This chapter details a series of steps that could be followed in order to provide an empirical basis for addressing some of the questions raised by the conceptual framework presented in Chapter 16.
These guidelines are developed as an aid to researchers attempting to draw insights from within-country, ex post analyses. The guidelines focus attention on the key aspects of the relationships depicted in Figure 16.1. These require empirical information additional to that already available in the literature, and sufficient to allow informed discussion of the two-stage links.
Given the complexity of the reforms, and the counteracting forces that reforms often generate with respect to agriculture performance and food security, identifying strong causal linkages between instances of reform and changes in the levels of food security is likely to be difficult, if not impossible. The guidelines set out below attempt to break down the linkages into their component parts, where the strength of the relevant relationships may be identified more easily. In bringing together the component parts it may be possible to infer likely directions of change in key indicators, and in some cases the relative values of such changes.
Using Figure 16.3 as a basis, the guidelines cover five main areas of investigation:
description of the main episodes of reform and changes in key indicators of agricultural production, trade and food security;
analysis of the characteristics of domestic agricultural markets in terms of changes in relative prices;
analysis of aggregate supply response;
analysis of the impact of agricultural change on household incomes and expenditure patterns;
assessment of the link between (a) agriculture sector change and national level food security; and (b) changes in household incomes and expenditures and household food security.
A focus on commodities
There are many ways of approaching the analysis of links between trade reform and food security. For example, one could start with indicators of food security and attempt to work backwards through changes in consumption patterns and then agriculture sector changes; alternatively, one could investigate the impact of a specific reform on incentives and supply response across the agriculture sector. Given the large number of possible modifying parameters, however, it is felt that focusing attention on a limited range of key commodities and tracking through the impact of a bundle of reforms on commodity price and production levels, will lead to the clearest understanding of the linkages.
Methodological approaches
The guidelines proposed below do not set out a rigid methodological approach. The method adopted for examining a particular aspect of the linkage may differ from country to country depending upon the data and expertise available and the complexity of the particular relationship being investigated, and therefore, the questions that will be relevant for each country. For example, in investigating the relationship between two sets of market prices, methodologies may range from a graphical representation to the use of co-integration analysis. Similarly, the classification of households may be based on key informant interviews or on a cluster analysis using key characteristics. The focus of these guidelines is on ex post analysis. In such cases, each step set out below may be analysed separately, in others, a more inclusive modelling approach may be used to analyse a number of the steps simultaneously, although this approach would be more generally used within an ex ante framework.
The investigation is approached as a series of sequential steps, following Figure 16.1. The framework is adapted in Figure 17.1 to concisely depict the key data requirements and proposed analytical approaches.
Figure 17.1 Analyses to link reforms, intermediate effects and food security outcomes
Step 1: Description of episodes of reform and trends in key variables
The key episodes of reform and the main policy instruments and institutional changes implemented within them should first be documented. The investigation would typically start with the situation before the reform, for example, the types of exchange rate controls, trade controls, pricing and marketing arrangements. Next past episodes of reforms should be described. Economic reforms, such as the SAPs, often come in episodes, for example, one could be in 1986, the next in 1992, then in 1997 and so on. This description would be in terms of the reform objectives, instruments in each package, the intensity of the reform, implementation (there are many examples of reversals of reforms initiated) and evaluation of effects[282].
There are some problems worth noting at this stage. One is deciding upon the time frame for the analysis. In particular, what is the pre-reform period for the benchmark against which to measure the impact? If the reform started in 1986, the pre-1986 period may look like a good candidate, but even if household survey data for that period are available - many structural changes could have occurred since then, which makes it difficult to justify the attribution of computed changes to the reform. A second problem is that several reform measures may be implemented gradually, for example, exchange rates may be devalued in steps, or trade reforms taken over several years.
A timeline of reform episodes could therefore form the basis of this documentation. For each episode of reform identified, information should be collated for the key policy instruments implemented.
Policy change and policy design: key policy instruments can be described in terms of their design and implementation. Indicators of quantitative change in the levels of the instruments are described in Table 17.1 by policy type. A qualitative ranking might judge policy reform as being ambitious, complicated, or simple.
Objectives: policies might be aimed at one or more of the following general targets:
i. economic stabilization;
ii. economic growth;
iii. diversification;
iv. agricultural or rural development;
v. income or asset redistribution;
vi. food security.
The overall trade policy stance may be characterized as one of:
i. import protection;
ii. outward oriented;
iii. two track reform (e.g. China).
