FAOSTAT (2016) data show that agricultural land supply declined by 7.6 per cent over 2008–2011 in per capita terms, which reflects the fact that Uganda is nearing (or may have reached) its land frontier (see Jayne, Chamberlin and Headey, 2014). This simulation, therefore, serves as a proxy for declining per capita availability of land. Although limited mechanisation and rising rural–urban migration also have implications for agricultural production, land constraints are arguably the most pressing issue in the current context in Uganda, hence the focus on this production factor. Given the volatility and heterogeneity in yield growth rates over 2008–2011 (FAO, 2016), we do not introduce productivity shocks into this simulation. The simulation, therefore, assumes a constant level of total factor productivity, but declining land availability, which will lead to an overall decline in crop production and endogenous price increases, subject to producers’ land allocation choices. Combining these domestic price shock simulations with international price shocks complements earlier CGE analyses of price shocks in Uganda, which have only focused on international prices (e.g. Boysen and Matthews, 2012; Matovu and Twimukye, 2009). 3.4. Closure rules Our analysis focuses on the medium to longer run in that we assume farmers are able to reallocate land to more profitable crops in response to the shocks imposed. We further assume that labour factors are fully employed and mobile across economic sectors, although the movement of family farm labour is restricted to agricultural subsectors to emphasise the difficulty family farm workers have in accessing non-farm employment opportunities. Capital stock, employed largely in non-agricultural sectors, is activity-specific or immobile to impose some rigidities on non-agricultural sectors’ production capacity levels. The overall level of capital stock is also fixed in static CGE models. This means that even as investment levels change, which is determined by the level of savings in the economy (i.e. a savings-driven investment closure), we do not consider the future impact this may have on overall levels of capital stock and production capacity. Foreign savings are fixed exogenously, but the trade balance and the exchange rate are flexible, with the latter serving the role of equilibrating variable to the current-account balance. The domestic consumer price index (CPI) is treated as the numeraire in the model, which means all endogenous price changes are expressed in real terms and relative to a fixed CPI. We, therefore, model the real economy and implications of relative price changes, but not the effect of inflation itself. One advantage of CGE models is the relative ease with which additional simulations or robustness checks under alternative closure rules can be conducted. For example, under an investment-driven savings closure, which assumes households adjust their savings rates to ensure a target level investment is achieved, the macroeconomic results would look similar, but household-level welfare is affected as consumption is shifted towards current investment. As a result of space constraints, and given the emphasis on comparing results across different modelling approaches, we only present results from a limited set of simulations under a single set of closure rules and assumptions which we feel are most appropriate to the situation at hand, although this remains a subjective decision. 3.5. Results and discussion Rising world prices tend to cause shifts in consumer demand towards cheaper, locally produced substitutes, while producers shift production toward more profitable export markets. Such structural shifts may sometimes result in a reduction in domestic absorption (ABS), defined as the sum of consumption (C), investment (I) and government spending (G), and considered a measure of aggregate domestic welfare. The effects of rising exports (X) and declining imports (M) are an improvement in the trade balance (X – M). Simple reorganisation of the GDP equation (3) shows why this will generally lead to a decline in domestic absorption – and hence aggregate domestic welfare – for a given level of GDP (Arndt et al., 2008): GDP=C+I+G+(X–M)=ABS+(X–M)ABS=GDP–(X–M) (3) However, if GDP rises by more than the increase in the trade surplus, absorption may still increase. Although this is fairly unlikely in instances where a country’s terms of trade worsens, it may be likely if the terms of trade improve and increased export opportunities lead to an increase in household disposable income and hence domestic absorption. Absorption may also increase as a result of rising government expenditure and/or investment expenditure. Table 1 reports key macroeconomic results from our simulations, including GDP measured at market prices. Table 2, in turn, presents GDP at factor cost (or value-added), both nationally and for individual sectors. We also present changes in factor incomes by type of factor. Commodity taxes and trade and transport margins explain the difference in GDP at market prices and GDP at factor costs. Table 1. Changes in GDP at market prices and macroeconomic indicators Changes relative to base (%) World food prices (sim1) Food & export crop prices (sim2) Food, export & fuel prices (sim3) Price shocks & domestic supply shock (sim4) Macroeconomic indicators Terms of trade (ToT) 3.2 16.3 15.9 15.9 Real exchange rate −5.5 −7.3 −7.0 −6.7 World price index 3.7 8.2 8.5 8.5 Domestic price index 0.1 0.7 0.6 0.0 Real GDP at market prices (% change) −0.3 0.0 −0.1 −0.9 Absorption 0.2 2.7 2.5 1.6 Consumption 0.6 4.3 4.0 2.7 Investment −1.3 −1.9 −1.9 −1.6 Government 0.0 0.0 0.0 0.0 Exports 1.8 −2.5 −2.4 −3.1 Imports 2.6 8.4 8.0 7.3 Changes relative to base (%) World food prices (sim1) Food & export crop prices (sim2) Food, export & fuel prices (sim3) Price shocks & domestic supply shock (sim4) Macroeconomic indicators Terms of trade (ToT) 3.2 16.3 15.9 15.9 Real exchange rate −5.5 −7.3 −7.0 −6.7 World price index 3.7 8.2 8.5 8.5 Domestic price index 0.1 0.7 0.6 0.0 Real GDP at market prices (% change) −0.3 0.0 −0.1 −0.9 Absorption 0.2 2.7 2.5 1.6 Consumption 0.6 4.3 4.0 2.7 Investment −1.3 −1.9 −1.9 −1.6 Government 0.0 0.0 0.0 0.0 Exports 1.8 −2.5 −2.4 −3.1 Imports 2.6 8.4 8.0 7.3 Source: CGE model results. Table 1. Changes in GDP at market prices and macroeconomic indicators Changes relative to base (%) World food prices (sim1) Food & export crop prices (sim2) Food, export & fuel prices (sim3) Price shocks & domestic supply shock (sim4) Macroeconomic indicators Terms of trade (ToT) 3.2 16.3 15.9 15.9 Real exchange rate −5.5 −7.3 −7.0 −6.7 World price index 3.7 8.2 8.5 8.5 Domestic price index 0.1 0.7 0.6 0.0 Real GDP at market prices (% change) −0.3 0.0 −0.1 −0.9 Absorption 0.2 2.7 2.5 1.6 Consumption 0.6 4.3 4.0 2.7 Investment −1.3 −1.9 −1.9 −1.6 Government 0.0 0.0 0.0 0.0 Exports 1.8 −2.5 −2.4 −3.1 Imports 2.6 8.4 8.0 7.3 Changes relative to base (%) World food prices (sim1) Food & export crop prices (sim2) Food, export & fuel prices (sim3) Price shocks & domestic supply shock (sim4) Macroeconomic indicators Terms of trade (ToT) 3.2 16.3 15.9 15.9 Real exchange rate −5.5 −7.3 −7.0 −6.7 World price index 3.7 8.2 8.5 8.5 Domestic price index 0.1 0.7 0.6 0.0 Real GDP at market prices (% change) −0.3 0.0 −0.1 −0.9 Absorption 0.2 2.7 2.5 1.6 Consumption 0.6 4.3 4.0 2.7 Investment −1.3 −1.9 −1.9 −1.6 Government 0.0 0.0 0.0 0.0 Exports 1.8 −2.5 −2.4 −3.1 Imports 2.6 8.4 8.0 7.3 Source: CGE model results. Table 2. Changes in GDP at factor cost (value added) for selected sectors GDP or factor income shares in the base (%) Changes relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (2007) (sim1) (sim2) (sim3) (sim4) National GDP at factor cost 100.0 −0.4 −0.4 −0.4 −1.2 Agriculture 22.7 −1.4 −1.0 −1.1 −4.8 Cereal crops 2.5 20.8 9.6 9.7 0.1 Roots crops 4.0 −4.0 −3.6 −3.7 −9.3 Matooke 2.4 −3.4 −3.1 −3.1 −7.6 Pulses and oil seeds 2.9 −12.0 −14.7 −14.7 −20.4 Horticulture 0.2 −3.5 −3.2 −3.3 −7.2 Export crops 2.2 −13.8 4.4 4.5 −1.6 Livestock 1.6 0.7 −0.8 −0.8 0.3 Industry 27.2 −0.1 −0.4 −0.4 −0.2 Meat processing 0.1 −0.9 −2.2 −2.2 −0.7 Grain processing 0.7 0.3 0.3 0.3 −0.3 Other food processing 1.5 6.5 3.7 3.8 4.2 Services 50.1 −0.1 0.0 −0.1 −0.2 Factor incomes 100.0 −0.4 −0.4 −0.4 −1.2 Labour 28.2 1.0 0.3 0.2 −2.2 Capital 61.2 −2.0 −3.1 −3.1 −4.7 Land 10.5 5.2 13.7 13.8 21.3 GDP or factor income shares in the base (%) Changes relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (2007) (sim1) (sim2) (sim3) (sim4) National GDP at factor cost 100.0 −0.4 −0.4 −0.4 −1.2 Agriculture 22.7 −1.4 −1.0 −1.1 −4.8 Cereal crops 2.5 20.8 9.6 9.7 0.1 Roots crops 4.0 −4.0 −3.6 −3.7 −9.3 Matooke 2.4 −3.4 −3.1 −3.1 −7.6 Pulses and oil seeds 2.9 −12.0 −14.7 −14.7 −20.4 Horticulture 0.2 −3.5 −3.2 −3.3 −7.2 Export crops 2.2 −13.8 4.4 4.5 −1.6 Livestock 1.6 0.7 −0.8 −0.8 0.3 Industry 27.2 −0.1 −0.4 −0.4 −0.2 Meat processing 0.1 −0.9 −2.2 −2.2 −0.7 Grain processing 0.7 0.3 0.3 0.3 −0.3 Other food processing 1.5 6.5 3.7 3.8 4.2 Services 50.1 −0.1 0.0 −0.1 −0.2 Factor incomes 100.0 −0.4 −0.4 −0.4 −1.2 Labour 28.2 1.0 0.3 0.2 −2.2 Capital 61.2 −2.0 −3.1 −3.1 −4.7 Land 10.5 5.2 13.7 13.8 21.3 Source: CGE model results. Table 2. Changes in GDP at factor cost (value added) for selected sectors GDP or factor income shares in the base (%) Changes relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (2007) (sim1) (sim2) (sim3) (sim4) National GDP at factor cost 100.0 −0.4 −0.4 −0.4 −1.2 Agriculture 22.7 −1.4 −1.0 −1.1 −4.8 Cereal crops 2.5 20.8 9.6 9.7 0.1 Roots crops 4.0 −4.0 −3.6 −3.7 −9.3 Matooke 2.4 −3.4 −3.1 −3.1 −7.6 Pulses and oil seeds 2.9 −12.0 −14.7 −14.7 −20.4 Horticulture 0.2 −3.5 −3.2 −3.3 −7.2 Export crops 2.2 −13.8 4.4 4.5 −1.6 Livestock 1.6 0.7 −0.8 −0.8 0.3 Industry 27.2 −0.1 −0.4 −0.4 −0.2 Meat processing 0.1 −0.9 −2.2 −2.2 −0.7 Grain processing 0.7 0.3 0.3 0.3 −0.3 Other food processing 1.5 6.5 3.7 3.8 4.2 Services 50.1 −0.1 0.0 −0.1 −0.2 Factor incomes 100.0 −0.4 −0.4 −0.4 −1.2 Labour 28.2 1.0 0.3 0.2 −2.2 Capital 61.2 −2.0 −3.1 −3.1 −4.7 Land 10.5 5.2 13.7 13.8 21.3 GDP or factor income shares in the base (%) Changes relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (2007) (sim1) (sim2) (sim3) (sim4) National GDP at factor cost 100.0 −0.4 −0.4 −0.4 −1.2 Agriculture 22.7 −1.4 −1.0 −1.1 −4.8 Cereal crops 2.5 20.8 9.6 9.7 0.1 Roots crops 4.0 −4.0 −3.6 −3.7 −9.3 Matooke 2.4 −3.4 −3.1 −3.1 −7.6 Pulses and oil seeds 2.9 −12.0 −14.7 −14.7 −20.4 Horticulture 0.2 −3.5 −3.2 −3.3 −7.2 Export crops 2.2 −13.8 4.4 4.5 −1.6 Livestock 1.6 0.7 −0.8 −0.8 0.3 Industry 27.2 −0.1 −0.4 −0.4 −0.2 Meat processing 0.1 −0.9 −2.2 −2.2 −0.7 Grain processing 0.7 0.3 0.3 0.3 −0.3 Other food processing 1.5 6.5 3.7 3.8 4.2 Services 50.1 −0.1 0.0 −0.1 −0.2 Factor incomes 100.0 −0.4 −0.4 −0.4 −1.2 Labour 28.2 1.0 0.3 0.2 −2.2 Capital 61.2 −2.0 −3.1 −3.1 −4.7 Land 10.5 5.2 13.7 13.8 21.3 Source: CGE model results. The combined effect of international food price movements in sim1 is a net improvement in Uganda’s terms of trade (3.2 per cent) and an associated appreciation of the real exchange rate (5.5 per cent). Favourable export opportunities for maize and rising profitability in producing import-competing rice and other cereals create incentives for farmers to allocate resources to the cereals sector, resulting in a 20.8 per cent increase in cereals GDP (Table 2). Although the agricultural sector as a whole contracts by 1.4 per cent, increasing returns to cereals benefits the large share of rural farm households that engage in cereals production, which explains the observed rise in consumption expenditure (0.6 per cent) and consequently domestic absorption (0.2 per cent) (Table 1). The latter is indicative of a marginal increase in aggregate welfare in the economy. Of course, while the decline in current investment has no impact on current capital stock levels, it could negatively affect future productive capacity in the economy, and hence also future welfare levels. Table 3 reports changes in consumer prices for selected food commodities. As anticipated, maize prices increase fairly sharply in sim1 (13.2 per cent), which has important spillover effects for prices in the grain processing (8.5 per cent), animal feeds (7.6 per cent), and cattle and sheep (10.6 per cent) sectors due to strong inter-industry linkages. Prices of key staples such as roots, matooke and beans all increase by around 6 per cent as land and labour factors are reallocated towards import-competing and export crop sectors. Table 3. Real changes in consumer prices for selected commodities Changes relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (sim1) (sim2) (sim3) (sim4) Agricultural prices Maize 13.2 14.3 14.1 19.2 Rice 9.3 10.4 10.2 15.4 Other cereals 8.8 9.1 9.0 13.5 Cassava 6.3 10.1 9.9 18.2 Irish potatoes 5.0 8.4 8.2 13.9 Sweet potatoes 6.3 10.4 10.1 18.3 Matooke 6.4 10.2 10.0 18.2 Oil seeds 5.9 9.2 9.0 16.0 Beans 6.8 10.8 10.6 18.4 Vegetables 4.2 7.3 7.1 10.4 Fruit 5.0 8.5 8.4 13.4 Cattle and sheep 10.6 11.2 11.1 8.8 Poultry 0.8 4.6 4.3 2.4 Processed food prices Meat processing 6.2 6.3 6.1 4.1 Grain processing 8.5 8.8 8.6 7.9 Animal feeds 7.6 6.5 6.3 9.1 Other food processing 5.0 4.0 3.8 2.6 Changes relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (sim1) (sim2) (sim3) (sim4) Agricultural prices Maize 13.2 14.3 14.1 19.2 Rice 9.3 10.4 10.2 15.4 Other cereals 8.8 9.1 9.0 13.5 Cassava 6.3 10.1 9.9 18.2 Irish potatoes 5.0 8.4 8.2 13.9 Sweet potatoes 6.3 10.4 10.1 18.3 Matooke 6.4 10.2 10.0 18.2 Oil seeds 5.9 9.2 9.0 16.0 Beans 6.8 10.8 10.6 18.4 Vegetables 4.2 7.3 7.1 10.4 Fruit 5.0 8.5 8.4 13.4 Cattle and sheep 10.6 11.2 11.1 8.8 Poultry 0.8 4.6 4.3 2.4 Processed food prices Meat processing 6.2 6.3 6.1 4.1 Grain processing 8.5 8.8 8.6 7.9 Animal feeds 7.6 6.5 6.3 9.1 Other food processing 5.0 4.0 3.8 2.6 Source: CGE model results. Note: Since the CPI is the numéraire in the model all commodity price changes are considered to be real price relative to a fixed CPI. By extension, this implies that relative food price increases (as shown in the table) would be offset by relative price decreases for non-food commodities (not presented in the table). Table 3. Real changes in consumer prices for selected commodities Changes relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (sim1) (sim2) (sim3) (sim4) Agricultural prices Maize 13.2 14.3 14.1 19.2 Rice 9.3 10.4 10.2 15.4 Other cereals 8.8 9.1 9.0 13.5 Cassava 6.3 10.1 9.9 18.2 Irish potatoes 5.0 8.4 8.2 13.9 Sweet potatoes 6.3 10.4 10.1 18.3 Matooke 6.4 10.2 10.0 18.2 Oil seeds 5.9 9.2 9.0 16.0 Beans 6.8 10.8 10.6 18.4 Vegetables 4.2 7.3 7.1 10.4 Fruit 5.0 8.5 8.4 13.4 Cattle and sheep 