class: center, middle, inverse, title-slide # Disparities in Regional Productivity, Capital Accumulation, and Efficiency across Indonesia: ## A Convergence Clubs Approach ### Carlos Mendez
https://carlos-mendez.rbind.io
Graduate School of International Development
Nagoya University
Mitsuhiko Kataoka
Graduate School of Business
Rikkyo University ### Prepared for the 2019 Applied Regional Science Conference (ARSC)
[ Slides and paper available at:
http://bit.ly/arsc2019
] --- ## Motivation: - Large regional inequality in Indonesia (Esmara 1975; Mishra 2009; Kataoka 2018) - A large literature focuses on the **_dynamics_** of regional inequality - MOST papers use the classical convergence approach of Barro and Sala-i-Martin (1991) - Describe the dynamics of the "average" region (Garcia and Soelistianingsih 1997; Resosudarmo and Vidyattama 2006; Hill, Resosudarmo, and Vidyattama 2007; Vidyattama 2013; 2014). - FEW papers study the dynamics of regional convergence "beyond the average" (Sakamoto 2007; Kurniawan et. al, 2019) - Kurniawan et. al (2019 Reg. Sci PP) use the novel convergence approach of Phillips and Sul (2007 Econometrica) to study regional income convergence in Indonesia **_beyond the average_** - Focus on **the role of heteregeneity** both across regions and over time - Provinces in Indonesia are converging to TWO separate clubs ## Research Objective: Study the determinants of regional per-capita income: labor productivity, capital accumulation and efficiency - Labor productivity = F(Physical capital, Human capital, Efficiency) --- class: middle ## Methods: - Nonlinear dynamic factor model (Phillips and Sul, 2007, 2009) - Clustering algorithm for panel data (Phillips and Sul, 2007, 2009) ## Data: - Labor productivity, phyical capital, human capital, efficiency (Katoka, 2013,2018) - 26 Indonesian provinces over the 1990-2010 period ## Main Results: 1. Cross-provincial dynamics of labor productivity are characterized by **TWO convergence clubs** 2. The dynamics of the proximate determinants of labor productivity show some mixed results: - Physical and human capital are characterized by **multiple convergence clubs** (FOUR and TWO convergence clubs, respectively) - Two alternative measures of efficiency are characterized by **ONE convergence club** --- class: middle # Outline of this presentation 1. Some stylized facts 2. Convergence framework (intuition) 3. Main results of the paper - Two convergence clubs in labor productivity - Four convergence clubs in physical capital - Two convergence clubs in human capital - One convergence club in efficiency (non-parametric and parametric) <br /> <br /> [ Slides and paper available at: http://bit.ly/arsc2019 ] --- class: center, middle # (1) Some stylized facts **Regional heterogeneity across Indonesia** --- class: middle,center ## Are there any signs of convergence in labor productivity? ![](figs/fig01.jpg) --- class: middle,center ## Are there any signs of convergence in the determinants of labor productivity? ![](figs/fig02.jpg) --- class: center, middle # (2) Convergence framework Convergence test (intuition) Convergence clubs (intuition) --- class: middle # Convergence framework (brief overview) - First, define a relative transition parameter, `\(h_{it}\)`, as `$$h_{it}=\frac{y_{it}}{\frac{1}{N}\sum_{i=1}^{N}y_{it}}$$` - Second, the convergence hypothesis is defined as `$$H_{t}=\frac{1}{N}\sum_{i=1}^{N}\left(h_{it}-1\right)^{2}\rightarrow 0$$` In other words, when the relative transition parameter converges to unity, `\(h_{it}\rightarrow1\)`, the cross-sectional variance converges to zero, `\(H_{t}\rightarrow0\)`. - Thrid, Phillips and Sul (2007) test this hypothesis by using the following log t regression model `$$log\left(\frac{H_{1}}{H_{t}} \right)-2log\left\{ log\left(t\right)\right\} = a+b\:log\left(t\right)+\epsilon_{t}$$` --- class: middle, center # Convergence test (intuition) ![](figs/convergence-test.jpg) --- class: middle, center # Convergence clubs (intuition) ![](figs/convergence-clubs.jpg) --- class: middle, center # (3) Main results Overall results Characteristics of the two convergence clubs in labor productivity Characteristics of the multiple convergence clubs in capital accumulation Characterstics of the unique convergence club in efficiency --- class: middle, center ## Overall results: Convergence and Divergence ![](figs/tab01.jpg) --- class: middle, center ## LABOR PRODUCTIVITY: Two convergence clubs ![](figs/fig03.jpg) --- class: middle, center ## LABOR PRODUCTIVITY: Members of the clubs ![](figs/fig04.jpg) --- class: middle, center ## PHYSICAL CAPITAL: Four convergence clubs ![](figs/fig05.jpg) --- class: middle, center ## PHYSICAL CAPITAL: Members of the clubs ![](figs/fig06.jpg) --- class: middle, center ## HUMAN CAPITAL: Two convergence clubs ![](figs/fig07.jpg) --- class: middle, center ## HUMAN CAPITAL: Members of the clubs ![](figs/fig08.jpg) --- class: middle, center ## EFFICIENCY: One convergence club ![](figs/fig09.jpg) Note: Efficiency is measured in absolute levels using a Cobb-Douglas production function with elasticity parameter of 0.33. The mean and the 95 confidence interval is computed for each year --- class: middle # Concluding Remarks Reject the (overall) convergence hypothesis in labor productivity and two of its determinants: physical capital and human capital - Labor productivity: Two largely separated clubs - Physical capital: Four clubs with separating trends at the extremes - Human capital: Two clubs with separating trends A unique convergence club in efficiency - Low effeciency club: Is this a low-efficiency trap? ## Implications, discussion, and further research Regional heterogeneity is a pervasive feature in Indonesia (and many developing countries) - The need for an analaytical framework that focuses on heterogeneity and goes beyond the average Convergence clubs may help us identify economies facing similar challenges - Call for better coordination and cooperation policies both within and between clubs Masked within provinces in Indonesia, there is still a high degree of heterogeneity that is worth exploring. - Using district level data is the next step and using firm level data is the following --- class: center, middle # Thank you very much for your attention https://carlos-mendez.rbind.io Slides and working paper available at: http://bit.ly/arsc2019 ![](figs/QuaRCS-lab-logo2.png) **Quantitative Regional and Computational Science lab** https://quarcs-lab.rbind.io *** C. Mendez: This research project was supported by JSPS KAKENHI Grant Number 19K13669