As a measure of policy stance, total levels of support and protection, such as the Net Protection Coefficient and Producer Subsidy Equivalent could be calculated
Drivers of reform: reform may occur for a wide range of reasons including:
i. a change in the system of governance, for example introduction of democracy or increased transparency;
ii. pressure from lobby groups;
iii. requirements to adjust in light of bilateral or international agreements e.g. WTO, loan conditionality.
Date initiated/sequencing: A timeline of policy events including start date and change dates where relevant should be constructed.
Adherence: The implementation of key policy reform should be tracked. Slippage against original plans and changes in the level of governmental commitment should be noted
Expected outcome: Against each key component of the reform, the expected direction of change in primary indicators reflecting the desired outcome should be recorded. Where evidence to the contrary is available, this should be noted. Both objectives and expected outcomes could be treated from both the government and from IMF/World Bank/donor perspectives.
Table 17.1 Examples of indicators associated with the reform of policy
Type of reform |
Indicators for design/extent of change in policy |
Possible data source |
|
Nominal exchange rate (T) Real exchange rate (T) Incidence of devaluation/revaluation Dates of government intervention |
Central bank IMF |
Monetary Money supply Inflation Savings Interest rates |
Inflation rate (Tm): CPI, PPI Savings: consumption ratio (T) Interest rate (savings) (T) Interest rate (borrowing) (T) Role of government Independence of central bank |
Statistical yearbook Central bank Treasury |
Fiscal Budgetary constraints Introduction of user fees Taxation policy |
Budget deficit (T) Government investment (T) General taxation (T) Sectoral taxation (T) |
Statistical yearbook Central bank Treasury Agricultural yearbook |
Trade Tariff NTB Export subsidies Export incentives Export credits/guarantees Export bans Trade agreements |
Tariff bindings (T); Applied tariffs (T); Tariff revenues (T) NTB tariff equivalents (T) Expenditure on export promotion (T) |
WTO schedules FAO green volumes for countries covered Regional trade agreement documentation (e.g. COMESA) Trade yearbooks |
Agriculture sector policy Subsidies:
Price controls (e.g. pan-territorial) Risk management Direct payments R&D, extension |
By crop (and if appropriate by region): Subsidy type and level at each change (P) Other interventions type and level at each change (P) Total Expenditure by amber, green and blue box categories |
WTO Domestic Support Schedules Agricultural yearbooks |
Poverty/Food security Social spending Food stock adjustment Safety nets Strategic food reserves |
Urban vs. rural expenditure (P) Programme type and level of expenditure at each change (P) |
WTO schedules National level data |
Institutional reform State withdrawal from distribution, infrastructure and service provision Role of NGOs |
Number and type of STE Proportion of activities by level in marketing chain in private sector (P) |
WTO schedules |
The following symbols are used to denote the minimum periodicity of data required for the subsequent analysis:
(P) = at discrete intervals i.e. 5 yearly
(T) = time series annual unless specified by e.g. (Tm) = monthly
(D) = derived from collected data
Data should be collected over the period 1970 - present as a minimum.
Indicators of change in agricultural production and trade
An aggregate-level depiction of the status of the agriculture sector and of food security should be developed as a basis for developing initial hypotheses relating to the impact of reform and to give context to the succeeding analyses.
For major food and non-food commodities, time series data should be collected on:
changes in aggregate production levels in terms of the quantity of production, area harvested and average yields;
changes in the crop mix, for example: the ratios of land area devoted to food and non-food production, to marketed and subsistence production, and to traditional and non-traditional export crop production;
changes in the value of domestic agricultural production by commodity could be calculated, in combination with the price data.
Of particular interest to this research is the associated impact on the agricultural trade balance and composition. Time series data would be required to illustrate the changes in, for example, total value of imports and exports, the contribution of agriculture and non-agriculture to trade and the value of imports and exports for major food and non-food commodities. For selected food and non-food commodities changes in the direction of trade should be noted.
Indicators of change in food security indicators
National level[283] indicators of food security can be categorized in terms of Availability, Stability and Accessibility. Various indicators have been described in Part I, and include:
per caput food production
Status Quo gap and nutrition gap indicators as calculated by USDA;
daily calorific intake per capita and proportion of individuals undernourished as calculated by FAO;
food commodity price stability;
per caput food imports
per caput total food supply
food import bill;
food import bill: export (total merchandise and/or agricultural) earnings;
food self sufficiency ratio (level and variation);
food aid: food imports.
Step 2 - Relating reforms to agricultural performance
Once an overall picture of the reforms and of trends in indicators of agricultural production and of food security has been established, the next step is to determine the extent to which the reforms can be related to the indicators.
