Nkechi S. Owoo and Wim Naudé Are Informal Household Enterprises also subject to Agglomeration Economies? Evidence from Rural Africa AG RICULTURE IN AFRICA T E L L I N G FA C T S FROM MYTHS Global Development Network (GDN) Conference June 18-20, 2014 INTRODUCTION • Non-farm enterprises ubiquitous in rural Africa o Brewing, retail, running a restaurant or coffee-shop, running a taxi, etc – 42% of rural households operate non-farm enterprises (Nagler and Naudé, 2014) – 40-50% of rural household income in Africa from rural non-farm enterprises (Rijkers and Costa, 2012; Haggblade et al., 2010) • Non-farm economies increasingly vital for job creation and livelihoods (De Brauw et al, 2013; Javry and Sadoulet, 2010) – Growth in rural populations – Declines in agricultural employment Page 2 LITERATURE REVIEW • Most literature on enterprise productivity deals with advanced economies – Productivity levels widely dispersed across firms o Managerial competence- Mano et al. (2012); Bloom and Van Reenen (2010) o Innovation and absorption of technology- Bernard (2010) o External shocks- Rijkers and Soderbom (2013) • Fewer studies on developing countries – Aspects of African business environment hindering firm growth o Market access, poor infrastructure, weak governance, financial services, etc Page 3 LITERATURE REVIEW • Productivity of firms depends on productivity of other firms in close proximity – Enterprises clustering together is advantageous for individual productivity • Again, most studies of spatial clustering of firms examined in advanced economies.... – Wennberg and Lindqvist (2010)- Sweden – Rupasingha and Contreras (2010)- rural USA – Baumgartner et al (2012)- rural Switzerland – Martin et al. (2011)- France • Spatial Effects DO matter! - Deller, 2010 Page 4 LITERATURE REVIEW • Fewer spatial studies of rural nonfarm enterprises in developing countries – spatial proximity important for firm performance o McCormick (1999); Siba et al (2012) • Ali and Peerlings (2011) & Ayele et al. (2009) – Clustering helps enterprises in handloom industry in Ethiopia to improve productivity • No explicit spatial techniques applied – Spatial nature of data biased estimates Page 5 RESEARCH QUESTION (MYTH OR FACT??) • Spatial effects matter for RNFEs in developing country settings • There are positive linkages between farming and non-farm enterprises in developing countries Page 6 DATA • 2011 Ethiopian Rural Socioeconomic Survey (ERSS) – 259 EA observations • 2010/11 Nigeria General Household Survey (NGHS) – 379 EA observations – Information on primarily rural areas • Basic demographic information – Education, health , labour, non-farm economic activities • GIS information – Analysis at EA level Page 7 STUDY VARIABLES: (BASED ON THEORETICAL LITERATURE) • Dependent Variable – Sales of RNFEs • Household-head Characteristics – Age – Sex – Marital status – Education – Religion – Household size • Location and Infrastructure Characteristics – Co-operative – Phone – Microfinance Institution – Distance to asphalt road – Distance to market (see paper Table 1A & 1B for summary statistics) Page 8 EMPIRICAL METHODOLOGY • Exploratory Spatial Data Analysis (ESDA) – Series of tests that account for spatial nature of data o Quantile Maps o Global Moran’s I Statistics o Local Indicators of Spatial Autocorrelation • Econometric Specification – Multivariate regression of RNFE performance on set of control variables o OLS o Spatial Lag o Spatial error Page 9 Exploratory Spatial Data Analysis (ESDA) Distribution of RNFE Performance in Ethiopia and Nigeria 10 GLOBAL MORAN’S I STATISTICS ETHIOPIA NIGERIA Page 11 LOCAL SPATIAL AUTOCORRELATION ETHIOPIA NIGERIA Page 12 ECONOMETRIC SPECIFICATION • OLS (BASE MODEL)• SPATIAL LAG MODEL• SPATIAL ERROR MODEL- ; where; Y is the dependent variable, X is the vector of household and community independent variables, β is the vector of regression co-efficients 𝜖 is the vector of errors p is the spatial lag co-efficient WY is the spatially lagged dependent variable W is the weight matrix 𝝺 is the spatial error co-efficient 𝝻 is the vector of errors Page 13 Empirical Results: EA/ Individual Level • Regression of RNFE performance on household and community variables – Other control variables omitted; have expected signs (see paper Tables 3A and 3B) SPATIAL PARAMETERS ETHIOPIA NIGERIA EA Individual EA Individual Rho (p) 0.527* (1.85) 0.572*** (21.77) 0.240 (0.74) 0.133*** (4.32) Lambda (𝝺) 0.310 (0.73) 0.582*** (21.91) 0.0263 (0.07) 0.136*** (4.39) Control Vars YES YES YES YES # Obs 259 1, 230 379 2, 001 t statistics in parentheses : * p < 0.10, ** p < 0.05, *** p < 0.01 Page 14 Conclusions from Empirical Estimations • Evidence of spatial correlation – EA vs. Individual level analyses • Education, religious affiliation and marital status of household head are important determinants of RNFE performance in Ethiopia • Age and sex of head, education and presence of microfinance institutions are important determinants of RNFE performance in Nigeria Page 15 Bivariate Relationship between RNFE Performance and Agricultural Activity • Spatial interactions between concentration of agricultural activities and RNFE performance – Are high performance RNFEs clustered, not to be near one another, but to be near high prevalence farming areas? • Strong linkages between farm and non-farm activity – Negative Relationship o De Janvry, 2005; Lanjouw and Lanjouw, 2001 – Positive Relationship o Haggblade et al. 2002; Deichmann et al. 2009 Page 16 Bivariate Relationship between RNFE Performance and Agricultural Activity Global Moran’s I Ethiopia Nigeria -0.0225082** (0.049) -0.0400949*** (0.002) • Negative spatial relationship between farm activity and RNFE performance – High (low)-performing non-farm enterprises are surrounded by other communities with low (high) engagement in farming activities Page 17 Bivariate Relationship between RNFE Performance and Agricultural Activity • Increases in farm activity not necessarily associated with increases in non-farm enterprise productivity in the same region • Contrary to ‘most prominent view amongst development practitioners’ (Deichmann et al., 2008: 1) • Requires more research – Type of rural non-farm enterprise? – Some other unexplained characteristics? Page 18 RESEARCH LIMITATIONS • Small sample size – only 259/ 379 observations • Scale of spatial analysis Page 19 Myth OR Fact? Summary FACT - Evidence of spatial effects in developing countries MYTHISH- There are positive linkages between farming and non-farm enterprises in developing countries RNFE performance highest in areas with lower farming activity Additional research required Page 20 POLICY IMPLICATIONS • Spatially differentiated approach to RNFE support • Encourage asset and knowledge accumulation of existing firms – Improve skills and technology of leading enterprises o Spillover to proximate enterprises • Encouragement of entrepreneurial and management education for enterprise performance • Need for investments in local infrastructure Page 21 THANK YOU FOR YOUR ATTENTION! Page 22