10.6 11.2 11.1 8.8 Poultry 0.8 4.6 4.3 2.4 Processed food prices Meat processing 6.2 6.3 6.1 4.1 Grain processing 8.5 8.8 8.6 7.9 Animal feeds 7.6 6.5 6.3 9.1 Other food processing 5.0 4.0 3.8 2.6 Changes relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (sim1) (sim2) (sim3) (sim4) Agricultural prices Maize 13.2 14.3 14.1 19.2 Rice 9.3 10.4 10.2 15.4 Other cereals 8.8 9.1 9.0 13.5 Cassava 6.3 10.1 9.9 18.2 Irish potatoes 5.0 8.4 8.2 13.9 Sweet potatoes 6.3 10.4 10.1 18.3 Matooke 6.4 10.2 10.0 18.2 Oil seeds 5.9 9.2 9.0 16.0 Beans 6.8 10.8 10.6 18.4 Vegetables 4.2 7.3 7.1 10.4 Fruit 5.0 8.5 8.4 13.4 Cattle and sheep 10.6 11.2 11.1 8.8 Poultry 0.8 4.6 4.3 2.4 Processed food prices Meat processing 6.2 6.3 6.1 4.1 Grain processing 8.5 8.8 8.6 7.9 Animal feeds 7.6 6.5 6.3 9.1 Other food processing 5.0 4.0 3.8 2.6 Source: CGE model results. Note: Since the CPI is the numéraire in the model all commodity price changes are considered to be real price relative to a fixed CPI. By extension, this implies that relative food price increases (as shown in the table) would be offset by relative price decreases for non-food commodities (not presented in the table). When food prices rise relative to a fixed numeraire (CPI), non-food prices become relatively cheaper. The welfare implications of the simulation are, therefore, not immediately evident, and depend on how households’ consumption is affected by relative food and non-food price shifts and income changes. We use the equivalent variation (EV) measure to investigate changes in welfare due to price changes. Table 4 shows that all households experience an increase in welfare in sim1, with rural farm and urban households gaining the most (0.6 and 1.0 per cent, respectively). Urban households in this instance are well positioned to benefit from increased trade opportunities particularly in processed food sectors. Non-farm households experience a slight decline in welfare levels as a result of sharp food price increases from which they do not benefit in the same way as farm households. Table 4. Long run changes in household welfare and poverty Changes in welfare (EV) relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (sim1) (sim2) (sim3) (sim4) All households 0.5 4.0 3.7 2.4 Rural farm households 0.6 5.3 5.1 3.9 Rural non-farm households −0.3 2.1 1.6 −0.1 Kampala metro 0.3 2.6 2.3 1.0 Urban farm households 1.0 4.0 3.5 2.2 Urban non-farm households 0.4 3.0 2.6 1.2 Changes in welfare (EV) relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (sim1) (sim2) (sim3) (sim4) All households 0.5 4.0 3.7 2.4 Rural farm households 0.6 5.3 5.1 3.9 Rural non-farm households −0.3 2.1 1.6 −0.1 Kampala metro 0.3 2.6 2.3 1.0 Urban farm households 1.0 4.0 3.5 2.2 Urban non-farm households 0.4 3.0 2.6 1.2 Poverty headcount in the base (%) (2007) Change in poverty relative to base (%-point) (sim1) (sim2) (sim3) (sim4) All households 30.0 −0.7 −6.0 −5.8 −4.5 Rural households 33.4 −0.8 −6.7 −6.4 −5.1 Rural farm 32.3 −0.9 −7.4 −7.2 −5.8 Rural non-farm 41.9 0.0 −1.1 −0.5 0.0 Urban households 11.6 −0.2 −2.5 −2.2 −1.2 Poverty headcount in the base (%) (2007) Change in poverty relative to base (%-point) (sim1) (sim2) (sim3) (sim4) All households 30.0 −0.7 −6.0 −5.8 −4.5 Rural households 33.4 −0.8 −6.7 −6.4 −5.1 Rural farm 32.3 −0.9 −7.4 −7.2 −5.8 Rural non-farm 41.9 0.0 −1.1 −0.5 0.0 Urban households 11.6 −0.2 −2.5 −2.2 −1.2 Source: CGE model and poverty module results. Note: The EV welfare measure is calculated in the CGE model and reported by representative household groups in the model; the poverty module calculates changes in poverty on the basis of the survey and therefore a different reporting structure can be adopted. Table 4. Long run changes in household welfare and poverty Changes in welfare (EV) relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (sim1) (sim2) (sim3) (sim4) All households 0.5 4.0 3.7 2.4 Rural farm households 0.6 5.3 5.1 3.9 Rural non-farm households −0.3 2.1 1.6 −0.1 Kampala metro 0.3 2.6 2.3 1.0 Urban farm households 1.0 4.0 3.5 2.2 Urban non-farm households 0.4 3.0 2.6 1.2 Changes in welfare (EV) relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (sim1) (sim2) (sim3) (sim4) All households 0.5 4.0 3.7 2.4 Rural farm households 0.6 5.3 5.1 3.9 Rural non-farm households −0.3 2.1 1.6 −0.1 Kampala metro 0.3 2.6 2.3 1.0 Urban farm households 1.0 4.0 3.5 2.2 Urban non-farm households 0.4 3.0 2.6 1.2 Poverty headcount in the base (%) (2007) Change in poverty relative to base (%-point) (sim1) (sim2) (sim3) (sim4) All households 30.0 −0.7 −6.0 −5.8 −4.5 Rural households 33.4 −0.8 −6.7 −6.4 −5.1 Rural farm 32.3 −0.9 −7.4 −7.2 −5.8 Rural non-farm 41.9 0.0 −1.1 −0.5 0.0 Urban households 11.6 −0.2 −2.5 −2.2 −1.2 Poverty headcount in the base (%) (2007) Change in poverty relative to base (%-point) (sim1) (sim2) (sim3) (sim4) All households 30.0 −0.7 −6.0 −5.8 −4.5 Rural households 33.4 −0.8 −6.7 −6.4 −5.1 Rural farm 32.3 −0.9 −7.4 −7.2 −5.8 Rural non-farm 41.9 0.0 −1.1 −0.5 0.0 Urban households 11.6 −0.2 −2.5 −2.2 −1.2 Source: CGE model and poverty module results. Note: The EV welfare measure is calculated in the CGE model and reported by representative household groups in the model; the poverty module calculates changes in poverty on the basis of the survey and therefore a different reporting structure can be adopted. Since poor households allocate a relatively large share of their budget to food items, the expectation is that rising food prices will negatively affect poverty, ceteris paribus. Yet, as shown in Table 4, poverty declines by 0.7 percentage points nationally in sim1. Rural farm households experience the greatest decline in poverty (0.9 percentage points), which suggests that the poor are not necessarily excluded from the welfare gains achieved by rural farm households as a whole. Even poor rural non-farm and urban households are seemingly not negatively affected by significant food price increases. However, closer inspection reveals that although households across the board are able to increase their real expenditure, the actual quantity of food consumed declines as households re-orient their budgets towards cheaper non-food items. For at least those households that are close to or below the food requirement, this consumption shift may have implications for food and nutrition security outcomes. The second simulation, sim2, introduces international export crop price shocks in addition to the food price shocks modelled in sim1. As this implies a further improvement in the terms of trade (16.3 per cent relative to the base) the exchange rate appreciation is also greater than before (7.3 per cent) (Table 1). In addition to cereals, traditional export crops now also become an attractive investment opportunity due to higher export prices (Table 2). As a result, GDP in the cereals sector does not expand by as much as before (9.6 per cent), which is due to two reasons: first, the sector now competes more directly with traditional export crops for limited resources; and second, the larger exchange rate appreciation lowers the competitiveness of Ugandan exports compared to sim1. Agricultural sector GDP still contracts, but not by as much as before. Once again, this simulation is associated with an increase in aggregate welfare, as evidenced by the 2.7 per cent rise in domestic absorption, which is driven mainly by the increase in consumption expenditure (4.3 per cent), which more than offsets the decline in current investment. However, somewhat surprisingly, the real value of exports declines by 2.5 per cent despite the seemingly improved opportunities to export and compete with imports. This reflects the eroding effect of the exchange rate appreciation on export earnings. At the same time, real imports rise substantially (8.4 per cent), which is partly facilitated by the exchange rate appreciation, but also relates to the shift in household consumption towards cheaper non-food items which have a greater import intensity. These trade balance effects are amplified in sim2 compared to sim1 given the larger average increases in food price now observed (Table 3). We can, therefore, conclude that rising household welfare levels, combined with relative price changes, ultimately increases the import intensity of the economy and weakens the trade balance, mainly because the export opportunities associated with relative world price changes are restricted to the food and agricultural sectors. Overall household welfare increases by 4 per cent in sim2. The measured decline in poverty is also larger than in sim1, with the national poverty rate declining by 6 per cent (Table 4). The latter is driven mostly by the 7.4 percentage point decline in the rural farm household poverty rate. As anticipated, the addition of fuel price increases in sim3 only marginally changes the results seen in sim2, owing to the fact that the worst effects of the 2008 fuel price spike had dissipated by 2011 and that fuel only accounts for 8.3 per cent of Uganda’s total imports. Nevertheless, rising fuel prices cause a slight deterioration in the terms of trade, such that the net effect in sim3 is a 15.9 per cent improvement in the terms of trade, while the exchange rate appreciation is 7 per cent. We note a slightly lower aggregate welfare effects (2.5 per cent increase in domestic absorption) as well as a slightly smaller increase in real imports than before (8 per cent) (Table 1). Sectoral GDP results are largely similar (Table 2) than in the previous simulation, while food prices increase by slightly less. Lastly, welfare and poverty effects are only slightly weaker given the inflationary effect of fuel prices for particularly trade and transport related services (Table 3). The final simulation, sim4, combines the international price shocks of sim3 with a 7.6 per cent decline in the area of land supply, which is designed to simulate the effect of a decline in cultivated land in per capita terms. The terms of trade effect in sim4 are similar to that in sim3 (i.e. 15.9 per cent improvement), but land constraints ultimately lead to a much-weakened aggregate welfare effect (domestic absorption rises only 1.6 per cent) and a larger decline in agricultural exports (3.1 per cent) (Table 1). Agricultural GDP contracts by 4.8 per cent, while gains observed in earlier simulations in some subsectors, such as cereals or export crops, are wiped out (Table 2). Particularly, sharp GDP losses are also recorded in many of the large non-tradable sectors, such as pulses and oil seeds, root crops and matooke. This simulation is associated with large agricultural crop price increases, in some cases almost double what was measured in earlier simulations (Table 3). Agricultural crop price increases ensure that, despite the contraction in agriculture, returns to factors of production, especially land and family farm labour, increase sharply. The implication is that producers – and by extension households linked to producers via the factor market – ultimately experience welfare gains, albeit not to the same extent as in earlier simulations. More specifically, with the exception of rural non-farm households, all Ugandan households experience an increase in welfare (EV), with rural farm households gaining most (3.9 per cent). Poverty rates also decline across the board, especially for rural farm households (by 5.8 percentage points), and to lesser extent urban households (1.2 percentage points decline). The national poverty rate declines by 4.5 percentage points. While a detailed discussion of results obtained under alternative model closures falls beyond the scope of the current study, this is certainly an area that warrants further investigation, especially as far as the savings-investment closure is concerned. For example, under an investment-driven closure where household saving rates adjust generate a target level of investment, overall welfare (EV) levels increase slightly more than under the savings-driven closure presented earlier. However, we no longer observe the large welfare gains for farm households, i.e. welfare gains now largely accrue to urban households. The decline in poverty is slightly smaller for all household subgroups under this closure, e.g. national poverty declines by 4.1 percentage points. In summary, the results for the medium- to longer run analysis appear to be driven largely by international price shocks (sim1) and the domestic agricultural supply shock (sim3). Ultimately, however, both the macroeconomic growth changes and household-level welfare outcomes are dominated by the international price shocks, which result in positive welfare and poverty effects for most household groups but not for rural non-farm households. In general, however, the significant improvement in the terms of trade represents an opportunity for Ugandan farmers to access export markets, with positive growth and welfare spillover effects. 