TITLE:
A Chance–Constrained Data Envelopment Analysis Approach to Problem Provincial Productivity Growth in Vietnamese Agriculture from 1995 to 2007
AUTHORS:
Nguyen Khac Minh, Pham Van Khanh
KEYWORDS:
Total Factor Productivity, Technical Efficiency Change, Technological Progress, Chance-Constrained Programming
JOURNAL NAME:
Open Journal of Statistics,
Vol.1 No.3,
October
20,
2011
ABSTRACT: This study employs a chance-constrained data envelopment analysis (CDEA) approach with two models (model A and model B) to decompose provincial productivity growth in Vietnamese agriculture from 1995 to 2007 into technological progress and efficiency change. The differences between the chance - constrained programming model A and model B are assumptions imposed on the covariance matrix. The decomposition allows us to identify the contributions of technical change and the improvement in technical efficiency to productivity growth in Vietnamese production. Sixty-one provinces in Vietnam are classified into Mekong - technology and other -technology categories. We conduct a Mann-Whitney test to verify whether the two samples, the Mekong technology province sample and the other technology sample, are drawn from the same productivity change populations. The result of the Mann-Whitney test indicates that the differences between the Mekong technology category and the other technology category from two models are more significant. Two important questions are whether some provinces in the samples could maintain their relative efficiency rank positions in comparison with the others over the study period and how to further examine the agreements between the two models. The Kruskal - Wallis test statistic shows that technical efficiency from both models for some provinces are higher than those of them in the study period. The Malmquist results show that production frontier has contracted by around 1.3 percent and 0.31 percent from chance-constrained model A and model B, respectively, a year on average over the sample period. To examine the agreements or disagreements in the total factor productivity indexes we compute the correlation between Malmquist indexes, which is positive and not very high. Thus there is a little discrepancy between the two Malmquist indexes, estimated from the chance - constrained models A and B.