Methodology

Abstract

This chapter introduces the book’s empirical investigation of the role that agricultural growth has played in aggregate economic growth in Ethiopia in the twenty-first century. This chapter describes the main method and data used for empirical investigation. It describes both the data contained in the three Social Accounting Matrices (SAMs) that form the book’s main empirical data source, as well as the methodological proceedings of the Semi-Input-Output multiplier model that the book applies.

To explore the role of agriculture in Ethiopia’s economic growth, this section quantitatively evaluates the country’s changing agricultural growth linkages. It uses a Social Accounting Matrix (SAM)-based economic multiplier model to calculate the agricultural sector’s growth linkages and compare them to the manufacturing sector. Along with econometric studies (Bravo-Ortega & Lederman, 2005; Tiffin & Irz, 2006; Self & Grabowski, 2007) and growth accounting studies (Martin & Mitra, 2001; Gulati et al., 2005; Bosworth & Collins, 2008), multiplier analysis is one of the main methods for assessing the role of agriculture in economic growth. The choice to use this method in the present study is made in light of the recognized limitations of growth accounting, which provides only a decomposition of the proximate sources of growth, and the difficulty of establishing convincing causal links through econometric studies (as discussed by Tsakok & Gardner, 2007).

The economic multiplier method allows us to address two specific questions: how large are the growth linkages of agriculture and manufacturing, and has their size changed over time relative to each other? Through these questions, the method can explore which sector was the best growth option at three snapshots during Ethiopia’s economic growth in 2002, 2006, and 2010. This approach also allows us to get closer to the counterfactual question of which sector was the strongest growth engine at these points and therefore would have been the most appropriate sector to invest in. Given the historical record, we know that both the agricultural sector and the economy grew over the period of investigation. However, the economic multiplier method allows for a greater understanding of whether Ethiopia’s agricultural growth was the main engine of its economic growth or whether the same (or higher) aggregate growth would have been generated had there been similar growth in the manufacturing sector instead. The method also tries to uncover whether the size of the growth linkages of agriculture and manufacturing have changed over time. Although 2002–2010 is a short time frame to observe significant economic change, it was a transformative period in the Ethiopian economy, representing a break toward higher economic growth than in previous periods. If successful industrialization was occurring in this period of economic growth—as theory and previous development experiences would suggest—then the multiplier analysis would show that the growth linkages from the manufacturing sector strengthened over time, while the agricultural growth linkages would shrink.

Data

To calculate growth linkages in an economy, information on the economy’s production technologies and consumption patterns is needed. This information can be obtained via the construction of a SAM. A SAM is a summary table for a given period (often a year) that provides a coherent, detailed database on the production, incomes, consumption, investment, external trade, and other flows in the economy, revealing a country’s economic structure. There is a large literature on the construction, usefulness, and application of SAMs (Pyatt & Round, 1979; Defourny & Thorbecke, 1984; Keuning & de Ruijter, 1988; Sadoulet & de Janvry, 1995; Thorbecke, 2000; Round, 2003; Breisinger et al., 2009). Several studies have employed SAM-based multiplier analysis to examine how an income injection in one part of an economic system affects the economy (Hayden & Round, 1982; Bell & Hazell, 1980; Thorbecke et al., 1992; Powell & Round, 2000; Tarp et al., 2002). SAMs are usually studied for one year at a time due to the extensive work and data collection that is required to construct them. While the use of multiple SAMs is therefore somewhat uncommon, previous studies using SAMs for multiple years include Cohen (1989), Hewings et al. (1998), Lima et al. (2004), and Llop (2007); other studies have used SAMs for multiple countries, including Vogel (1994) and Arndt et al. (2012). The multiple SAMs available for Ethiopia during this period offer a unique opportunity to provide a detailed understanding of the relationship between agricultural and economic growth in a low-income country.