To understand the extent to which the episodes of reform documented in step 1 promoted change in the agriculture sector, it is necessary to understand the changes in the incentives that different types of producers faced, and how they responded to them.
Changes in relative prices and in quantities produced and consumed are the two most important intermediate outcomes of reforms. The challenge here is both to quantify the level of the outcomes and to establish the connection or linkages between reform measures and outcomes. Again, there is no easy approach or method to doing this in ex post analyses. Researchers often use a number of approaches and methods.
Sharma[284] provides a useful insight into the types of approaches that might be used to further inform the analysis by breaking down the intermediate impact of reform. The two most common elements of analysis at this stage are estimating price effects and supply responses. The impact resulting from changes in relative prices only (i.e. prior to changes in quantities produced and consumed) may be called first-round effects while the second-round effects would be those after taking into account quantity responses also. In some cases, both effects are analysed simultaneously, as part of the single model (discussed below). But it can be more revealing to conduct the analysis sequentially, i.e. first of prices and then quantity responses. In Figure 1, these effects are called intermediate effects. The focus in adopting these types of approaches is on one or a limited number of markets within and economy. However, a limitation of these approaches is that household expenditure is assumed to remain fixed and that changes in demand due to changes in incomes and prices tend to be ignored - i.e. the second round effects may not be fully captured.
It is proposed that these impacts be investigated on an individual commodity basis (for the reasons given above) by examining both changes in relative prices and the sources of changes in the level of commodity output. A range of commodities should be selected, according to criteria such as a ranking of products by agricultural production value, but also including emerging non-traditional crops. Although the focus of this step is upon the impact of reform on producer incentives and supply response, the selection should consider the importance of the products to the livelihoods of vulnerable groups both as producers and/or as consumers (see also Step 4).
Analysis of price data
The analysis of the impact on relative prices is the starting point because most reforms aim at correcting prices, and relative prices invariably change following economic reforms.
Thus, one can form expectations about the direction to which relative prices in the post-reform period would move. Beyond this, the analysis is complicated because it is very difficult to estimate the magnitude of the change in ex post studies (in ex ante studies one simulates a model to determine the size of the change). At times, different instruments in the reform package may impact on prices differently, some raising prices and others depressing them. It is for this reason that analysts often use a number of tools to conduct price analysis so that informed conclusions may be drawn.
For a better understanding of the influence of policy reforms on the relative prices faced by producers it may be necessary to investigate the extent of price transmission and market integration. For example, in analysing the impact of trade policy reforms it is important to establish the extent to which changes in international prices are transmitted to domestic prices. If the purpose of reform is to increase the openness of the economy, then one may expect the transmission of changes in international price levels to be more fully reflected in changes to domestic prices. The results of such analysis will provide insights into the extent to which incentives are modified by reform. It is therefore proposed that for each main commodity selected, a set of price series data derived from different segments of the domestic market within a country be analysed.
If changes in price series can be related to episodes of reform, it may therefore be possible to assess the extent to which margins, and in some cases, the strength of price transmission, have changed. Judgements could then be made about the impact of reforms on incentives. To facilitate this, a timeline of reforms impacting (directly and indirectly) on the commodity could be related to trends in the price series. A decomposition of price by source may also help to identify the extent to which reforms are contributing to price changes. Comparisons of the characteristics of price series across commodities should also be made to inform discussion of how reforms have impacted on relative prices.
On the basis of the preceding discussion, three components of price analysis may be warranted:
Price decomposition
This type of analysis is used to determine whether, and to what extent, it is the reform process that explains the evolution of market prices (producer, consumer or wholesale) of commodities. In reality, domestic prices are influenced by a number of factors, notably world prices, exchange rates, tariffs, domestic marketing margins and the institutional environment. It is useful to know the percentage contribution of these various factors to changes in market prices so that some informed conclusion may be drawn about the relationship between changes in price incentives and the reform process.
One way to conduct this analysis is to use a technique to decompose the contributions to price changes.