4. Conclusions This study analyses the impact of price changes, roughly between 2008 and 2011, on Uganda’s economy. We consider the combined impact of changes in prices of a basket of commodities on welfare and poverty. First-order results generated with an extended version of Deaton equation for measuring the short-run welfare impact of price changes are compared with longer run results generated with the aid of a full-scale CGE model that accounts for economy-wide direct and indirect effects. In the first-order analysis, we find that commodity price movements had only small welfare implications, generally changing poverty by less than 1 percentage point. However, these averages hide substantial heterogeneity, with urban households, especially those living just above the poverty line, likely to lose most as a result of higher prices. We also find that the relative price movements are important. We find that price patterns towards that end of our time series are damaging, when the price of cassava, often bought by poor households, shoots up, while the price of maize, often sold by poor farm households, remains flat. Our long-run analysis, which explicitly accounts for behavioural responses to price changes, albeit at an aggregated household group level, predicts that recent international food price changes would lead to an improvement in Uganda’s terms of trade given the country’s position as a net exporter of food and agricultural products. Domestic price shocks erode some of the gains from the terms of trade improvement associated with international price changes, but the net effect is still an important improvement in welfare and reduction in poverty, with the exception perhaps of rural non-farm households. Overall, in the long run, poverty declines by about 4.5 percentage points in our combined scenario, ceteris paribus. These diverging results should not be interpreted as evidence of superiority of either method over the other. Even though we present them side by side, it remains difficult to compare the results given the differences in the underlying assumptions. The first-order approximation is essentially an assessment a theoretical concept to assess immediate impact worst-case scenarios.8 The frugal nature of the model is its strength: its simplicity and low data requirements make it very popular among policy makers and practitioners. By using household survey data, it allows estimation of the welfare impact for any definable subgroup of households in the population. In contrast, it is hard to simulate historical changes in domestic prices with a CGE model because the prices of non-tradable crops in such a model are endogenous. The welfare impact depends partly on whether the price change is due to rising international prices, a domestic supply shock, or other exogenous factors. The fact that CGE models also work with representative economic agents puts limits on how far one can go in exploring heterogeneous responses to price changes. CGE models are also complex, requiring expert knowledge and custom data, such as a SAM and estimates of various commodity- and household-specific elasticities. At the same time, they do provide a much more comprehensive and complete picture of the impact of price changes and are flexible in what aspects of the economy should be included in the model. For example, it allows policy makers to introduce particular features of the economy, such as market imperfections. Our results may serve to add an additional layer of nuance to the debate on the costs and benefits of higher food prices (Swinnen and Squicciarini, 2012). Perhaps more importantly, the contrasting first-order and general-equilibrium outcomes should be taken as a warning for policy makers not to rely too much on a single approach. Most likely, policy makers in developing countries will be biased towards first-order analyses because of its ease of interpretation and limited data requirements mentioned above. The likely outcome that especially the urban consumers are hurt may attract more policy action because they are more vocal and visible than rural smallholder farmers. This may mean policy makers will try to insulate (urban) consumers from price increases through price support and trade policy such as export restrictions and import tariff reductions. Focusing too much on the negative short-run effects may thus jeopardise long-run benefits if policy makers attempt to dampen price signals that are essential in driving behavioural change among rural producers. At the same time, relying too much on general-equilibrium models may result in policy makers neglecting the plight of poor households with limited options to substitute or adapt. Confronted with higher food prices, even in the short run, these households may be forced to liquidate productive assets, potentially pushing them into a poverty trap (Carter and Lybbert, 2012). Therefore, policy makers may need to provide social safety nets to safeguard future productive capacity. Both analyses are thus relevant to our understanding of appropriate policy design: first-order price impact models can inform household-level emergency responses to price shocks. They may also be better at capturing heterogeneity of impacts due the fact that the unit of analysis are actual households as opposed to representative households in a CGE analysis. In the context of rapidly urbanising societies, such models may be especially useful in prompting governments to design well-targeted social safety nets such as food for work or cash or in-kind transfers. General-equilibrium analyses can help policy makers identify longer term interventions needed to support the adjustment to new economic realities. This will allow governments to create an enabling environment in which farm households can seize the opportunity created by higher prices for their products. Footnotes 1 It derives its name because it is the first-order Taylor-series expansion of compensating variation (if based on the original quantities) or EV (if based on the final quantities after adjustment to price changes) associated with a price change. 2 Probably closest to our work, a recent paper by Tiberti and Tiberti (2018) compare first-order effects to a partial equilibrium model that allows for consumption, production and labour/leisure adjustments. 3 If the percentage change in consumer prices is equal to the percentage change in the producer prices (such as when there is a proportional margin between consumer and producer prices) and if there is just one commodity, this expression collapses to equation (1), proposed by Deaton (1989). However, substantial evidence of asymmetric price transmission in agricultural commodity markets, where price reductions are passed on to producers but price increases are not, means many studies that assess first-order impact of price changes now value changes in consumer and producer prices separately (Meyer and Von Cramon-Taubadel, 2004). Dawe and Maltsoglou (2014), Minot and Dewina (2015) and Van Campenhout, Lecoutere and D’Exelle (2015) show that accounting for marketing margins, so that producer price changes differ from consumer price changes, can have a noticeable effect on the welfare impact. 4 Note that wholesale prices may be substantially higher than the price farmers receive due to transaction costs in transporting goods from the point of sale to the wholesale market. Similarly, the cost of what a household consumes from the market is likely to be higher than the retail price, as the household needs to get the crop from the market. However, we are not interested in levels, but in proportional changes in consumer prices over time (Δpj/pj) and in changes in producer prices over time (Δwi/wi). If the wholesale price in the market is the farm-gate producer price plus a proportional transaction cost to bring the crop to the market, then the proportional change in the wholesale price will be equal to the proportional change in the producer price. Similarly, if the margin between the wholesale price and the consumer price is proportional, then they will change by the same proportion. However, it is likely that some portion of the marketing margins is fixed rather than proportional. 5 We decided to take the average price level in 2008 as the baseline price level. We use simple linear interpolation to accommodate gaps in the price series data. 6 Looking back at Figure 1, we see that for the available price data, price levels are at their highest in December 2011. As mentioned before, the choice of points in time is often driven by data availability or represents extreme situations (e.g. the lowest versus the highest prices over a certain period). In this case, a government official may be tempted to take the prices situation in December 2011 as the endpoint, which would lead to the conclusion that welfare declined 2.5 per cent (and a large increase in poverty, see below). However, such a conclusion would not be consistent with the fact that, during much of the period, the first-order impact of higher prices actually seemed to increase welfare. 7 Analysis of the impact of individual commodities on welfare reveals that maize is largely responsible for this positive effect. 8 First-order estimates do not necessarily result in worst-case results, but they do generate an over-estimate of the welfare impact (positive or negative) of a price change in competitive markets. Through behavioural changes, households either mitigate negative effects or exploit positive effects. Furthermore, effects may also depend on the market distortions/failures that exit in the considered economy. Price shocks can lead to more detrimental effects if, for instance, monopolistic importers increase their margins on import flows. However, the markets we consider in this analysis are reasonably competitive. References Arndt , C. , Benfica , R. , Maximiano , N. , Nucifora , A. and Thurlow , J. ( 2008 ). Higher fuel and food prices: impacts and responses for Mozambique . Agricultural Economics 39 ( s1 ): 497 – 511 . Google Scholar Crossref Search ADS Bellemare , M. ( 2015 ). Rising food prices, food price volatility, and social unrest . American Journal of Agricultural Economics 97 ( 1 ): 1 – 21 . Google Scholar Crossref Search ADS Benson , T. , Minot , N. , Pender , P. , Robles , M. and von Braun , J. ( 2013 ). Information to guide policy responses to higher global food prices: the data and analyses required . Food Policy 38 : 47 – 58 . Google Scholar Crossref Search ADS Benson , T. , Mugarura , S. and Wanda , K. ( 2008 ). Impacts in Uganda of rising global food prices: the role of diversified staples and limited price transmission . Agricultural Economics 39 ( s1 ): 513 – 524 . Google Scholar Crossref Search ADS Boysen , O. and Matthews , A. ( 2012 ). The differentiated effects of food price spikes on poverty in Uganda. Paper prepared for the 123rd European Association of Agricultural Economists. Carter , M. R. and Lybbert , T. J. ( 2012 ). Consumption versus asset smoothing: testing the implications of poverty trap theory in Burkina Faso . Journal of Development Economics 99 ( 2 ): 255 – 264 . Google Scholar Crossref Search ADS Dawe , D. and Maltsoglou , I. ( 2014 ). Marketing margins and the welfare analysis of food price shocks . Food Policy 46 : 50 – 55 . Google Scholar Crossref Search ADS Deaton , A. ( 1989 ). Rice prices and income distribution in Thailand: a non-parametric analysis . The Economic Journal 99 ( 395 ): 1 – 37 . Google Scholar Crossref Search ADS FAO (Food and Agriculture Organization of the United Nations) . 2016 . FAOSTAT. Rome. http://faostat.fao.org/ Ferreira , F. H. G. , Fruttero , A. , Leite , P. and Lucchetti , L. ( 2013 ). Rising food prices and household welfare: evidence from Brazil in 2008 . Journal of Agricultural Economics 64 ( 1 ): 151 – 176 . Google Scholar Crossref Search ADS Headey , D. ( 2016 ). Food prices and poverty. World Bank Economic Review doi.10.1093/wber/lhw064 . Headey , D. , Malaiyandi , S. and Fan , S. ( 2010 ). Navigating the perfect storm: reflections on food, energy and financial crises . Agricultural Economics 41 ( s1 ): 217 – 228 . Google Scholar Crossref Search ADS Headey , D. and Martin , W. ( 2016 ). The impact of food prices on poverty and food security . Annual Review of Resource Economics 8 : 329 – 351 . Google Scholar Crossref Search ADS Ivanic , M. , Martin , W. and Zaman , H. ( 2012 ). Estimating the short-run poverty impacts of the 2010–11 surge in food prices . World Development 40 ( 11 ): 2302 – 2317 . Google Scholar Crossref Search ADS Jacoby , H. G. ( 2016 ). Food prices, wages, and welfare in rural India . Economic Inquiry 54 ( 1 ): 159 – 176 . Google Scholar Crossref Search ADS Jayne , T. S. , Chamberlin , J. and Headey , D. ( 2014 ). Land pressures, the evolution of farming systems, and development strategies in Africa: a synthesis . Food Policy 48 : 1 – 17 . Google Scholar Crossref Search ADS Löfgren , H. , Harris , R. and Robinson , S. ( 2002 ). A standard Computable General Equilibrium (CGE) model in GAMS. IFPRI Trade and Macroeconomics Discussion Paper 75. Washington, DC: International Food Policy Research Institute. Matovu , J. and Twimukye , E. ( 2009 ). Increasing World Prices: Blessing or Curse? EPRC Research Series, No. 61 . Kampala, Uganda : Economic Policy Research Centre . Meyer , J. and Von Cramon-Taubadel , S. ( 2004 ). Asymmetric price transmission: a survey . Journal of Agricultural Economics 55 ( 3 ): 581 – 611 . Google Scholar Crossref Search ADS Minot , N. and Dewina , R. ( 2015 ). Are we overestimating the negative impact of higher food prices? Evidence from Ghana . Agricultural Economics 46 ( 4 ): 579 – 593 . Google Scholar Crossref Search ADS Minot , N. and Goletti , F. ( 1998 ). Export liberalization and household welfare: the case of rice in Viet Nam . American Journal of Agricultural Economics 80 ( 4 ): 738 – 749 . Google Scholar Crossref Search ADS Robinson , S. , Yúnez-Naude , A. , Hinojosa-Ojeda , R. , Lewis , J. D. and Devarajan , S. ( 1999 ). From stylized to applied models: building multisector CGE models for policy analysis . The North American Journal of Economics and Finance 10 ( 1 ): 5 – 38 . Google Scholar Crossref Search ADS Simler , K. ( 2010 ). The Short-Term Impact of Higher prices on Uganda. Policy Research Series. No. 5210 . Washington DC : World Bank . Stephens , E. C. and Barrett , C. B. ( 2010 ). Incomplete credit markets and commodity marketing behaviour . Journal of Agricultural Economics 62 ( 1 ): 1 – 24 . Google Scholar Crossref Search ADS Swinnen , J. and Squicciarini , P. ( 2012 ). Mixed messages on prices and food security . Science 27-335 ( 6067 ): 405 – 406 . Google Scholar Crossref Search ADS Thurlow , J. ( 2012 ). A 2007 Social Accounting Matrix for Uganda . Washington, DC : International Food Policy Research Institute . Tiberti , L. and Tiberti , M. ( 2018 ). Food price changes and household welfare: what do we learn from two different approaches? Journal of Development Studies 54 ( 1 ): 72 – 92 . Google Scholar Crossref Search ADS UBOS (Uganda Bureau of Statistics) ( 2012 ). Statistical Abstract 2012 . Kampala, Uganda . http://www.ubos.org/onlinefiles/uploads/ubos/pdf%20documents/2012StatisticalAbstract.pdf. Van Campenhout , B. , Lecoutere , E. and D’Exelle , B. ( 2015 ). Inter-temporal and spatial price dispersion patterns and the well-being of maize producers in southern Tanzania . Journal of African Economies 24 ( 2 ): 230 – 253 . Google Scholar Crossref Search ADS Weber , M. T. , Staatz , J. M. , Holtzman , J. S. , Crawford , E. W. and Bernsten , R. H. ( 1988 ). Informing food security decisions in Africa: empirical analysis and policy dialogue . American Journal of Agricultural Economics 70 ( 5 ): 1044 – 1052 . Google Scholar Crossref Search ADS Wiebelt , M. , Pauw , K. , Matovu , J. M. , Twimukye , E. and Benson , T. ( 2011 ). Managing future oil revenues in Uganda for agricultural development and poverty reduction: A CGE analysis of challenges and options. Kiel Institute Working Papers No. 1696 (May 2011), University of Kiel, Germany. Author notes Review coordinated by Jack Peerlings © Oxford University Press and Foundation for the European Review of Agricultural Economics 2018; all rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png European Review of Agricultural Economics Oxford University Press http://www.deepdyve.com/lp/oxford-university-press/the-impact-of-food-price-shocks-in-uganda-first-order-effects-versus-HhXq9HcGoU

