The Effects of Agro-clusters on Rural Poverty: A Spatial Perspective for West Java of Indonesia Dadan Wardhana, Rico Ihle, Wim Heijman (Agricultural Economics and Rural Policy) Prepared to be presented in EAAE 150 th Seminar, Edinburgh, UK, October 22-23,
Presentation Outline: Introduction Theory Data Methodology ● Cluster measurements ● Model Specifications Results and Discussion Policy Remarks 2
Introduction Role of the agricultural sector in rural economies Concentration of non-farm economy vs. concentrated agricultural sector Shifting from agriculture to more developed non-farm to boost economies However, Agricultural productivity growth associated with higher economic performance and lower poverty rates (de Janvry & Sadoulet, 2010) More investment in agricultural sector increases welfare (World Bank, 2008) GDP growth in agriculture has at least three times larger positive impact on poverty reduction than for the rest of the economy How? 3
What are agro-clusters? Economic Agglomeration Collective efficiency (Schmitz & Nadvi, 1999) Competitive advantages (Porter, 1990) Pool of labourers, information & knowledge exchange (Krugman, 1995) Farmer Group Manufacturing Tourism Restaurants Exports Source: Authors 4
Theory of Spatial Agglomeration Effects Productivity curve Cost curve Number of labourers (Cluster size) Productivity per hectare Costs Profits Number of labourers (Cluster size) Profit per hectare “Spatial concentration in agriculture reduces rural poverty” (Kiminami & Kiminami, 2009) Source: Duraton, et al. (2010). p.34 Productivity Curve The higher size of clusters, the higher productivity. Cost Curve The higher size of clusters, the higher production costs Profit Curve concave profit vs. size of the cluster 5
Description of West Java B J Note: B : Bandung Metropolitan, one of cities in West Java J : Jakarta Metropolitan, a capital city of Indonesia Source: Authors 6
Data The Indonesia programme for the census of agriculture of the office of national statistics, The Indonesia programme for the census of poverty of BPS, 2011 District Statistical Yearbook published by BPS for all districts. The raw shapefile of West Java map. We focus on 545 sub-districts of West Java of Indonesia by using aggregated data at sub-district level. Poverty Rates (% of population) Source: Authors 7
Cluster Measurements 8
Spatial Regression Specifications ModelsSpecifications SAR (Spatial Autoregressive Model) Spatial lags in dependent variable (lnpov) for two baseline models SDM (Spatial Durbin Model) Spatial lags in both dependent variable and main explanatory variables (hc s and K s ) SEM (Spatial Error Model) Spatial lags in error terms for two baseline models 9
Baseline models: poverty rates and agro-clusters sq_hc s : squared cluster size; Xi,s : control variables, including Growth of farmer number, % smallholders, % farmers with 55 years old, population density, total area, % wetland, productivity, distance and time travel to the nearest city. D: dummy variable (1 = rural regions; 0 = urban areas) Independent Variables Cluster size (hc s ) Specialization index (K s ) 10
Results Variable (y= lnpov) SARSEMSDMSARSEMSDM Original variables Cluster size (hc) Square cluster size Krugman index (Ks) Growth of farmer number % of smallholders % of farmers > 55 years Population density Total area of sub-district % of wetland in total Productivity Distance Travel time Dummy rural region Spatially lagged variables Cluster size (hc) 0 Krugman index (Ks) 0 Population density 0 0 Travel time Note: + and – are significant at 5% 11
Conclusions The higher number of farm labourers, the lower poverty is, but after reaching the optimal average number, the poverty rate rise. The more specialization in agriculture or non-agriculture, sub- district may experience the lower poverty rate. The direct effects of agro-clusters on poverty rates in the sub- district itself are larger than the spillover effects. 12
Policy Remarks In the sense of specialization foster the specialisation of the sub-districts establish production nuclei control input availability and prices improve market access In terms of maintaining cluster size improve farmers’ capabilities and knowledge. enforce networks Spatial spillover effects improve the availability and quality of infrastructures 13
Agriculture in West Java Economic Agglomeration
Thank you
Testing Spatial Dependence OLS? Diagnostic tests on baseline models Contiguity spatial weight matrix Moran I’s spatial autoregression,
Spatial Regression Specifications ModelsSpecifications SAR SDM SEM
Findings Variable (y= lnpov) hc as X variableKs as X variable SARSEMSDMSARSEMSDM Original variables Cluster size (hc) *** *** *** Square cluster size **0.0038**0.0035** Krugman index (Ks) ** ** ** Growth of farmer number0.0004*** **0.0003*** % of smallholders % of farmers > 55 years *** *** *** *** *** *** Population density *** *** *** *** *** *** Total area of sub-district0.0008*0.0010** % of wetland in total0.0021***0.0025***0.0019***0.0017**0.0021***0.0015** Productivity *** ** *** ** * ** Distance Travel time0.0897***0.1192***0.1142***0.0746***0.1001***0.0994*** D Spatially lagged variables Cluster size (hc) Krugman index (Ks) Population density Travel time ** * ***2.5814***1.6542***1.8900***2.6569***1.7605*** *** ***0.3864*** *** *** *** Note: One, two and three asterisks denote significance at the 10%, 5% and 1% level, respectively.