Assume that the domestic price of a product, for some period 0, is determined as in equation (17.1)[285]:
Pd0 = Pw0 * E0 * t0 * c0 (17.1)
where Pd is domestic price (e.g. at farm level), Pw is world price, E is exchange rate, t (or (1+t) to be exact) represents ad valorem tariff and c (or (1+c) to be exact) represents other (proportional) costs (e.g. transport and marketing costs). A similar relationship is defined for some other period (period 1) as in equation (17.2):
Pd1 = Pw1 * E1 * (1 + t1) * (1+c1) (17.2)
Taking logs (ln) of both equations and subtracting (17.1) from (17.2), one obtains equation (17.3):
(lnPd1 - lnPd0) = (lnPw1 - lnPw0) + (lnE1 - lnE0) + (lnt1- lnt0) + (lnc1- lnc0) (17.3)
Since the first-order difference of logs gives the (approximate) percentage change (after multiplying by 100), the change in the domestic price is decomposed in a way that the contributions of the four factors sum to 100 percent. The data for the first three variables are often easily available; it is only in the last two cases that researchers will face data problems. In view of this, many analysts simply combine the last two terms when conducting the decomposition - which is also useful in that this term represents the effect of domestic factors (policies and others). Table 17.2 shows an example from an application by Quiroz and Valdés[286]
Table 17.2:.Decomposition of the sources of change in domestic prices of agricultural importables, Malawi, 1980-87
Period |
Change in domestic price |
Change in world price |
Exchange rate |
Other factors |
1980-82 |
67.5 |
-16.5 |
22.1 |
61.8 |
1982-85 |
-20.3 |
-5.8 |
18.3 |
-32.8 |
1985-87 |
-10.3 |
-5.8 |
-5.0 |
0.5 |
1980-87 |
36.9 |
-28.1 |
35.4 |
29.6 |
Source: Quiroz and Valdés (1993). Decompositions based on real price and exchange rate data.
Price transmission analysis
There is a separate literature on price transmission in the context of market integration. The purpose of price transmission analysis is to determine the degree of integration of two or more markets, notably the world and domestic markets, but also spatially separated markets within a country. One of the potential effects of policy reforms is to increase the degree of market integration post reform. For example, changes in world market prices would be more strongly reflected in domestic prices following the removal of quantitative controls on trade. Similarly, competitive procurement and free internal trade following the removal of the monopoly right of marketing boards should result in much stronger transmission of price signals across domestic markets.
Price transmission analysis can therefore be used, in principle, to determine whether there is stronger evidence of transmission in the post reform period.
To conduct the analysis, one can specify a regression relationship between prices in their lagged form. In its simplest form, the specification can be written as:
Pdt = j1 + j2 Pwt + et. (17.4)
where Pdt is domestic price of a product in period t, Pwt is world price, ö1 and ö2 are parameters to be estimated and åt is the stochastic error term.
The interpretation of the estimates of the parameters should however be undertaken with caution. The estimated parameter j2 has been treated as an estimate of the transmission elasticity. This implies that as the price Pw rises by 1 percent the price Pd will rise by b percent in the long run. This type of conclusion could be misleading. For instance, the two prices could be unrelated except that they are both driven by a common trend (such as general inflation). They will therefore be cointegrated. However, a rise in Pw that was unrelated to this trend (and was a temporary shock), does not require that the price of Pd change at all. If the parameter ö2 is being thought of as a partial derivative of Pd with respect to Pw, the estimate of ö2 will almost certainly not estimate this partial derivative[287].
In analysing the extent of price transmission, given the qualifications regarding the use of the transmission elasticity as discussed above, the establishment of causality is an important first step.
In practice, the analysis of price transmission can be approached using a Vector Autoregression (VAR) model. A VAR is a simultaneous equation system in which all variables in the model are regarded as endogenous and expressed as a linear function of each variables past, or lagged values, and the lagged values of the other variables.
For example, a VAR model of Pd and Pw using 2 lags can be expressed as follows:
(17.5)
(17.6)
Where the ms are the stochastic error terms and are assumed to be uncorrelated. Each equation is estimated by ordinary least squares regression. The objective is to test whether statistically one can detect the direction of causality between Pd and Pw in a lead-lag relationship. This is done by implementing the Granger causality test[288], which consists of using an F-test to test the hypothesis, for example, with respect to equation (17.5), that = 0; that is, lagged Pw terms do not belong in regression. If the statistical test rejects this hypothesis, then we cannot say that Pw does not Granger-cause Pd.
If there is transmission from the world market to the domestic market, then Pw would commonly be found to Granger-cause Pd. Thus, one might interpret causality as being evidence of transmission.
It is important to emphasise that a lack of Granger causality does not necessarily imply an absence of transmission. Signals may be transmitted instantaneously. Therefore, it is possible that there is no Granger causality, yet transmission might be perfect. If all relevant shocks were transmitted instantaneously then this could show up as correlation between m1t and m2t.
An interpretation of the estimated coefficients and of the restrictions placed upon the model can therefore be used to provide evidence as to whether the world price is a better explanatory of the domestic price following the implementation of reforms.