A SAM is an accounting framework represented as a square matrix in which each cell represents a flow of funds from a column account to a row account. A basic SAM structures the economy into seven types of accounts: activities, commodities, factors of production, households, government, savings and investment (S-I), and rest of the world (RoW) (Breisinger et al., 2009). In this structure, “activities” are the entities that produce goods and services, and “commodities” are the goods and services that activities produce. In most SAMs, including the SAM used here, the values in the activity accounts are measured in producer prices. Therefore, the SAM structure used also includes an eighth account for “margins,” which include the marketing and transportation costs associated with commodity flows.

Three national Ethiopian SAMs are available from previous research, for 2002 (EDRI, 2008), 2006 (EDRI, 2009), and 2010 (Aragie, 2014).¹ As two of the previous SAMs were constructed by the Ethiopian Development Research Institute (EDRI), and the third is an updated version of the 2006 SAM, the SAMs share a similar structure and are suitable for comparison. The rich and detailed data contained in the SAMs are sourced from both macro and micro sources. The main data sources for the 2002 SAM are national accounts statistics (MOFED, 2006), supply-use tables (MOFED, 2007), industry surveys (CSA, 2003a, 2003b, 2003c), agricultural census data (CACC, 2003), labor force survey data (CSA, 2006a), balance-of-payments statistics (CSA, 2004), and household survey data (CSA, 2001), as well as internal government revenue, government expenditure, and customs data files (EDRI, 2008). These were constructed as a square matrix of 133 × 133 cells in Excel containing information about 42 production activities, 61 commodity groups, five primary factors, two household groups, and 17 tax instruments, as well as aggregate accounts for trade margins, transport margins, government, investment, and RoW. The 2006 SAM is an extended version of the 2002 SAM that uses similar data sources, but is updated if available (EDRI, 2009). Updates include industry survey data (CSA, 2006b, 2006c, 2006d), agricultural census data (CACC, 2007a, 2007b), balance-of-payments statistics (CSA, 2006e), and household survey data (CSA, 2006f). The SAM is a square matrix of 255 × 255 cells containing information on 99 production activities, 91 commodity groups, five factors of production, 14 household groups, 17 tax instruments, and aggregate accounts for trade margins, transport margins, government, investment, and RoW. The 2010 SAM is an updated version of the 2006 SAM constructed by Aragie (2014). It updates the 2006 SAM with a new Household Consumption and Expenditure survey for 2011 and national accounts for 2011 and includes finer disaggregation between home production and marketed production. It is a square matrix of 236 × 236 cells containing information on 50 production activities, 39 commodity groups, 36 primary factors, 27 household groups, five tax instruments, and aggregate accounts for trade margins, transport margins, government, investment, and RoW (Aragie, 2014).

To make the three SAMs comparable, they were recoded into the same structure of eight activities, eight commodity groups, two factors of production, two household types, and aggregate accounts for margins, government, S-I, and RoW, as summarized in Table 6.1.² As such, the research is based on three SAMs, each structured into a 24 × 24 square matrix. The high level of aggregation into eight activities and commodities is implemented in order for the three SAMs to be comparable over time.

Table 6.1 Structure (accounts) of the 2002, 2006, and 2010 SAMs

Method

To capture the size and change of economic linkages in Ethiopia, the study calculates Semi-Input-Output (SIO) multipliers based on the three SAMs. SIO multiplier analysis is an economic model that assumes that all relationships in each SAM are linear and prices are fixed (in the short run). The work follows the guidelines for SIO multiplier analysis as outlined in Breisinger et al. (2009).

SIO multiplier analysis indicates the size of a sector’s contribution to aggregate growth through its linkages with other sectors. It shows how much the overall economy would grow if one sector grows. This is estimated by calculating how much the overall economy would grow if one sector experienced an exogenous demand-side shock—for example, due to increased export demand, public spending, or aid—considering both the direct and indirect effects. The indirect effects are also called “demand linkages” and include backward and forward production linkages as well as consumption linkages. Together, the direct and indirect effects make up the total multiplier effect. The total size of a sector’s growth linkages depends on the interdependencies among an economy’s sectors in terms of production technologies and household consumption patterns.