Loading next page...

References (23)

Hanan Jacoby (2013)
Food Prices, Wages, and Welfare in Rural India
Macroeconomics: Prices
J. Swinnen, P. Squicciarini (2012)
Mixed Messages on Prices and Food Security
Science, 335
D. Dawe, I. Maltsoglou (2014)
Marketing margins and the welfare analysis of food price shocks
Food Policy, 46
T. Benson, Samuel Mugarura, K. Wanda (2008)
Impacts in Uganda of rising global food prices: the role of diversified staples and limited price transmission
Agricultural Economics, 39
D. Headey, W. Martin (2016)
The Impact of Food Prices on Poverty and Food Security
Annual Review of Resource Economics, 8
N. Minot, Reno Dewina (2015)
Are we overestimating the negative impact of higher food prices? Evidence from Ghana
Agricultural Economics, 46
S. Robinson, A. Yúnez-Naude, Raúl Hinojosa-Ojeda, Jeffrey Lewis, S. Devarajan (1999)
From stylized to applied models:: Building multisector CGE models for policy analysis
The North American Journal of Economics and Finance, 10
L. Tiberti, M. Tiberti (2018)
Food Price Changes and Household Welfare: What Do We Learn from Two Different Approaches?
The Journal of Development Studies, 54
Marc Bellemare (2015)
Rising Food Prices, Food Price Volatility, and Social Unrest
Wiley-Blackwell: American Journal of Agricultural Economics
D. Headey, S. Malaiyandi, S. Fan (2010)
Navigating the perfect storm: reflections on the food, energy, and financial crises
Agricultural Economics, 41
T. Jayne, J. Chamberlin, D. Headey (2014)
Land pressures, the evolution of farming systems, and development strategies in Africa: A synthesis
Food Policy, 48
T. Benson, N. Minot, J. Pender, Miguel Robles, J. Braun (2013)
Information to guide policy responses to higher global food prices: The data and analyses required
Food Policy, 38
Francisco Ferreira, Anna Fruttero, P. Leite, L. Lucchetti (2011)
Rising Food Prices and Household Welfare: Evidence from Brazil in 2008
Agricultural & Natural Resource Economics eJournal
Angus Deaton (1989)
Rice Prices and Income Distribution in Thailand: A Non-parametric Analysis
The Economic Journal, 99
C. Arndt, Rui Benfica, Nelson Maximiano, A. Nucifora, J. Thurlow (2008)
Higher fuel and food prices: impacts and responses for Mozambique
Agricultural Economics, 39
N. Minot, F. Goletti (1998)
Export Liberalization and Household Welfare: The Case of Rice in Vietnam
ERN: Other Econometrics: Applied Econometric Modeling in Microeconomics - Microeconometric Models of Household Behavior (Topic)
Emma Stephens, C. Barrett (2011)
Incomplete Credit Markets and Commodity Marketing Behaviour
Journal of Agricultural Economics, 62
Michael Weber, J. Staatz, E. Crawford, R. Bernsten, J. Holtzman (1988)
Informing Food Security Decisions in Africa: Empirical Analysis and Policy Dialogue
American Journal of Agricultural Economics, 70
M. Carter, Travis Lybbert (2012)
Consumption versus asset smoothing: testing the implications of poverty trap theory in Burkina Faso
Journal of Development Economics, 99
Meir Salem, Y. Siddiqi (2012)
A Social Accounting Matrix for Canada
B. Campenhout, Els Lecoutere, Ben d’Exelle (2015)
Inter-Temporal and Spatial Price Dispersion Patterns and the Well-Being of Maize Producers in Southern Tanzania
Journal of African Economies, 24
M. Ivanic, W. Martin, H. Zaman (2011)
Estimating the Short-Run Poverty Impacts of the 2010-11 Surge in Food Prices
Agricultural & Natural Resource Economics eJournal
J. Meyer, S. Cramon-Taubadel (2004)
Asymmetric Price Transmission: A Survey
Journal of Agricultural Economics, 55

Publisher: Oxford University Press
Copyright: © Oxford University Press and Foundation for the European Review of Agricultural Economics 2018; all rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN: 0165-1587
eISSN: 1464-3618
DOI: 10.1093/erae/jby013
Publisher site: See Article on Publisher Site