Analysis of marketing, transport and other margins
Economic reforms should reduce the cost of doing business, or margins facing economic agents. Thus, margins are important intermediate impact indicators in themselves. They are also indispensable statistics for related analysis: to explain the share of consumer price received by producers; to explain the price transmission results; and to examine differential terms of trade facing small and large farmers. The analysis of margins plays an important role because many studies have shown that smaller farmers invariably face much higher margins (and transaction costs) than larger farmers, and this explains a great deal of their differential participation in commercialisation.
These analyses of price series should help to illustrate and explain why different sets of producers are differentially affected and allow comparisons across both crops and countries. The analyses would stimulate hypotheses as to the key factors causing constraints to integration and market functionality.
Data limitations will determine what is feasible, but a number of options for segmentation of the market for analysis could be considered:
market segmentation by region to give an indication as to the degree to which incentives affected by reforms pass through to remoter areas;
market segmentation by level in the commodity chain (e.g. border, wholesale, retail, farm) in order to examine the extent to which these markets are integrated;
market segmentation by producer type, for example sales by smallholders to communal markets and commercial farms to marketing boards (e.g. Zimbabwean domestic beef sales). In addition, information on the extent of engagement by semi-subsistence producers in the market should be collected.
Supply response
The analyses set out above provide a mechanism for understanding how policy change is related to price changes and thus to changes in the incentives facing producers. Having analysed the incentives faced by producers and how changes in these incentives are related to reforms, a next step is to investigate how producers have responded. Different sets of producers will be more or less able to respond to identical changes in output prices depending on the current policy and institutional environment and on agro-climatic constraints. Establishing the supply response is therefore a critical link in explaining the impact of reforms on income levels (see below).
Good quality survey data is essential for measuring the response. Even where the quantification of elasticities using econometrics is not feasible, insights should be gained through other techniques. Interviews with farmers, key informants and extension workers often reveal whether producers responded to incentives following a policy change, whether the response was weak or strong, and what non-price factors were important for facilitating the response. Where there was no response, these surveys should point to constraints limiting the response. One technique that has been used to study supply response to prices and other constraints at the farm level is to use linear programming methods. This would provide a what if type of analysis, rather than quantifying past responses, but could be useful for insights about how farmers would have responded.
Output response by source. Data on aggregate commodity production should be analysed to determine the proportion of changes in the level of production that can be attributed to changes in the harvested area and the proportion attributed to changes in yield. The sources of agricultural production growth are likely to vary across commodities and across countries as a result of the different incentives, opportunities and constraints faced, as Figures 17.2 and 17.3 demonstrate. Rice sector expansion in Mali during the past two decades appears to be driven mainly be area expansion, although there is some evidence of yield increases. By contrast, the dominant driver in Viet Nam has been yield increase with more limited area expansion. Questions need to be asked as to why the different responses occurred and whether they were the most desirable.
Figure 17.2 Sources of rice sector expansion, Mali, 1980-1998
Source: FAOSTAT (2000)
Figure 17.3 Sources of rice sector expansion, Viet Nam, 1980-1998
Source: FAOSTAT (2000)
Examples of analyses that may be required to understand the relationships are provided in Box 17.1.
The insights derived from the analysis of prices and production levels could be combined to inform analysis of the impact of episodes of reform on agricultural supply response.
Box 17.1 Investigations for relating output response and the modifying parameters An investigation of access to tradable inputs may include analysis of:
Land tenure systems may be characterized by:
Formal and informal mechanisms for credit may be characterized by:
Regional differences may be characterized by
|
Step 3 - Impact of agricultural change on household incomes and expenditure
The second stage of the link depicted in Figure 17.1 can be broken down further to aid analysis:
the relationship between agriculture sector change and household food security
the relationship between agricultural sector performance and national indicators of food security
The household level relationship can be considered in two stages: first, the impact of agriculture sector change on household income and expenditures; and second, the relationship between changes in household incomes and expenditures and household food security. In step 3, the first stage is examined, with the other components being investigated in the next step of the analysis.
Reform implications by category of rural household
To understand how households have been differentially affected by agriculture sector change brought about by reform, it is necessary first to categorize household types. Indeed, the proper identification of the household typology is often crucial. The most appropriate classification will vary between countries, but ideally would emerge in part, from the market segmentation of step 2. For example, if the market for key crops has been segmented regionally, household types could be classified within each of the regions.