To calculate the multiplier effect, the SAM accounts are divided into endogenous and exogenous accounts such that a change in the latter influences the former. In the model, the government, capital (S-I), and RoW accounts are treated as exogenous. As such, the model only considers two sets of agents (activities and households) interacting through two sets of markets (commodities and factors). The model also requires the classification of sectors into those with perfectly elastic supply and those that are supply-constrained. If demand for a supply-unconstrained sector increases, domestic output increases to match the increased demand. However, if demand for a supply-constrained sector increases, it is satisfied by imports. The degree of supply responsiveness in a sector depends on technological and resource constraints and the capacity to utilize available technologies and resources. In this study, two scenarios are modeled: one where agriculture is supply-unconstrained and another where agriculture is supply-constrained.³ In both scenarios, the public service sector is treated as supply-constrained, which is consistent with the literature. All other sectors are treated as supply-unconstrained.⁴

The SIO multiplier analysis is used to simulate the insertion of equivalent investments leading to equal-sized demand increases in either the agricultural sector or the manufacturing sector for each of the three years. The exogenous injection represents increased demand from any of the exogenous accounts, such as increased public spending, increased investment demand from either domestic or international capital, or foreign aid.⁵ The resulting multiplier effect reveals how much the economy would grow—given its production technologies and consumption patterns in the year described by the SAM—if one sector experienced such an exogenous demand-side shock. It does not attempt to explain why such a demand-side shock would occur, nor why some sectors respond more or less than others. Instead, it gives a numerical description of how an exogenous inflow into one sector would affect the other sectors in the economy once the structural (demand and supply) interconnections are fully taken into account. To calculate the total multiplier effects, the SIO model uses matrix algebra. The equations are detailed below, and the equation legend is provided in Table 6.2. The equations are developed based on the guidelines in Breisinger et al. (2009).

Table 6.2 Equation legend, values and shares

$$ {\displaystyle \begin{array}{l}{Z}_1={a}_{11}{X}_1+{a}_{12}{X}_2+{c}_1Y+{E}_1\\ {}{Z}_2={a}_{21}{X}_1+{a}_{22}{X}_2+{c}_2Y+{E}_2\end{array}} $$

(1)

$$ {\displaystyle \begin{array}{l}{Z}_1={a}_{11}{b}_1{Z}_1+{a}_{12}{b}_2{Z}_2+{c}_1\left({v}_1{b}_1{Z}_1+{v}_2{b}_2{Z}_2\right)+{E}_1\\ {}{Z}_2={a}_{21}{b}_1{Z}_1+{a}_{22}{b}_2{Z}_2+{c}_2\left({v}_1{b}_1{Z}_1+{v}_2{b}_2{Z}_2\right)+{E}_2\end{array}} $$

(2)

$$ {\displaystyle \begin{array}{l}{Z}_1-{a}_{11}{b}_1{Z}_1-{c}_1{v}_1{b}_1{Z}_1-{a}_{12}{b}_2{Z}_2-{c}_1{v}_2{b}_2{Z}_2={E}_1\\ {}-{a}_{21}{b}_1{Z}_1-{c}_2{v}_1{b}_1{Z}_1+{Z}_2-{a}_{22}{b}_2{Z}_2-{c}_2{v}_2{b}_2{Z}_2={E}_2\end{array}} $$

(2)

$$ {\displaystyle \begin{array}{l}\left(1-{a}_{11}{b}_1-{c}_1{v}_1{b}_1\right){Z}_1+\left(-{a}_{12}{b}_2-{c}_1{v}_2{b}_2\right){Z}_2={E}_1\\ {}\left(-{a}_{21}{b}_1-{c}_2{v}_1{b}_1\right){Z}_1+\left(1-{a}_{22}{b}_2-{c}_2{v}_2{b}_2\right){Z}_2={E}_2\end{array}} $$