Abstract

Abstract For developing countries, whose governments are faced with volatile world food prices, the appropriate policy response hinges on who are the likely winners and losers. Therefore, it is necessary to predict the impact of higher commodity prices on different subgroups of society. We compare the results of a method that is popular with policy makers because of its parsimony and ease of interpretation with the results of a more complex and data-intensive general-equilibrium model. Using historical prices between 2008 and 2011 for Uganda, we find that both methods predict high prices benefit poor rural farmers, but more so if a more elaborate model is used. 1. Introduction The world food price crisis of 2007–2008 has been an eye opener for many governments in developing countries. In response to the ensuing social unrest and political instability, governments have been scrambling to manage the local impact of global food price surges. However, higher food prices may also signal new economic opportunities. To guide the choice of the appropriate policy instruments, governments must determine who are the likely winners and losers of food price surges. This is essentially an empirical question that depends on production and consumption patterns among households as well as the pass-through of global commodity prices (Benson, Mugarura and Wanda, 2008). Estimates of who wins and who loses also depend on the complexity of the models used, which in part is influenced by the availability of data and analytical capacity. At one extreme, the impact of a price increase of a commodity at the household level can be estimated using a first-order approximation of the welfare impact.1 The first-order welfare impact of a price change is the impact without taking into account any response by economic agents to price changes. As such, it can be considered the ‘short-term’ or ‘immediate’ welfare impact. Modest data requirements and a simple formula with a straightforward interpretation make this the workhorse model among many government agencies within developing countries (Headey, 2016). At the other extreme, the impact of a price increase of a commodity is assessed through full-scale computable general-equilibrium (CGE) models. These models trace the impact of an exogenous price shock through the entire economy, incorporating adjustments in consumption and production decisions of households, secondary effects through the labour market, capital and land market consequences, and so on. These models are much more data intensive and require specialised software and advanced modelling skills (Benson et al., 2013). The empirical literature suggests that studies of food price hikes that rely on more elaborate models arrive at more positive (or less negative) welfare conclusions than studies that only consider first-order effects (Headey and Martin, 2016). A recent multi-country assessment of first-order welfare impacts of food price hikes during the second half of 2010 by Ivanic, Martin and Zaman (2012) finds poverty increased by an average 1.1 per cent in low-income countries. For Uganda, Benson, Mugarura and Wanda (2008) evaluate the immediate welfare impact of the 2008 food price crisis and conclude that the impact was likely to be negative, but small. A similar study by Simler (2010) finds the food price crisis increased poverty in Uganda by 2.6 percentage points. A recent example of a study that goes beyond first-order impacts is Jacoby (2016), who finds that in India, rural households across the income spectrum benefit from higher agricultural prices, mainly due to wage effects. For Uganda, Matovu and Twimukye (2009) use a CGE model to look at the effects of rising cereals prices. Their simulated price increases lead to large improvements in well-being. Boysen and Matthews (2012) run an integrated CGE–micro simulation model to analyse the 2006–2008 increase in commodity prices in Uganda. Their results suggest increased commodity prices reduced national poverty by about 2.7 percentage points. Inspired by Robinson et al. (1999), in this paper, we juxtapose the kind of stylised models prevalent in policy environments in developing countries with an ideal situation where economy-wide effects are incorporated in the analysis. In particular, we present the results of both a first-order analysis and a long-run general-equilibrium analysis based on the same case study: historical price movements of a set of key commodities in Uganda over the period 2008–2011. This article contributes to the study the effect of food price changes on welfare in several ways. First, to our knowledge, this is the first study that runs both a first-order effects analysis and a complete CGE analysis on the same case.2 While previous studies suggest substantially more positive welfare effects from price increases when long-run general-equilibrium models are used than when a first-order approach is taken, it is still unclear how much of this difference should be attributed to the difference in method and how much to the fact that different case studies are used. In addition, instead of looking at a hypothetical price increase of one particular commodity, we try as much as possible to base both analyses on actual, historical price movements of a range of commodities. As such, our simulations also serve as an impact evaluation of price movements, but controlling for other external shocks that may have affected socio-economic outcomes during the same period. This study also adds to the available evidence of the welfare impact of food prices in Uganda, where this is a politically contentious issue (Bellemare, 2015). Finally, by comparing the pragmatic first-order approach to a comprehensive, economy-wide analysis, we also aim to provide an additional explanation for the often opposing views held on the costs and benefits of higher commodity prices (Swinnen and Squicciarini, 2012). The remainder of this article is organised as follows. The next section presents the first-order analysis, including sub-sections on the methodology, data and results. Section 3 presents the long-run general-equilibrium analysis, including descriptions of the social accounting matrix (SAM), behavioural assumptions underlying the CGE model, scenarios we simulate, closure rules and results. The fourth section concludes. 2. First-order impact of food prices on household welfare 2.1. A welfare compensation index Farm households in developing countries characterised by semi-substance agriculture typically participate in the market as both buyer and seller for a range of commodities. Household surveys show that, in many developing countries, many rural households (often a majority) are net buyers of major crops, even staple food crops (Weber et al., 1988). In this context, Deaton (1989) proposed a simple first-order approximation of the welfare impact of price changes in a study of rice farmers in Thailand, who both produce for the market and consume from the market, using data from a standard household survey: Δyy=(rpy−qpy)Δpp (1) where y is a measure of household welfare, p is the price of the commodity, r is the quantity of sales and q is the quantity of purchases. The expression in parentheses is called the net benefit ratio (NBR) of a commodity and can be considered the short-term elasticity of welfare with respect to the commodity price. If the household is a net seller of the commodity, the NBR is positive, and price increases are associated with higher welfare. If the household is a net buyer, the NBR is negative and higher prices mean lower welfare. Deaton’s expression has been extended to include second-order terms (accounting for household response to price changes), to distinguish between changes in producer and consumer prices and to measure the welfare effect of the price changes in multiple commodities (Minot and Goletti, 1998; Boysen and Matthews, 2012; Dawe and Maltsoglou, 2014; Tiberti and Tiberti, 2016). In this study, we extend Deaton’s expression to multiple commodities and allow differences in the proportional changes in producer and consumer prices, as shown in the following equation3: Δyy=∑iriwiyΔwiwi−∑jqjpjyΔpjpj (2) where ri is the quantity of commodity i sold, wi is its sale price and pj is now the price of purchased good j. The first term on the right side sums the share of income from each commodity (riwi/y) multiplied by the corresponding proportional change in producer prices (⁠ Δwi/wi ⁠). The second term on the right side sums the share of the budget spent on each commodity consumed (qjpj/y) multiplied by the corresponding proportional change in consumer price (⁠ Δpj/pj ⁠). Equation (2) is calculated for each household and then aggregated to some level. For instance, one can simply take the average to get the overall first-order proportional impact on welfare. However, it is often more instructive to compare the average impact of different groups of households that have different consumption and production patterns, such as urban and rural households or farm and non-farm households (Benson et al., 2013). Most studies using Deaton’s method adopt a before-and-after approach, where price changes are calculated as the difference between price levels of all commodities sold and bought at two distinct points in time (⁠ Δwi=wi,t=1-wi,t=0 and Δpj=pj,t=1-pj,t=0 ⁠, where wi,t=0 and pj,t=0 are, respectively, the producer and consumer prices before the price change and wi,t=1 and pj,t=1 are the producer and consumer prices after the price change). These studies are very sensitive to the choice of these points in time. Often, the choice of points is arbitrary, is determined by data availability, or represents extreme situations (e.g. the lowest versus the highest prices over a certain period). In general, however, price data are much more readily available than data on quantities purchased and sold, which are typically derived from household surveys. The availability of time series data on prices allows one to define price changes over a range of periods, fixing baseline price levels at one point in time and calculating the difference with various end points. Evaluating equation (2) at different end points in time allows one to create a ‘welfare compensation index’. As noted above, a standard Deaton first-order welfare analysis, as represented by equation (1) above, quantities sold in the market (r) and quantities purchased through the market (q) are assumed to be fixed and measured before the price change happens. This measure does not take into account changes in household behaviour as a result of the price change, even though it is very likely that households will buy less if a commodity becomes more expensive and sell more if the price increases. It is an approximation of reality that is most appropriate in the very short run, for small price changes, or for commodities with inelastic supply and demand. Furthermore, it neglects changes in factor markets and other indirect effects of the price change (Ferreira et al., 2013). As one may argue that adding second-order effects to the short-run analysis would make our application more realistic, we want to reiterate that we want to compare results obtained from a simple method that is very popular among policy makers because of its ease of use and light data requirements to a more elaborate empirical assessment of the impact of price changes. While second-order estimates of the welfare impact are relatively easy to calculate, they do rely on estimates of the elasticity of supply and demand for the relevant commodities. In addition, particularly when working with staple crops for which supply and demand are usually assumed to be inelastic (around −0.3 and 0.3), our experience suggests that adding second-order effects does not make much difference (e.g. Minot and Dewina, 2015). The first-order method also abstracts from several other real-life features. For instance, it is assumed that all households face the same changes in consumer and producer prices. However, as already mentioned in footnote 3, the existence of marketing margins drives a wedge between consumer and producer prices. In addition, the dates for prices at which commodities are bought and sold are assumed to be the same, while in reality, farmers often sell at one point in time and buy at another (Stephens and Barrett, 2010). As such, for groups of farmers that are particularly affected by intertemporal and spatial price dispersion, such as farmers living in remote areas that sell most of the harvest immediately after the harvest and are forced to turn to the market in the lean season, the metric in equations (1) and (2) may provide a poor description of what actually happens to their welfare (Dawe and Maltsoglou, 2014; Minot and Dewina, 2015; Van Campenhout, Lecoutere and D’Exelle, 2015). The extent to which spatial and intertemportal price dispersion can be accounted for depends on data availability. In this study, we have access to prices collected in different locations, allowing for some degree of spatial price dispersion. 2.2. Data We use two data sources for our application. For producer prices (wi) and consumer prices (pj), we use data from FIT-Uganda, a business development consultancy that runs the Infotrade market information service. FIT-Uganda collects prices for 44 commodities from 23 towns and cities throughout Uganda three times a week. For most commodities, both wholesale (which we use to proxy producer prices) and retail prices (which we use as consumer prices) are available.4 We include 10 of the most important agricultural commodities in Uganda, and aggregate the series to a monthly frequency. For the quantities produced (ri) and consumed (qj), we use the 2009/10 wave of the Uganda National Panel Survey, which has a detailed agricultural module. Given the importance of prices in a first-order analysis, it may be useful to start by describing the price data we will use. The data have three dimensions: time, location and commodity. Figure 1 graphs the time dimension (2008–2011), averaging over space and commodities. Although the average price level during the analysis period is 1,076 Ugandan shillings (UGX) for wholesale prices and UGX 1,256 for retail prices, such an average was clearly not representative anymore in 2011. World prices for staple foods were on the rise between 2006 and 2008 and accelerated sharply in the beginning of 2008. While food prices dropped significantly by the end of 2008, they rose again in 2009 and remain high by historical standards in both international and local markets (Headey, Malaiyandi and Fan, 2010). Thus, the 2011 spike happened on top of historically high prices. Another interesting feature is the difference between wholesale and retail prices. Although prices at both the wholesale and the retail level rose sharply at the beginning of 2011, wholesale prices retracted more than retail prices. In fact, around May 2011, retail prices and wholesale prices seemed to move in opposite directions. A widening gap between wholesale and retail prices is likely to hurt farmers who participate in the market, as their revenue from sales decreases whereas the cost of purchases increases. Fig. 1. View largeDownload slide Average prices of three commodities over time. Source: Authors’ calculations based on FIT Uganda. Note: The figure shows the average over commodities (matooke, maize and groundnuts) and five markets. Fig. 1. View largeDownload slide Average prices of three commodities over time. Source: Authors’ calculations based on FIT Uganda. Note: The figure shows the average over commodities (matooke, maize and groundnuts) and five markets. Figure 2a shows the evolution of prices of three unprocessed staple foods over time (averaged over the different market locations). Since these are all staple foods, one would expect these prices to move closely together as substitution in consumption keeps the prices from diverging too much. However, there were some episodes when prices for different staple crops moved in opposite directions. For instance, from June 2009 to November 2009, maize became cheaper whereas cassava became more expensive. This is likely to affect the terms of trade of households and can be detrimental to households that sold maize and bought cassava with the revenues. Figure 2b shows the evolution over time of higher value or lightly processed commodities. Also here, there are marked differences. The price of groundnuts increased significantly over the period, whereas the price of maize flour increased only moderately. Fig. 2. View largeDownload slide Price evolution of selected crops. Source: Authors’ calculations based on FIT Uganda. Note: Figures show the average of wholesale and retail prices and average across five markets. Fig. 2. View largeDownload slide Price evolution of selected crops. Source: Authors’ calculations based on FIT Uganda. Note: Figures show the average of wholesale and retail prices and average across five markets. 