The following mechanisms for farm household classification may be appropriate:
by degree of commercialization (subsistence, pre-commercial, commercial);
by types of main products (export crops, import crops, non-tradables);
by organization of farming (owner-operated, cooperative, contract farming, sharecropping, plantation);
by mode of market access (e.g. how output and credit markets are accessed);
by location[289]
by scale of resources access (particularly land);
other non-farm rural household types (e.g. rural labourers).
For each selected household type the information should be documented on income patterns (percent of total income by source) and on expenditure patterns (proportion on food and non-food items). For farm households, data should be documented on the production of each commodity and on the proportion of each that is sold. Ideally, the information should be collected for two or more discrete periods of time in order for changes to be identified.
A commonly used approach is to construct an accounting framework to record all incomes and expenditures of a household in order to compute the net change, and could also incorporate quantity responses to price changes. Applications based on this approach to quantifying the impact of reform on poverty are numerous, as reported in McCulloch (2002). Sahn and Sarris suggest that there are two issues that require careful analysis when adopting this approach. One relates to the structure of household income and consumption and how this might change during the reform period, and the other is the effect of price changes on the magnitudes of these values.
The following framework is somewhat simplified, but adequate for most cases.
For a selected household type, and a given period, the full income and consumption expenditures are defined as in equations (17.6) and (17.7) after McCulloch (2002).
Income: Y = (S pojqoj - S pikqik) + w.L + T (17.6)
where the first term within the bracket is value added (income) from own production (the ps indicating prices and qs indicating quantities, the superscript o output and subscript j commodities produced). The w.L term is to measure labour income (wage rate times employment) and T is transfers received by the household (e.g. from government, pension, remittance etc.).
Consumption expenditure: C = (S pcjqcj) (17.7)
with variables as above (the superscript c standing for consumption).
The purpose is to measure change over two periods. Where survey data are available for two periods, the Y and C in equations (17.6) and (17.7) are measured for two periods. Where this is not possible, analysts have measured the impact of price changes (induced by policy reforms, as estimated in the previous steps, or shocks) assuming that quantities do not change.
Using the change operator D, equations (17.8) and (17.9) are used for this purpose:
DY = (SD pojqoj -SD pikqik) + D w.L + D T (17.8)
Similarly, the change in consumption assuming that consumed quantities remain fixed is:
DC = (SDpcjqcj) (17.9)
A first-order approximation of the change in money metric utility, or welfare W[290], resulting from changes in prices (including wage rates) is given by expression (17.10):
DW = DY - DC (17.10)
With this framework, it is possible to analyse the impact of changes in many variables, including output prices, food prices, input prices, wage rates, transfers etc. In particular, the focus of many studies has been on the impact of a change in food prices. This is because an increase in food prices not only leads to higher income but also higher consumption expenditure. The net gain will depend on the net position of the household (net seller or net buyer) and the weights (income and consumption shares).
Many studies also embed quantity responses, i.e. price and income elasticities of supply and demand, in equations (17.6)-(17.10) in order to derive expressions that permit the analysis of multiplier effects[291]. However, analysts need to weigh up the benefits and costs (in terms of data and expertise required) of attempting to estimate these elasticities. Consumption elasticities are likely to be of less interest than production elasticities. The model can also be used for ex ante simulation studies, under a variety of assumptions about markets, e.g. complete and fully functioning markets, missing markets, or imperfect markets.
Data sources
Data availability is likely to impose constraints on the options available for collating and examining these changes. In most situations, household survey data could be used as the basis for household categorization and the analysis of changes in expenditure and income levels. A particular constraint is likely to be data availability on rural labour markets, notably wages, and this may preclude analysis of the impact on household income levels as a result of labour reallocation. Where the required information is not available from existing surveys, rapid rural appraisal (RRA) approaches may be undertaken to provide an approximation of the values associated with the key variables.
In some cases, only one survey may be available during the period under analysis. In such cases, these surveys will provide the quantity information required in equations 17.6 and 17.7 above, and the analysis will reveal the impact of changing prices and wage rates under the assumption that the structure of the households income and consumption patterns remain constant. In other words the household income and consumption levels are projected forward from the initial situation.
Where two surveys are available it is possible to incorporate quantity changes into the analysis. In this case, it would also be feasible to undertake the projection as explained above, but with the added advantage of being able to cross check the results against the situation described by the second survey.
Whether one or more surveys are available, it is important to construct a counterfactual. This can be achieved by determining the expected income, given past trends in income and productivity levels, in the absence of reforms. It would then be possible to judge the impact of reform by determining the difference between the counterfactual and the observed level of income. Any divergence could then be related to the policy reforms that had been implemented during the period.