(3)

$$ \left(\begin{array}{cc}1-{a}_{11}{b}_1-{c}_1{v}_1{b}_1& 0\\ {}-{a}_{21}{b}_1-{c}_2{v}_1{b}_1& -1\end{array}\right)\left(\begin{array}{c}{Z}_1\\ {}{E}_2\end{array}\right)=\left(\begin{array}{cc}1& {a}_{12}{b}_2+{c}_1{v}_2{b}_2\\ {}0& -1+{a}_{22}{b}_2+{c}_2{v}_2{b}_2\end{array}\right)\left(\begin{array}{c}{E}_1\\ {}{Z}_2\end{array}\right) $$

(4)

$$ \left(\begin{array}{cc}1-{a}_{11}{b}_1-{c}_1{v}_1{b}_1& 0\\ {}-{a}_{21}{b}_1-{c}_2{v}_1{b}_1& -1\end{array}\right)=I-{M}^{\ast } $$

(4.1)

$$ \left(\begin{array}{cc}1& {a}_{12}{b}_2+{c}_1{v}_2{b}_2\\ {}0& -1+{a}_{22}{b}_2+{c}_2{v}_2{b}_2\end{array}\right)=B $$

(4.2)

$$ \left(I-{M}^{\ast}\right)\left(\begin{array}{c}{Z}_1\\ {}{E}_2\end{array}\right)=B\left(\begin{array}{c}{E}_1\\ {}{Z}_2\end{array}\right) $$

(4.2)

$$ \left(\begin{array}{c}{Z}_1\\ {}{E}_2\end{array}\right)={\left(I-{M}^{\ast}\right)}^{-1}B\left(\begin{array}{c}{E}_1\\ {}{Z}_2\end{array}\right) $$

(5)

The final SIO multiplier equation (5) shows that an exogenous increase in demand for the unconstrained sectors (E₁) leads to a final increase in total demand for these sectors (Z₁), including all of the forward and backward linkages ((I−M*)⁻¹). However, the exogenous demand for constrained sectors (E₂) is leaked to imports (because the final demand for the supply-constrained sectors (Z₂) is met through increased imports), eliminating any linkages for those sectors. The information regarding linkage effects from the SAM is incorporated into the multiplier model through the coefficient matrix M.

The above equations calculate accounting multipliers based on average patterns, not “fixed-price” multipliers based on marginal responses (despite the name, note that both multipliers are formally “fixed-price,” as prices are fixed in the short run). While “fixed price” multipliers may be conceptually closer to the underlying reality, as the marginal responses in the system may be different from the average one, Pyatt and Round (1979) compared computations for both types, and found that numerical differences were very small. As such, accounting multipliers are deemed sufficient for this study.

The SIO model suffers from several shortcomings. These include its assumptions regarding fixed prices and its presentation of the results as if the economic system adjusted immediately to exogenous changes without addressing the institutional barriers that can prevent such adjustment. In addition, while using three SAMs, the SAM methodology is not optimal for the study of structural change, as the model assumes that the structure is fixed in each year (although changing across the SAMs). As such, this study cannot speak to the relationship of agricultural growth to structural change but is limited to exploring the power of agriculture (and manufacturing) to generate growth. In combination with the high level of aggregation of the sectors—specified to make the SAMs comparable across time—this implies that the model specified here cannot plausibly claim to capture the full complexity of the connection between agriculture and the aggregate economy in Ethiopia. As a result, the study is primarily a tool for measuring the broad changes in economic multipliers in Ethiopia over time rather than a tool for detailed policy analysis. Given that the SIO analysis does not include the institutional barriers to linkages, the results are best understood as the upper bounds of economic linkages. The interpretation of the results should therefore focus on comparisons and patterns of change rather than on exact multiplier sizes.