2.3. Welfare and poverty impact of price changes Figures 3a plots the percentage change in welfare, proxied by consumption expenditure per capita, as a result of actual price changes between June 2008 and December 2011. In particular, we calculate the metric defined in equation (2) and then take the average over all households. Since we have monthly data for prices between 2008 and 2011, we can calculate equation (2) evaluated at prices in each month over this period relative to some baseline price level, hence the results are plotted as a time series.5 In this way, we do not have to choose an endpoint at which to evaluate the price evolutions, but we can investigate what would have been the outcome if, for instance, the government of Uganda decided to evaluate the impact at a particular point in time. Fig. 3. View largeDownload slide Welfare impact index and short-run poverty. Source: Authors’ calculations based on FIT Uganda and UNPS 2009/10. Fig. 3. View largeDownload slide Welfare impact index and short-run poverty. Source: Authors’ calculations based on FIT Uganda and UNPS 2009/10. We see that if the government would evaluate the evolution of prices anywhere between September 2008 and June 2009, the first-order impact would have yielded a positive welfare impact. However, effects are generally less than 1 per cent of welfare. If wi,t=1 and pj,t=1 are fixed after June 2009, welfare changes turn negative again, and become larger over time, reaching about 2 per cent for the comparison between the baseline period and September 2010. The year 2011 is volatile with a welfare increase of more than 2.5 per cent comparing the baseline with March 2011 and a welfare loss of 2.5 per cent comparing the baseline with December 2011. The evolution of the welfare impact over time is in line with the movement of the price series in Figures 1 and 2. Episodes where prices are higher than at baseline, such as around March and April 2011, correspond to an increase in welfare, suggesting many farmers are net sellers of the affected commodities. Note also that at the end of the sample, wholesale prices decrease while retail prices level out. At the same time, the price of cassava, a crop bought by the poor, shoots up, while the price of maize, the main income earner for smallholders, levels out. Together, this explains the sharp reduction in first-order welfare impact between the baseline and December 2011.6 To calculate the impact of price changes on poverty, we use the first-order welfare impact to calculate real per capita consumption expenditure for each household at different sets of prices. In particular, for each household, we add (∑iriΔwi-∑jqjΔpj) to per capita consumption expenditure at baseline price levels, and compare this to the official poverty line in Uganda. Figure 3b plots the poverty headcount, defined as the percentage of people that fall below the poverty line, for each monthly set of prices. In the reference period, headcount poverty stands at 23.1 per cent, represented by the horizontal line. The correspondence between the welfare changes depicted in Figure 3a and the poverty rates seem reasonable, with positive first-order welfare impacts corresponding to poverty reductions, and vice versa. The exception seems to be when the price situation around the beginning of 2011 is considered the endpoint of a first-order analysis. A sharp increase in welfare seems to coincide with an equally sharp increase in poverty. This suggests that the price pattern in the beginning of 2011 benefited richer farmers, while at the same time hurting farmers that were just above the poverty threshold. In other words, there is considerable heterogeneity in the price impact of price changes on individual farmer’s well-being and position relative to the poverty line. We can explore some of this heterogeneity a bit further by looking at the first-order effects of price changes for particular groups of people. Because urban households are generally net buyers and would be more adversely affected by higher food prices, we first divided the sample into rural and urban households and consider the evolution of welfare separately (see Figure 4a). However, since there is a great deal of agricultural activity in peri-urban areas in Uganda (and to facilitate comparison to the CGE results below that use similar groups of representative households), we also distinguish rural farmers and urban non-farming households who reside either in secondary towns/cities or in Kampala metropolitan area. Results are in Figure 4b. Fig. 4. View largeDownload slide Welfare impact for different types of households. Source: Authors’ calculations based on FIT Uganda and UNPS 2009/10. Fig. 4. View largeDownload slide Welfare impact for different types of households. Source: Authors’ calculations based on FIT Uganda and UNPS 2009/10. Figure 4a shows that, for most of the period, recorded price levels have benefited, or at least not hurt, the rural population as judged by the first-order effects and relative to average price levels of 2008. It seems that, in general, the higher prices observed between 2008 and 2011 were beneficial to rural households, suggesting many of them are net sellers in the market for the commodities that saw the highest price hikes and net buyers for commodities for which prices increased less dramatically.7 Especially, the price hike in the beginning of 2011 resulted in large welfare impacts of over 5 per cent of household welfare. Only in mid-2010, when prices were lowest, we find a negative first-order impact for the rural population, but the magnitude is small. As expected, urban households were adversely affected by high prices, as they are generally less likely to sell on the market. Losses are substantial, reaching well over 10 per cent if December 2011 is taken as the endpoint. Figure 4b shows that these losses become more important for urban dwellers that are not involved in farming. 3. General-equilibrium impact of food price shocks on the Ugandan economy 3.1. Model and data CGE models are particularly suited to the analysis of shocks with complex economy-wide implications, including direct (first-order) and indirect (second-order) effects that become particularly relevant under a longer time frame. First, they account for behavioural responses – primarily to relative price changes – and interaction between producers, consumers and government within a whole-economy framework that includes markets for labour, capital and commodities. Second, through simulation, the impact of complex policy or economic shocks can be better understood and decomposed, e.g. by sector/product or household type, or in the short or longer run, with the latter characterised by assumptions of greater flexibility in producers’ or consumers’ responses to changing market conditions, particularly relative price shocks. Third, by tracking changes in prices, outputs and household incomes, these models – particularly more disaggregated ones – provide a theoretically consistent framework for analysing the welfare effects of policies and external shocks. For this study, we calibrate the International Food Policy Research Institute (IFPRI) standard CGE model (Löfgren, Harris and Robinson, 2002) to a 2007 SAM for Uganda developed by Thurlow (2012). SAM accounts define the economic agents and markets included in the CGE model, such as productive activities, commodity markets, factor markets, current accounts of domestic institutions (i.e. households, government and incorporated business enterprises), a capital account that captures savings and investments and a rest-of-the-world account. The SAM includes 50 activities and 50 commodity accounts, including 21 subsectors in the agriculture, forestry and fishing sectors as well as five agro-processing sectors. There are also six types of production factors (family farm labour, unskilled labour, skilled labour, capital, cattle stock and land) and five representative household groups (rural farm, rural non-farm, urban farm, urban non-farm and Kampala households). 3.2. Model behavioural assumptions While SAM entries record transactions that took place between economic agents during a particular year, the CGE model itself is defined by behavioural relationships. Economic relationships are modelled as a mix of nonlinear and linear equations that govern how economic agents in the model respond to changes in market conditions brought about by exogenous policy or economic shocks. A detailed model description appears in Löfgren, Harris and Robinson (2002); below we focus briefly on some of the key features of the model. In the CGE model, we assume that producers (activities) maximise profits when combining intermediate inputs with primary factors of production such as land, labour and capital. Production is specified using nested constant elasticity of substitution (CES) functions. These reflect sector-specific technologies and allow for imperfect substitution between factors of production. Factor incomes are distributed to their owners. Income from self-employed farm labour, land and cattle stock accrues to farm households, while income from the remaining factors accrues to non-farm and urban households. Agricultural land is an interesting case worth highlighting. We assume that all returns to land accrue to those who actively farm the land, irrespective of whether they own or rent the land. In cases where land is rented from land owners, who may or may not be farmers themselves, rental income is channelled via the capital account in the same way any investment income is channelled. Households also earn income from government transfers or in the form of remittances from other households or from abroad. Households save and pay taxes and use the balance of income for consumption expenditure. The SAM was constructed on the assumption that all consumption is marketed consumption, i.e. home consumption is valued at market prices. Consumption demand is governed by a linear expenditure system (LES) which assumes that a certain portion of disposable income is allocated to subsistence consumption that is unresponsive to price changes. During calibration of the model, we arbitrarily set the Frisch parameters to −1 for all households, and use income elasticities estimated by Wiebelt et al. (2011), which results in approximately 14 per cent of disposable income being allocated to subsistence consumption in the base. Producers supply their output to national product markets. Transaction costs and consumption taxes separate producer and consumer prices. A constant elasticity of transformation (CET) function allows production to shift imperfectly between domestic and foreign markets depending on the relative prices of exports and domestic products. Similarly, a CES function governs the mix of imported and domestically supplied goods demanded locally, depending on relative prices of imported or domestically produced goods. Elasticity values for CES and CET functions are taken from Wiebelt et al. (2011). In order to facilitate a more nuanced analysis of household poverty, we link our modelled changes in prices and consumption at the representative household group level to corresponding member households in the Uganda National Household Survey (2005/06). A survey-based poverty module, a standard feature of IFPRI CGE models, then computes changes in income poverty measures by applying model-estimated percentage changes in consumption for household groups to consumption reported by individual households in the survey. The initial and new consumption levels are compared against an official poverty line, defined for rural and urban households, and estimates of changes in poverty are obtained. We report poverty changes at national level and for various household subgroups. This allows for a direct comparison with the first-order welfare impact estimated in the previous section. At least one perceived limitation of CGE models is their assumption that information flows freely and markets perform well. This is not completely unrealistic for agriculture and food markets in a country such as Uganda, where there are many producers (sellers) and consumers (buyers) and where markets are reasonable well integrated spatially. Standard CGE models also capture some rigidities in markets: for example, in instances where traders extract significant rents due to lack of information or market access by farmers, these are reflected in high trade and transport margins. As discussed later, we restrict mobility of some factors of production to better represent the functioning of factor markets. 3.3. Modelled simulations The main objective of the CGE simulations is to explore the impact of agricultural, food and fuel price changes over the period 2008–2011 relative to a base case scenario representing the economy of Uganda in 2007. As a small country, Uganda is considered a price taker in world markets. Domestic prices, on the other hand, are endogenously determined and can be influenced by exogenous changes in world prices (via trade linkages), policy shocks that alter domestic demand and supply relationships, or economic shocks that affect costs of production. Whereas our world price shock simulations are based on actual observed price changes over 2008–2011, we cannot exogenously replicate observed domestic price changes, since these are endogenously determined in the model. The four simulations presented in this study (sim1–sim4) are run sequentially, each building on the previous simulation by adding a new dimension to the simulated shock. Simulation results are reported as changes relative to the base case in 2007, but given the simulation design, results can also be compared against previous simulations to identify the marginal change of each additional shock introduced. In the first simulation, sim1, foreign currency-denominated world import and export prices of agricultural and processed food commodities are shocked exogenously, based on observed world price changes during 2008–2011. In particular, we model a 50 per cent increase in world maize prices; a 25 per cent increase in rice and other cereals prices; a 35 per cent increase in processed grain prices (flour); and a 25 per cent increase in processed meat, fish and other food products. The second simulation, sim2, implements the same food price shock as sim1, but also increases world prices of key export products, specifically coffee prices (100 per cent) and tea, cocoa and vanilla prices (300 per cent). This particular simulation, therefore, captures the combined effect of observed changes in international food and cash crop prices for both imported and exported commodities. The third simulation, sim3, models the commodity price shocks of sim2 and adds to this a 6 per cent rise in global fuel prices. Fuel prices have been identified as a major cause of commodity price increases globally (Headey, Malaiyandi and Fan, 2010). Domestically, they cause transport cost increases, which tend to be passed on to consumers. For example, UBOS (2012) attributes inflation of 18.7 per cent in 2011 partly to food, beverage, and clothing price increases, but also to high transport costs (transport fares increased by 17.8 per cent that year), which in turn have been linked to rising fuel prices. World fuel prices spiked in 2008 but subsequently retreated to lower levels again, so that the net increase during 2008–2011 was only 6 per cent. The fourth simulation, sim4, simulates the effect of a 7.6 per cent decline in the availability of agricultural cropland, together with the world price shocks of sim3. Although in the static version of the model we do not account for population changes over time, FAOSTAT (2016) data show that agricultural land supply declined by 7.6 per cent over 2008–2011 in per capita terms, which reflects the fact that Uganda is nearing (or may have reached) its land frontier (see Jayne, Chamberlin and Headey, 2014). This simulation, therefore, serves as a proxy for declining per capita availability of land. Although limited mechanisation and rising rural–urban migration also have implications for agricultural production, land constraints are arguably the most pressing issue in the current context in Uganda, hence the focus on this production factor. Given the volatility and heterogeneity in yield growth rates over 2008–2011 (FAO, 2016), we do not introduce productivity shocks into this simulation. The simulation, therefore, assumes a constant level of total factor productivity, but declining land availability, which will lead to an overall decline in crop production and endogenous price increases, subject to producers’ land allocation choices. Combining these domestic price shock simulations with international price shocks complements earlier CGE analyses of price shocks in Uganda, which have only focused on international prices (e.g. Boysen and Matthews, 2012; Matovu and Twimukye, 2009). 3.4. Closure rules Our analysis focuses on the medium to longer run in that we assume farmers are able to reallocate land to more profitable crops in response to the shocks imposed. We further assume that labour factors are fully employed and mobile across economic sectors, although the movement of family farm labour is restricted to agricultural subsectors to emphasise the difficulty family farm workers have in accessing non-farm employment opportunities. Capital stock, employed largely in non-agricultural sectors, is activity-specific or immobile to impose some rigidities on non-agricultural sectors’ production capacity levels. The overall level of capital stock is also fixed in static CGE models. This means that even as investment levels change, which is determined by the level of savings in the economy (i.e. a savings-driven investment closure), we do not consider the future impact this may have on overall levels of capital stock and production capacity. Foreign savings are fixed exogenously, but the trade balance and the exchange rate are flexible, with the latter serving the role of equilibrating variable to the current-account balance. The domestic consumer price index (CPI) is treated as the numeraire in the model, which means all endogenous price changes are expressed in real terms and relative to a fixed CPI. We, therefore, model the real economy and implications of relative price changes, but not the effect of inflation itself. One advantage of CGE models is the relative ease with which additional simulations or robustness checks under alternative closure rules can be conducted. For example, under an investment-driven savings closure, which assumes households adjust their savings rates to ensure a target level investment is achieved, the macroeconomic results would look similar, but household-level welfare is affected as consumption is shifted towards current investment. As a result of space constraints, and given the emphasis on comparing results across different modelling approaches, we only present results from a limited set of simulations under a single set of closure rules and assumptions which we feel are most appropriate to the situation at hand, although this remains a subjective decision. 