In some circumstances, existing modelling exercises which analyse the impact of policy reform may exist, and these could also be used to cross check or verify these micro level analyses.
Once the sources of incomes and the patterns of expenditure have been established, attempts should be made to relate these to reforms of the agriculture sector.
Step 4 - Explaining changes in food security
The information derived from the analyses above can now be used to infer relationships between agriculture sector performance and both national and household food security.
National food security
FAO has been publishing, on an annual basis since 1999, estimates of the number of undernourished population for some 125 developing countries and economies in transition[292]. By using national-level data on the availability of food calories per capita, requirements and their distribution across population groups, estimates are made of the percentage of people in each country whose average calorie intake falls below the minimum required for living and light activity. Although national-level data are used to derive these estimates, the methodology yields the prevalence of undernourishment in terms of individuals in a country.
Most countries in the world have some food insecure. Thus, one can not say that a country is food secure or insecure, but countries can be ranked. For example, one may categorize countries on the basis of the severity of food insecurity, extremely low food insecurity when the prevalence rate is 2.5 percent or less and very high food insecurity when the rate is over 35 percent.
There is tendency to use some food-related indicators to infer the state of national food security, notable ones being per caput food production and self-sufficiency ratios (SSRs). A fall in the value of these indicators is then said to indicate a deterioration in national food security. Although one may find some correlations between these indicators and the household/individual based indicators of food security, the inference could be misleading. There are many instances of countries where these indicators worsen but the country becomes more food secure. In the context of economic reforms and trade liberalization, what happens to the food sector following a reform episode will depend on the situation prior to the reform. For example, if food sectors are protected with tariffs while cash crops are taxed, a trade reform (lowering of tariffs on foods and export taxes on cash crops) should lead to an expansion of the cash crop sector while the food sector will shrink, resulting in a deterioration in food security situation based on the two food-based indicators. If overall agricultural incomes rise as a result of this result (notably if the overall income gains are also captured by food insecure households), the country should be importing more food than before the reform. If so, per caput consumption or total supply of food would be higher. The new outcome should also be reflected in terms of the ratio of food imports to agricultural exports, which should fall or should not increase.
It is also possible that trade reform may even lead to a deterioration in agriculture as a whole (both food and non-food sectors) if the sector was supported, to the extent that it is no longer profitable to produce the same volume at world market prices. Additionally, economic growth has historically been associated with declines in the share of agriculture in overall GDP, while industries and service sectors expand. In that case, all three indicators (per caput food production, SSR and ratio of food imports to agricultural exports) may deteriorate.
Even in these cases, there is no guarantee that household food security has improved, since it is possible that increased food imports go to feed livestock for the rich, for example, maize and soy products.
At the national level, broad indicators of food security should therefore be related to broad indicators of agricultural sector change collected for step 1 with caution. The main objective at this stage is to analyse the effect of the modifying parameters on this relationship. These relate to the structural diversity of economies and of the role of the agriculture sector within them. Examples are provided in Box 17.2.
Box 17.2 Characterizing structural diversity The importance of agriculture may be characterized by indicators such as:
The degree of trade dependence may be characterized by:
Poverty levels may be characterized by
|
The country context
In addition, the context within which the reforms are taking place will also affect their impact on food security either directly or indirectly. In making comparisons across countries, context will be especially important. Indicators describing the country context are suggested in Table 17.3
Table 17.3 Indicators of country-level context
CATEGORY |
DESCRIPTION OF INDICATOR |
INDICATOR |
Demographic |
Population |
Total (P) Density (P) Growth rate (P) |
Rural versus urban |
Number in rural areas (P) Number in urban areas (P) Ratio of population in rural to urban areas (D) |
|
Age profile |
Dependency ratio (P) |
|
Education level |
Literacy rate (P) |
|
Cultural diversity |
Ethnic group ratios (P) |
|
Macroeconomic |
GDP |
GDP (T) GDP growth rate (D) Per capita income (D) |
Sectoral contribution |
Share of agriculture, industry, services (T) |
|
Unemployment |
Proportion of workforce unemployed (P) |
|
Employment by sector |
Numbers of employed by sector (T) Proportion of workforce by sector (D) |
|
Economic stability |
Inflation (T) Consumer price index (T) Low-income price index (T) |
|
Resource endowments |
Land area |
Total Total agricultural (P) Total arable (P) Per capita arable (D) |
Suitability/potential |
Proportion potential agricultural utilised (P) Proportion potential arable utilised (P) |
|
Climatice.g. drought risk |
Rainfall (Tm) Evapotranspiration (Tm) |
|
Irrigation |
Proportion of arable land irrigated (P) Irrigable area (potential) (P) Irrigation type (e.g. surface, groundwater, harvesting) Irrigation recharge (e.g. snowmelt versus. rainfall) |
|
Forest resource |
Forest coverage (P) |
|
Water availability |
Use to recharge (P) |
|
Non-economic events |
Civil war, mineral discovery, political change, famines, floods etc |
Timeline to record instances |
The following symbols are used to denote the minimum periodicity of data required for the subsequent analysis:
(P) = at discrete intervals i.e. 5 yearly
(T) = time series annual unless specified by e.g. (Tm) = monthly
(D) = derived from collected data
Data should be collected over the period from 1970 to the present as a minimum.