3.5. Results and discussion Rising world prices tend to cause shifts in consumer demand towards cheaper, locally produced substitutes, while producers shift production toward more profitable export markets. Such structural shifts may sometimes result in a reduction in domestic absorption (ABS), defined as the sum of consumption (C), investment (I) and government spending (G), and considered a measure of aggregate domestic welfare. The effects of rising exports (X) and declining imports (M) are an improvement in the trade balance (X – M). Simple reorganisation of the GDP equation (3) shows why this will generally lead to a decline in domestic absorption – and hence aggregate domestic welfare – for a given level of GDP (Arndt et al., 2008): GDP=C+I+G+(X–M)=ABS+(X–M)ABS=GDP–(X–M) (3) However, if GDP rises by more than the increase in the trade surplus, absorption may still increase. Although this is fairly unlikely in instances where a country’s terms of trade worsens, it may be likely if the terms of trade improve and increased export opportunities lead to an increase in household disposable income and hence domestic absorption. Absorption may also increase as a result of rising government expenditure and/or investment expenditure. Table 1 reports key macroeconomic results from our simulations, including GDP measured at market prices. Table 2, in turn, presents GDP at factor cost (or value-added), both nationally and for individual sectors. We also present changes in factor incomes by type of factor. Commodity taxes and trade and transport margins explain the difference in GDP at market prices and GDP at factor costs. Table 1. Changes in GDP at market prices and macroeconomic indicators Changes relative to base (%) World food prices (sim1) Food & export crop prices (sim2) Food, export & fuel prices (sim3) Price shocks & domestic supply shock (sim4) Macroeconomic indicators Terms of trade (ToT) 3.2 16.3 15.9 15.9 Real exchange rate −5.5 −7.3 −7.0 −6.7 World price index 3.7 8.2 8.5 8.5 Domestic price index 0.1 0.7 0.6 0.0 Real GDP at market prices (% change) −0.3 0.0 −0.1 −0.9 Absorption 0.2 2.7 2.5 1.6 Consumption 0.6 4.3 4.0 2.7 Investment −1.3 −1.9 −1.9 −1.6 Government 0.0 0.0 0.0 0.0 Exports 1.8 −2.5 −2.4 −3.1 Imports 2.6 8.4 8.0 7.3 Changes relative to base (%) World food prices (sim1) Food & export crop prices (sim2) Food, export & fuel prices (sim3) Price shocks & domestic supply shock (sim4) Macroeconomic indicators Terms of trade (ToT) 3.2 16.3 15.9 15.9 Real exchange rate −5.5 −7.3 −7.0 −6.7 World price index 3.7 8.2 8.5 8.5 Domestic price index 0.1 0.7 0.6 0.0 Real GDP at market prices (% change) −0.3 0.0 −0.1 −0.9 Absorption 0.2 2.7 2.5 1.6 Consumption 0.6 4.3 4.0 2.7 Investment −1.3 −1.9 −1.9 −1.6 Government 0.0 0.0 0.0 0.0 Exports 1.8 −2.5 −2.4 −3.1 Imports 2.6 8.4 8.0 7.3 Source: CGE model results. Table 1. Changes in GDP at market prices and macroeconomic indicators Changes relative to base (%) World food prices (sim1) Food & export crop prices (sim2) Food, export & fuel prices (sim3) Price shocks & domestic supply shock (sim4) Macroeconomic indicators Terms of trade (ToT) 3.2 16.3 15.9 15.9 Real exchange rate −5.5 −7.3 −7.0 −6.7 World price index 3.7 8.2 8.5 8.5 Domestic price index 0.1 0.7 0.6 0.0 Real GDP at market prices (% change) −0.3 0.0 −0.1 −0.9 Absorption 0.2 2.7 2.5 1.6 Consumption 0.6 4.3 4.0 2.7 Investment −1.3 −1.9 −1.9 −1.6 Government 0.0 0.0 0.0 0.0 Exports 1.8 −2.5 −2.4 −3.1 Imports 2.6 8.4 8.0 7.3 Changes relative to base (%) World food prices (sim1) Food & export crop prices (sim2) Food, export & fuel prices (sim3) Price shocks & domestic supply shock (sim4) Macroeconomic indicators Terms of trade (ToT) 3.2 16.3 15.9 15.9 Real exchange rate −5.5 −7.3 −7.0 −6.7 World price index 3.7 8.2 8.5 8.5 Domestic price index 0.1 0.7 0.6 0.0 Real GDP at market prices (% change) −0.3 0.0 −0.1 −0.9 Absorption 0.2 2.7 2.5 1.6 Consumption 0.6 4.3 4.0 2.7 Investment −1.3 −1.9 −1.9 −1.6 Government 0.0 0.0 0.0 0.0 Exports 1.8 −2.5 −2.4 −3.1 Imports 2.6 8.4 8.0 7.3 Source: CGE model results. Table 2. Changes in GDP at factor cost (value added) for selected sectors GDP or factor income shares in the base (%) Changes relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (2007) (sim1) (sim2) (sim3) (sim4) National GDP at factor cost 100.0 −0.4 −0.4 −0.4 −1.2 Agriculture 22.7 −1.4 −1.0 −1.1 −4.8 Cereal crops 2.5 20.8 9.6 9.7 0.1 Roots crops 4.0 −4.0 −3.6 −3.7 −9.3 Matooke 2.4 −3.4 −3.1 −3.1 −7.6 Pulses and oil seeds 2.9 −12.0 −14.7 −14.7 −20.4 Horticulture 0.2 −3.5 −3.2 −3.3 −7.2 Export crops 2.2 −13.8 4.4 4.5 −1.6 Livestock 1.6 0.7 −0.8 −0.8 0.3 Industry 27.2 −0.1 −0.4 −0.4 −0.2 Meat processing 0.1 −0.9 −2.2 −2.2 −0.7 Grain processing 0.7 0.3 0.3 0.3 −0.3 Other food processing 1.5 6.5 3.7 3.8 4.2 Services 50.1 −0.1 0.0 −0.1 −0.2 Factor incomes 100.0 −0.4 −0.4 −0.4 −1.2 Labour 28.2 1.0 0.3 0.2 −2.2 Capital 61.2 −2.0 −3.1 −3.1 −4.7 Land 10.5 5.2 13.7 13.8 21.3 GDP or factor income shares in the base (%) Changes relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (2007) (sim1) (sim2) (sim3) (sim4) National GDP at factor cost 100.0 −0.4 −0.4 −0.4 −1.2 Agriculture 22.7 −1.4 −1.0 −1.1 −4.8 Cereal crops 2.5 20.8 9.6 9.7 0.1 Roots crops 4.0 −4.0 −3.6 −3.7 −9.3 Matooke 2.4 −3.4 −3.1 −3.1 −7.6 Pulses and oil seeds 2.9 −12.0 −14.7 −14.7 −20.4 Horticulture 0.2 −3.5 −3.2 −3.3 −7.2 Export crops 2.2 −13.8 4.4 4.5 −1.6 Livestock 1.6 0.7 −0.8 −0.8 0.3 Industry 27.2 −0.1 −0.4 −0.4 −0.2 Meat processing 0.1 −0.9 −2.2 −2.2 −0.7 Grain processing 0.7 0.3 0.3 0.3 −0.3 Other food processing 1.5 6.5 3.7 3.8 4.2 Services 50.1 −0.1 0.0 −0.1 −0.2 Factor incomes 100.0 −0.4 −0.4 −0.4 −1.2 Labour 28.2 1.0 0.3 0.2 −2.2 Capital 61.2 −2.0 −3.1 −3.1 −4.7 Land 10.5 5.2 13.7 13.8 21.3 Source: CGE model results. Table 2. Changes in GDP at factor cost (value added) for selected sectors GDP or factor income shares in the base (%) Changes relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (2007) (sim1) (sim2) (sim3) (sim4) National GDP at factor cost 100.0 −0.4 −0.4 −0.4 −1.2 Agriculture 22.7 −1.4 −1.0 −1.1 −4.8 Cereal crops 2.5 20.8 9.6 9.7 0.1 Roots crops 4.0 −4.0 −3.6 −3.7 −9.3 Matooke 2.4 −3.4 −3.1 −3.1 −7.6 Pulses and oil seeds 2.9 −12.0 −14.7 −14.7 −20.4 Horticulture 0.2 −3.5 −3.2 −3.3 −7.2 Export crops 2.2 −13.8 4.4 4.5 −1.6 Livestock 1.6 0.7 −0.8 −0.8 0.3 Industry 27.2 −0.1 −0.4 −0.4 −0.2 Meat processing 0.1 −0.9 −2.2 −2.2 −0.7 Grain processing 0.7 0.3 0.3 0.3 −0.3 Other food processing 1.5 6.5 3.7 3.8 4.2 Services 50.1 −0.1 0.0 −0.1 −0.2 Factor incomes 100.0 −0.4 −0.4 −0.4 −1.2 Labour 28.2 1.0 0.3 0.2 −2.2 Capital 61.2 −2.0 −3.1 −3.1 −4.7 Land 10.5 5.2 13.7 13.8 21.3 GDP or factor income shares in the base (%) Changes relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (2007) (sim1) (sim2) (sim3) (sim4) National GDP at factor cost 100.0 −0.4 −0.4 −0.4 −1.2 Agriculture 22.7 −1.4 −1.0 −1.1 −4.8 Cereal crops 2.5 20.8 9.6 9.7 0.1 Roots crops 4.0 −4.0 −3.6 −3.7 −9.3 Matooke 2.4 −3.4 −3.1 −3.1 −7.6 Pulses and oil seeds 2.9 −12.0 −14.7 −14.7 −20.4 Horticulture 0.2 −3.5 −3.2 −3.3 −7.2 Export crops 2.2 −13.8 4.4 4.5 −1.6 Livestock 1.6 0.7 −0.8 −0.8 0.3 Industry 27.2 −0.1 −0.4 −0.4 −0.2 Meat processing 0.1 −0.9 −2.2 −2.2 −0.7 Grain processing 0.7 0.3 0.3 0.3 −0.3 Other food processing 1.5 6.5 3.7 3.8 4.2 Services 50.1 −0.1 0.0 −0.1 −0.2 Factor incomes 100.0 −0.4 −0.4 −0.4 −1.2 Labour 28.2 1.0 0.3 0.2 −2.2 Capital 61.2 −2.0 −3.1 −3.1 −4.7 Land 10.5 5.2 13.7 13.8 21.3 Source: CGE model results. The combined effect of international food price movements in sim1 is a net improvement in Uganda’s terms of trade (3.2 per cent) and an associated appreciation of the real exchange rate (5.5 per cent). Favourable export opportunities for maize and rising profitability in producing import-competing rice and other cereals create incentives for farmers to allocate resources to the cereals sector, resulting in a 20.8 per cent increase in cereals GDP (Table 2). Although the agricultural sector as a whole contracts by 1.4 per cent, increasing returns to cereals benefits the large share of rural farm households that engage in cereals production, which explains the observed rise in consumption expenditure (0.6 per cent) and consequently domestic absorption (0.2 per cent) (Table 1). The latter is indicative of a marginal increase in aggregate welfare in the economy. Of course, while the decline in current investment has no impact on current capital stock levels, it could negatively affect future productive capacity in the economy, and hence also future welfare levels. Table 3 reports changes in consumer prices for selected food commodities. As anticipated, maize prices increase fairly sharply in sim1 (13.2 per cent), which has important spillover effects for prices in the grain processing (8.5 per cent), animal feeds (7.6 per cent), and cattle and sheep (10.6 per cent) sectors due to strong inter-industry linkages. Prices of key staples such as roots, matooke and beans all increase by around 6 per cent as land and labour factors are reallocated towards import-competing and export crop sectors. Table 3. Real changes in consumer prices for selected commodities Changes relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (sim1) (sim2) (sim3) (sim4) Agricultural prices Maize 13.2 14.3 14.1 19.2 Rice 9.3 10.4 10.2 15.4 Other cereals 8.8 9.1 9.0 13.5 Cassava 6.3 10.1 9.9 18.2 Irish potatoes 5.0 8.4 8.2 13.9 Sweet potatoes 6.3 10.4 10.1 18.3 Matooke 6.4 10.2 10.0 18.2 Oil seeds 5.9 9.2 9.0 16.0 Beans 6.8 10.8 10.6 18.4 Vegetables 4.2 7.3 7.1 10.4 Fruit 5.0 8.5 8.4 13.4 Cattle and sheep 10.6 11.2 11.1 8.8 Poultry 0.8 4.6 4.3 2.4 Processed food prices Meat processing 6.2 6.3 6.1 4.1 Grain processing 8.5 8.8 8.6 7.9 Animal feeds 7.6 6.5 6.3 9.1 Other food processing 5.0 4.0 3.8 2.6 Changes relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (sim1) (sim2) (sim3) (sim4) Agricultural prices Maize 13.2 14.3 14.1 19.2 Rice 9.3 10.4 10.2 15.4 Other cereals 8.8 9.1 9.0 13.5 Cassava 6.3 10.1 9.9 18.2 Irish potatoes 5.0 8.4 8.2 13.9 Sweet potatoes 6.3 10.4 10.1 18.3 Matooke 6.4 10.2 10.0 18.2 Oil seeds 5.9 9.2 9.0 16.0 Beans 6.8 10.8 10.6 18.4 Vegetables 4.2 7.3 7.1 10.4 Fruit 5.0 8.5 8.4 13.4 Cattle and sheep 10.6 11.2 11.1 8.8 Poultry 0.8 4.6 4.3 2.4 Processed food prices Meat processing 6.2 6.3 6.1 4.1 Grain processing 8.5 8.8 8.6 7.9 Animal feeds 7.6 6.5 6.3 9.1 Other food processing 5.0 4.0 3.8 2.6 Source: CGE model results. Note: Since the CPI is the numéraire in the model all commodity price changes are considered to be real price relative to a fixed CPI. By extension, this implies that relative food price increases (as shown in the table) would be offset by relative price decreases for non-food commodities (not presented in the table). Table 3. Real changes in consumer prices for selected commodities Changes relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (sim1) (sim2) (sim3) (sim4) Agricultural prices Maize 13.2 14.3 14.1 19.2 Rice 9.3 10.4 10.2 15.4 Other cereals 8.8 9.1 9.0 13.5 Cassava 6.3 10.1 9.9 18.2 Irish potatoes 5.0 8.4 8.2 13.9 Sweet potatoes 6.3 10.4 10.1 18.3 Matooke 6.4 10.2 10.0 18.2 Oil seeds 5.9 9.2 9.0 16.0 Beans 6.8 10.8 10.6 18.4 Vegetables 4.2 7.3 7.1 10.4 Fruit 5.0 8.5 8.4 13.4 Cattle and sheep 10.6 11.2 11.1 8.8 Poultry 0.8 4.6 4.3 2.4 Processed food prices Meat processing 6.2 6.3 6.1 4.1 Grain processing 8.5 8.8 8.6 7.9 Animal feeds 7.6 6.5 6.3 9.1 Other food processing 5.0 4.0 3.8 2.6 Changes relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (sim1) (sim2) (sim3) (sim4) Agricultural prices Maize 13.2 14.3 14.1 19.2 Rice 9.3 10.4 10.2 15.4 Other cereals 8.8 9.1 9.0 13.5 Cassava 6.3 10.1 9.9 18.2 Irish potatoes 5.0 8.4 8.2 13.9 Sweet potatoes 6.3 10.4 10.1 18.3 Matooke 6.4 10.2 10.0 18.2 Oil seeds 5.9 9.2 9.0 16.0 Beans 6.8 10.8 10.6 18.4 Vegetables 4.2 7.3 7.1 10.4 Fruit 5.0 8.5 8.4 13.4 Cattle and sheep 10.6 11.2 11.1 8.8 Poultry 0.8 4.6 4.3 2.4 Processed food prices Meat processing 6.2 6.3 6.1 4.1 Grain processing 8.5 8.8 8.6 7.9 Animal feeds 7.6 6.5 6.3 9.1 Other food processing 5.0 4.0 3.8 2.6 Source: CGE model results. Note: Since the CPI is the numéraire in the model all commodity price changes are considered to be real price relative to a fixed CPI. By extension, this implies that relative food price increases (as shown in the table) would be offset by relative price decreases for non-food commodities (not presented in the table). When food prices rise relative to a fixed numeraire (CPI), non-food prices become relatively cheaper. The welfare implications of the simulation are, therefore, not immediately evident, and depend on how households’ consumption is affected by relative food and non-food price shifts and income changes. We use the equivalent variation (EV) measure to investigate changes in welfare due to price changes. Table 4 shows that all households experience an increase in welfare in sim1, with rural farm and urban households gaining the most (0.6 and 1.0 per cent, respectively). Urban households in this instance are well positioned to benefit from increased trade opportunities particularly in processed food sectors. Non-farm households experience a slight decline in welfare levels as a result of sharp food price increases from which they do not benefit in the same way as farm households. Table 4. Long run changes in household welfare and poverty Changes in welfare (EV) relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (sim1) (sim2) (sim3) (sim4) All households 0.5 4.0 3.7 2.4 Rural farm households 0.6 5.3 5.1 3.9 Rural non-farm households −0.3 2.1 1.6 −0.1 Kampala metro 0.3 2.6 2.3 1.0 Urban farm households 1.0 4.0 3.5 2.2 Urban non-farm households 0.4 3.0 2.6 1.2 Changes in welfare (EV) relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (sim1) (sim2) (sim3) (sim4) All households 0.5 4.0 3.7 2.4 Rural farm households 0.6 5.3 5.1 3.9 Rural non-farm households −0.3 2.1 1.6 −0.1 Kampala metro 0.3 2.6 2.3 1.0 Urban farm households 1.0 4.0 3.5 2.2 Urban non-farm households 0.4 3.0 2.6 1.2 Poverty headcount in the base (%) (2007) Change in poverty relative to base (%-point) (sim1) (sim2) (sim3) (sim4) All households 30.0 −0.7 −6.0 −5.8 −4.5 Rural households 33.4 −0.8 −6.7 −6.4 −5.1 Rural farm 32.3 −0.9 −7.4 −7.2 −5.8 Rural non-farm 41.9 0.0 −1.1 −0.5 0.0 Urban households 11.6 −0.2 −2.5 −2.2 −1.2 Poverty headcount in the base (%) (2007) Change in poverty relative to base (%-point) (sim1) (sim2) (sim3) (sim4) All households 30.0 −0.7 −6.0 −5.8 −4.5 Rural households 33.4 −0.8 −6.7 −6.4 −5.1 Rural farm 32.3 −0.9 −7.4 −7.2 −5.8 Rural non-farm 41.9 0.0 −1.1 −0.5 0.0 Urban households 11.6 −0.2 −2.5 −2.2 −1.2 Source: CGE model and poverty module results. Note: The EV welfare measure is calculated in the CGE model and reported by representative household groups in the model; the poverty module calculates changes in poverty on the basis of the survey and therefore a different reporting structure can be adopted. Table 4. Long run changes in household welfare and poverty Changes in welfare (EV) relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (sim1) (sim2) (sim3) (sim4) All households 0.5 4.0 3.7 2.4 Rural farm households 0.6 5.3 5.1 3.9 Rural non-farm households −0.3 2.1 1.6 −0.1 Kampala metro 0.3 2.6 2.3 1.0 Urban farm households 1.0 4.0 3.5 2.2 Urban non-farm households 0.4 3.0 2.6 1.2 Changes in welfare (EV) relative to base (%) World food prices Food & export crop prices Food, export & fuel prices Price shocks & domestic supply shock (sim1) (sim2) (sim3) (sim4) All households 0.5 4.0 3.7 2.4 Rural farm households 0.6 5.3 5.1 3.9 Rural non-farm households −0.3 2.1 1.6 −0.1 Kampala metro 0.3 2.6 2.3 1.0 Urban farm households 1.0 4.0 3.5 2.2 Urban non-farm households 0.4 3.0 2.6 1.2 Poverty headcount in the base (%) (2007) Change in poverty relative to base (%-point) (sim1) (sim2) (sim3) (sim4) All households 30.0 −0.7 −6.0 −5.8 −4.5 Rural households 33.4 −0.8 −6.7 −6.4 −5.1 Rural farm 32.3 −0.9 −7.4 −7.2 −5.8 Rural non-farm 41.9 0.0 −1.1 −0.5 0.0 Urban households 11.6 −0.2 −2.5 −2.2 −1.2 Poverty headcount in the base (%) (2007) Change in poverty relative to base (%-point) (sim1) (sim2) (sim3) (sim4) All households 30.0 −0.7 −6.0 −5.8 −4.5 Rural households 33.4 −0.8 −6.7 −6.4 −5.1 Rural farm 32.3 −0.9 −7.4 −7.2 −5.8 Rural non-farm 41.9 0.0 −1.1 −0.5 0.0 Urban households 11.6 −0.2 −2.5 −2.2 −1.2 Source: CGE model and poverty module results. Note: The EV welfare measure is calculated in the CGE model and reported by representative household groups in the model; the poverty module calculates changes in poverty on the basis of the survey and therefore a different reporting structure can be adopted. Since poor households allocate a relatively large share of their budget to food items, the expectation is that rising food prices will negatively affect poverty, ceteris paribus. Yet, as shown in Table 4, poverty declines by 0.7 percentage points nationally in sim1. Rural farm households experience the greatest decline in poverty (0.9 percentage points), which suggests that the poor are not necessarily excluded from the welfare gains achieved by rural farm households as a whole. Even poor rural non-farm and urban households are seemingly not negatively affected by significant food price increases. However, closer inspection reveals that although households across the board are able to increase their real expenditure, the actual quantity of food consumed declines as households re-orient their budgets towards cheaper non-food items. For at least those households that are close to or below the food requirement, this consumption shift may have implications for food and nutrition security outcomes. The second simulation, sim2, introduces international export crop price shocks in addition to the food price shocks modelled in sim1. As this implies a further improvement in the terms of trade (16.3 per cent relative to the base) the exchange rate appreciation is also greater than before (7.3 per cent) (Table 1). In addition to cereals, traditional export crops now also become an attractive investment opportunity due to higher export prices (Table 2). As a result, GDP in the cereals sector does not expand by as much as before (9.6 per cent), which is due to two reasons: first, the sector now competes more directly with traditional export crops for limited resources; and second, the larger exchange rate appreciation lowers the competitiveness of Ugandan exports compared to sim1. Agricultural sector GDP still contracts, but not by as much as before. Once again, this simulation is associated with an increase in aggregate welfare, as evidenced by the 2.7 per cent rise in domestic absorption, which is driven mainly by the increase in consumption expenditure (4.3 per cent), which more than offsets the decline in current investment. However, somewhat surprisingly, the real value of exports declines by 2.5 per cent despite the seemingly improved opportunities to export and compete with imports. This reflects the eroding effect of the exchange rate appreciation on export earnings. At the same time, real imports rise substantially (8.4 per cent), which is partly facilitated by the exchange rate appreciation, but also relates to the shift in household consumption towards cheaper non-food items which have a greater import intensity. These trade balance effects are amplified in sim2 compared to sim1 given the larger average increases in food price now observed (Table 3). We can, therefore, conclude that rising household welfare levels, combined with relative price changes, ultimately increases the import intensity of the economy and weakens the trade balance, mainly because the export opportunities associated with relative world price changes are restricted to the food and agricultural sectors. Overall household welfare increases by 4 per cent in sim2. The measured decline in poverty is also larger than in sim1, with the national poverty rate declining by 6 per cent (Table 4). The latter is driven mostly by the 7.4 percentage point decline in the rural farm household poverty rate. As anticipated, the addition of fuel price increases in sim3 only marginally changes the results seen in sim2, owing to the fact that the worst effects of the 2008 fuel price spike had dissipated by 2011 and that fuel only accounts for 8.3 per cent of Uganda’s total imports. Nevertheless, rising fuel prices cause a slight deterioration in the terms of trade, such that the net effect in sim3 is a 15.9 per cent improvement in the terms of trade, while the exchange rate appreciation is 7 per cent. We note a slightly lower aggregate welfare effects (2.5 per cent increase in domestic absorption) as well as a slightly smaller increase in real imports than before (8 per cent) (Table 1). Sectoral GDP results are largely similar (Table 2) than in the previous simulation, while food prices increase by slightly less. Lastly, welfare and poverty effects are only slightly weaker given the inflationary effect of fuel prices for particularly trade and transport related services (Table 3). The final simulation, sim4, combines the international price shocks of sim3 with a 7.6 per cent decline in the area of land supply, which is designed to simulate the effect of a decline in cultivated land in per capita terms. The terms of trade effect in sim4 are similar to that in sim3 (i.e. 15.9 per cent improvement), but land constraints ultimately lead to a much-weakened aggregate welfare effect (domestic absorption rises only 1.6 per cent) and a larger decline in agricultural exports (3.1 per cent) (Table 1). Agricultural GDP contracts by 4.8 per cent, while gains observed in earlier simulations in some subsectors, such as cereals or export crops, are wiped out (Table 2). Particularly, sharp GDP losses are also recorded in many of the large non-tradable sectors, such as pulses and oil seeds, root crops and matooke. This simulation is associated with large agricultural crop price increases, in some cases almost double what was measured in earlier simulations (Table 3). Agricultural crop price increases ensure that, despite the contraction in agriculture, returns to factors of production, especially land and family farm labour, increase sharply. The implication is that producers – and by extension households linked to producers via the factor market – ultimately experience welfare gains, albeit not to the same extent as in earlier simulations. More specifically, with the exception of rural non-farm households, all Ugandan households experience an increase in welfare (EV), with rural farm households gaining most (3.9 per cent). Poverty rates also decline across the board, especially for rural farm households (by 5.8 percentage points), and to lesser extent urban households (1.2 percentage points decline). The national poverty rate declines by 4.5 percentage points. While a detailed discussion of results obtained under alternative model closures falls beyond the scope of the current study, this is certainly an area that warrants further investigation, especially as far as the savings-investment closure is concerned. For example, under an investment-driven closure where household saving rates adjust generate a target level of investment, overall welfare (EV) levels increase slightly more than under the savings-driven closure presented earlier. However, we no longer observe the large welfare gains for farm households, i.e. welfare gains now largely accrue to urban households. The decline in poverty is slightly smaller for all household subgroups under this closure, e.g. national poverty declines by 4.1 percentage points. In summary, the results for the medium- to longer run analysis appear to be driven largely by international price shocks (sim1) and the domestic agricultural supply shock (sim3). Ultimately, however, both the macroeconomic growth changes and household-level welfare outcomes are dominated by the international price shocks, which result in positive welfare and poverty effects for most household groups but not for rural non-farm households. In general, however, the significant improvement in the terms of trade represents an opportunity for Ugandan farmers to access export markets, with positive growth and welfare spillover effects. 