Household-level food security
In the previous stage of the analysis, agriculture sector change was related to changes in the levels and sources of incomes and expenditures in different categories of households.
At this stage of the analysis attempts should be made to relate the information on changes in incomes and expenditures to indicators of food security. Following Sens entitlement approach, analysts might characterize the relationship in terms of:
the implications of changes in income and expenditure levels for purchasing power;
changes in subsistence strategies;
changes in average calorie consumption[293].
The objective is to identify whether positive shifts in income and expenditure patterns that have been attributed to changes in the agriculture sector are associated with improvements in food security status and in doing so, to identify groups of households that have lost or gained from these changes.
Indicators that may be used to inform this identification may include:
relative proportion of own production, cash, exchange for labour, safety net, remittance in incomes;
cost of food relative to wages and other income;
share of total expenditure on food;
stability of food basket costs.
Although, it is probably not possible to forge a direct link between a specific reform and the status of food security, the research strategy proposed above should improve the existing understanding relating to the factors which can modify the strength of any potential change brought about by reform. Understanding the effect of these parameters is likely to be more important than the ability to predict the exact response to reform, as it is an understanding of these parameters that will guide appropriate policy reform and the design of required complementary policy and institutional reforms.
[281] Prepared with inputs
from Ramesh Sharma and Jamie Morrison. [282] WTOs Trade Policy Reviews are a good source for identifying the range of trade policy instruments used in the past and currently. [283] Household level food security indicators are derived in steps 3 and 4. [284] Sharma, R. 2003. An Overview of the FAO Studies on International Trade and Household Food Security and Methodologies. INFOSAMAK Expert Consultation on International Fish Trade and Food Security. Casablanca 27 - 30 January [285] This expression for domestic price determination applies to trade products. For non-trade products, domestic prices are determined by forces of domestic demand and supply. [286] Quiroz,.J & Valdés, A. 1993. Decompositions based on real price and exchange rate data. [287] In addition, this technique has been found to suffer from several problems, including the non-stationarity of the data and high degrees of positive autocorrelation. Most economic time series are non-stationary in their level forms, which means that they will tend to drift in a random way following a shock. So, the relationship obtained from the static regression analysis could be misleading. [288] According to Granger (1969), since the future cannot predict the past, if variable x (Granger) causes variable y, then changes in x should precede changes in y. Hence, in a regression of y on other variables (including its own past values) if past or lagged values of x are included and it significantly improves the prediction of y, then we can say that x (Granger) causes y. In contexts where price transmission is occurring, but not instantaneously, Granger causality would be expected. A test for Granger causality therefore provides an indication as to whether, and in which direction, price transmission is occurring between series. [289] McCulloch (2002) notes that remoteness can determine the extent to which farm households are engaged in market related activities as opposed to production for home consumption. [290] Most poverty-focussed studies measure welfare in income terms. In food security-oriented studies, there is a justification for extending the analysis to quantify the impact in terms of food consumption, e.g. by measuring food energy (kilocalories) and protein. Several studies on agricultural commercialization in von Braun and Kennedy (1994) extend the analysis from incomes/expenditures to food energy. [291] There are numerous references on agricultural household models and their applications to cite here. A good exposition of the theory and application is in Chapter 6 of Sadoulet and de Janvry (1995). Sahn and Sarris (1991) apply similar method in their study of five countries in Africa (they trace welfare changes for 15 years, 1975-89). The Ethiopian study of Dercon (2001) uses similar technique. Minot and Goletti (2000) apply the method in their study of the impact on poverty of Vietnams rice market liberalization. [292] FAO, The State of Food Insecurity in the World available on-line at www.fao.org. [293] It is not expected that anthropometric data will be used in the analysis. |