4. Conclusions This study analyses the impact of price changes, roughly between 2008 and 2011, on Uganda’s economy. We consider the combined impact of changes in prices of a basket of commodities on welfare and poverty. First-order results generated with an extended version of Deaton equation for measuring the short-run welfare impact of price changes are compared with longer run results generated with the aid of a full-scale CGE model that accounts for economy-wide direct and indirect effects. In the first-order analysis, we find that commodity price movements had only small welfare implications, generally changing poverty by less than 1 percentage point. However, these averages hide substantial heterogeneity, with urban households, especially those living just above the poverty line, likely to lose most as a result of higher prices. We also find that the relative price movements are important. We find that price patterns towards that end of our time series are damaging, when the price of cassava, often bought by poor households, shoots up, while the price of maize, often sold by poor farm households, remains flat. Our long-run analysis, which explicitly accounts for behavioural responses to price changes, albeit at an aggregated household group level, predicts that recent international food price changes would lead to an improvement in Uganda’s terms of trade given the country’s position as a net exporter of food and agricultural products. Domestic price shocks erode some of the gains from the terms of trade improvement associated with international price changes, but the net effect is still an important improvement in welfare and reduction in poverty, with the exception perhaps of rural non-farm households. Overall, in the long run, poverty declines by about 4.5 percentage points in our combined scenario, ceteris paribus. These diverging results should not be interpreted as evidence of superiority of either method over the other. Even though we present them side by side, it remains difficult to compare the results given the differences in the underlying assumptions. The first-order approximation is essentially an assessment a theoretical concept to assess immediate impact worst-case scenarios.8 The frugal nature of the model is its strength: its simplicity and low data requirements make it very popular among policy makers and practitioners. By using household survey data, it allows estimation of the welfare impact for any definable subgroup of households in the population. In contrast, it is hard to simulate historical changes in domestic prices with a CGE model because the prices of non-tradable crops in such a model are endogenous. The welfare impact depends partly on whether the price change is due to rising international prices, a domestic supply shock, or other exogenous factors. The fact that CGE models also work with representative economic agents puts limits on how far one can go in exploring heterogeneous responses to price changes. CGE models are also complex, requiring expert knowledge and custom data, such as a SAM and estimates of various commodity- and household-specific elasticities. At the same time, they do provide a much more comprehensive and complete picture of the impact of price changes and are flexible in what aspects of the economy should be included in the model. For example, it allows policy makers to introduce particular features of the economy, such as market imperfections. Our results may serve to add an additional layer of nuance to the debate on the costs and benefits of higher food prices (Swinnen and Squicciarini, 2012). Perhaps more importantly, the contrasting first-order and general-equilibrium outcomes should be taken as a warning for policy makers not to rely too much on a single approach. Most likely, policy makers in developing countries will be biased towards first-order analyses because of its ease of interpretation and limited data requirements mentioned above. The likely outcome that especially the urban consumers are hurt may attract more policy action because they are more vocal and visible than rural smallholder farmers. This may mean policy makers will try to insulate (urban) consumers from price increases through price support and trade policy such as export restrictions and import tariff reductions. Focusing too much on the negative short-run effects may thus jeopardise long-run benefits if policy makers attempt to dampen price signals that are essential in driving behavioural change among rural producers. At the same time, relying too much on general-equilibrium models may result in policy makers neglecting the plight of poor households with limited options to substitute or adapt. Confronted with higher food prices, even in the short run, these households may be forced to liquidate productive assets, potentially pushing them into a poverty trap (Carter and Lybbert, 2012). Therefore, policy makers may need to provide social safety nets to safeguard future productive capacity. Both analyses are thus relevant to our understanding of appropriate policy design: first-order price impact models can inform household-level emergency responses to price shocks. They may also be better at capturing heterogeneity of impacts due the fact that the unit of analysis are actual households as opposed to representative households in a CGE analysis. In the context of rapidly urbanising societies, such models may be especially useful in prompting governments to design well-targeted social safety nets such as food for work or cash or in-kind transfers. General-equilibrium analyses can help policy makers identify longer term interventions needed to support the adjustment to new economic realities. This will allow governments to create an enabling environment in which farm households can seize the opportunity created by higher prices for their products. Footnotes 1 It derives its name because it is the first-order Taylor-series expansion of compensating variation (if based on the original quantities) or EV (if based on the final quantities after adjustment to price changes) associated with a price change. 2 Probably closest to our work, a recent paper by Tiberti and Tiberti (2018) compare first-order effects to a partial equilibrium model that allows for consumption, production and labour/leisure adjustments. 3 If the percentage change in consumer prices is equal to the percentage change in the producer prices (such as when there is a proportional margin between consumer and producer prices) and if there is just one commodity, this expression collapses to equation (1), proposed by Deaton (1989). However, substantial evidence of asymmetric price transmission in agricultural commodity markets, where price reductions are passed on to producers but price increases are not, means many studies that assess first-order impact of price changes now value changes in consumer and producer prices separately (Meyer and Von Cramon-Taubadel, 2004). Dawe and Maltsoglou (2014), Minot and Dewina (2015) and Van Campenhout, Lecoutere and D’Exelle (2015) show that accounting for marketing margins, so that producer price changes differ from consumer price changes, can have a noticeable effect on the welfare impact. 4 Note that wholesale prices may be substantially higher than the price farmers receive due to transaction costs in transporting goods from the point of sale to the wholesale market. Similarly, the cost of what a household consumes from the market is likely to be higher than the retail price, as the household needs to get the crop from the market. However, we are not interested in levels, but in proportional changes in consumer prices over time (Δpj/pj) and in changes in producer prices over time (Δwi/wi). If the wholesale price in the market is the farm-gate producer price plus a proportional transaction cost to bring the crop to the market, then the proportional change in the wholesale price will be equal to the proportional change in the producer price. Similarly, if the margin between the wholesale price and the consumer price is proportional, then they will change by the same proportion. However, it is likely that some portion of the marketing margins is fixed rather than proportional. 5 We decided to take the average price level in 2008 as the baseline price level. We use simple linear interpolation to accommodate gaps in the price series data. 6 Looking back at Figure 1, we see that for the available price data, price levels are at their highest in December 2011. As mentioned before, the choice of points in time is often driven by data availability or represents extreme situations (e.g. the lowest versus the highest prices over a certain period). In this case, a government official may be tempted to take the prices situation in December 2011 as the endpoint, which would lead to the conclusion that welfare declined 2.5 per cent (and a large increase in poverty, see below). However, such a conclusion would not be consistent with the fact that, during much of the period, the first-order impact of higher prices actually seemed to increase welfare. 7 Analysis of the impact of individual commodities on welfare reveals that maize is largely responsible for this positive effect. 8 First-order estimates do not necessarily result in worst-case results, but they do generate an over-estimate of the welfare impact (positive or negative) of a price change in competitive markets. Through behavioural changes, households either mitigate negative effects or exploit positive effects. Furthermore, effects may also depend on the market distortions/failures that exit in the considered economy. Price shocks can lead to more detrimental effects if, for instance, monopolistic importers increase their margins on import flows. However, the markets we consider in this analysis are reasonably competitive. References Arndt , C. , Benfica , R. , Maximiano , N. , Nucifora , A. and Thurlow , J. ( 2008 ). Higher fuel and food prices: impacts and responses for Mozambique . Agricultural Economics 39 ( s1 ): 497 – 511 . Google Scholar Crossref Search ADS Bellemare , M. ( 2015 ). Rising food prices, food price volatility, and social unrest . American Journal of Agricultural Economics 97 ( 1 ): 1 – 21 . Google Scholar Crossref Search ADS Benson , T. , Minot , N. , Pender , P. , Robles , M. and von Braun , J. ( 2013 ). Information to guide policy responses to higher global food prices: the data and analyses required . Food Policy 38 : 47 – 58 . Google Scholar Crossref Search ADS Benson , T. , Mugarura , S. and Wanda , K. ( 2008 ). Impacts in Uganda of rising global food prices: the role of diversified staples and limited price transmission . Agricultural Economics 39 ( s1 ): 513 – 524 . Google Scholar Crossref Search ADS Boysen , O. and Matthews , A. ( 2012 ). The differentiated effects of food price spikes on poverty in Uganda. Paper prepared for the 123rd European Association of Agricultural Economists. Carter , M. R. and Lybbert , T. J. ( 2012 ). Consumption versus asset smoothing: testing the implications of poverty trap theory in Burkina Faso . Journal of Development Economics 99 ( 2 ): 255 – 264 . Google Scholar Crossref Search ADS Dawe , D. and Maltsoglou , I. ( 2014 ). Marketing margins and the welfare analysis of food price shocks . Food Policy 46 : 50 – 55 . Google Scholar Crossref Search ADS Deaton , A. ( 1989 ). Rice prices and income distribution in Thailand: a non-parametric analysis . The Economic Journal 99 ( 395 ): 1 – 37 . Google Scholar Crossref Search ADS FAO (Food and Agriculture Organization of the United Nations) . 2016 . FAOSTAT. Rome. http://faostat.fao.org/ Ferreira , F. H. G. , Fruttero , A. , Leite , P. and Lucchetti , L. ( 2013 ). Rising food prices and household welfare: evidence from Brazil in 2008 . Journal of Agricultural Economics 64 ( 1 ): 151 – 176 . Google Scholar Crossref Search ADS Headey , D. ( 2016 ). Food prices and poverty. World Bank Economic Review doi.10.1093/wber/lhw064 . Headey , D. , Malaiyandi , S. and Fan , S. ( 2010 ). Navigating the perfect storm: reflections on food, energy and financial crises . Agricultural Economics 41 ( s1 ): 217 – 228 . Google Scholar Crossref Search ADS Headey , D. and Martin , W. ( 2016 ). The impact of food prices on poverty and food security . Annual Review of Resource Economics 8 : 329 – 351 . Google Scholar Crossref Search ADS Ivanic , M. , Martin , W. and Zaman , H. ( 2012 ). Estimating the short-run poverty impacts of the 2010–11 surge in food prices . World Development 40 ( 11 ): 2302 – 2317 . Google Scholar Crossref Search ADS Jacoby , H. G. ( 2016 ). Food prices, wages, and welfare in rural India . Economic Inquiry 54 ( 1 ): 159 – 176 . Google Scholar Crossref Search ADS Jayne , T. S. , Chamberlin , J. and Headey , D. ( 2014 ). Land pressures, the evolution of farming systems, and development strategies in Africa: a synthesis . Food Policy 48 : 1 – 17 . Google Scholar Crossref Search ADS Löfgren , H. , Harris , R. and Robinson , S. ( 2002 ). A standard Computable General Equilibrium (CGE) model in GAMS. IFPRI Trade and Macroeconomics Discussion Paper 75. Washington, DC: International Food Policy Research Institute. Matovu , J. and Twimukye , E. ( 2009 ). Increasing World Prices: Blessing or Curse? EPRC Research Series, No. 61 . Kampala, Uganda : Economic Policy Research Centre . Meyer , J. and Von Cramon-Taubadel , S. ( 2004 ). Asymmetric price transmission: a survey . Journal of Agricultural Economics 55 ( 3 ): 581 – 611 . Google Scholar Crossref Search ADS Minot , N. and Dewina , R. ( 2015 ). Are we overestimating the negative impact of higher food prices? Evidence from Ghana . Agricultural Economics 46 ( 4 ): 579 – 593 . Google Scholar Crossref Search ADS Minot , N. and Goletti , F. ( 1998 ). Export liberalization and household welfare: the case of rice in Viet Nam . American Journal of Agricultural Economics 80 ( 4 ): 738 – 749 . Google Scholar Crossref Search ADS Robinson , S. , Yúnez-Naude , A. , Hinojosa-Ojeda , R. , Lewis , J. D. and Devarajan , S. ( 1999 ). From stylized to applied models: building multisector CGE models for policy analysis . The North American Journal of Economics and Finance 10 ( 1 ): 5 – 38 . Google Scholar Crossref Search ADS Simler , K. ( 2010 ). The Short-Term Impact of Higher prices on Uganda. Policy Research Series. No. 5210 . Washington DC : World Bank . Stephens , E. C. and Barrett , C. B. ( 2010 ). Incomplete credit markets and commodity marketing behaviour . Journal of Agricultural Economics 62 ( 1 ): 1 – 24 . Google Scholar Crossref Search ADS Swinnen , J. and Squicciarini , P. ( 2012 ). Mixed messages on prices and food security . Science 27-335 ( 6067 ): 405 – 406 . Google Scholar Crossref Search ADS Thurlow , J. ( 2012 ). A 2007 Social Accounting Matrix for Uganda . Washington, DC : International Food Policy Research Institute . Tiberti , L. and Tiberti , M. ( 2018 ). Food price changes and household welfare: what do we learn from two different approaches? Journal of Development Studies 54 ( 1 ): 72 – 92 . Google Scholar Crossref Search ADS UBOS (Uganda Bureau of Statistics) ( 2012 ). Statistical Abstract 2012 . Kampala, Uganda . http://www.ubos.org/onlinefiles/uploads/ubos/pdf%20documents/2012StatisticalAbstract.pdf. Van Campenhout , B. , Lecoutere , E. and D’Exelle , B. ( 2015 ). Inter-temporal and spatial price dispersion patterns and the well-being of maize producers in southern Tanzania . Journal of African Economies 24 ( 2 ): 230 – 253 . Google Scholar Crossref Search ADS Weber , M. T. , Staatz , J. M. , Holtzman , J. S. , Crawford , E. W. and Bernsten , R. H. ( 1988 ). Informing food security decisions in Africa: empirical analysis and policy dialogue . American Journal of Agricultural Economics 70 ( 5 ): 1044 – 1052 . Google Scholar Crossref Search ADS Wiebelt , M. , Pauw , K. , Matovu , J. M. , Twimukye , E. and Benson , T. ( 2011 ). Managing future oil revenues in Uganda for agricultural development and poverty reduction: A CGE analysis of challenges and options. Kiel Institute Working Papers No. 1696 (May 2011), University of Kiel, Germany. Author notes Review coordinated by Jack Peerlings © Oxford University Press and Foundation for the European Review of Agricultural Economics 2018; all rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Journal

European Review of Agricultural Economics – Oxford University Press

Published: Dec 1, 2018

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

The impact of food price shocks in Uganda: first-order effects versus general-equilibrium consequences

The impact of food price shocks in Uganda: first-order effects versus general-equilibrium consequences

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

The impact of food price shocks in Uganda: first-order effects versus general-equilibrium consequences

The impact of food price shocks in Uganda: first-order effects versus general-equilibrium consequences

References (23)

Abstract

Journal

Recommended Articles

There are no references for this article.

Our policy towards the use of cookies