36 5 2015 5 Research on Economics and Management Vol. 36 No. 5 May 2015 490 490 F323. 9 A DOI:10.13502/j.cnki.issn1000-7636.2015.05.007 1000-7636 2015 05-0052 - 10 2008 836 70% 1. 2 2010 1 2 3 2015-03 - 13 08AJY042 100872 52
Research on Economics and Management No. 5 2015 2015 5 4-7 8 9 2004 8 2009 9 self - selection 490 Propensity Score Matching PSM 1 lny ij = β 0j + β 1j lnfs ij + β 2j lnsize ij + β 3j Stru ij + β 4j Infr ij + β 5j Type ij + μ ij 1 Y ij j = 1 2 3 i j FS ij i Size ij Stru ij Infr ij Type ij Size ij i 1 OLS 0 0 1 53
2015 5 Research on Economics and Management No. 5 2015 Stru ij i Infr ij i Type ij i 1 1 μ ij β ij 1 1 1 Y ij 8. 80 1. 73 FS ij Size ij Stru ij Infr ij Tape ij 5. 86 1. 47 7. 44 1. 82 0 2. 03 2. 29 5. 32 1. 28 5. 07 1. 28 8. 82 1. 25 3. 42 1. 24 1 0 0. 62 100% 0. 72 0. 30 1 0 0. 40 1 0 0. 42 1 0 0. 52 1 0 0. 38 1 0 0. 48 1 0 0. 44 FS ij β 1j j = 1 j = 2 j = 3 β 1j cause - effect relationship Rosenbaum & Rubin 1983 10 selection bias 1 2 PSM p X = Pr D = 1 X = E D X 2 D D = 1 D = 0 X p X propensity score 1 Logistic 2 11 12 13-14 15 16 17 54
Research on Economics and Management No. 5 2015 2015 5 PS 2 Average effect of Treatment on the Treated ATT Becker & Ichino 2002 18 i p X i ATT ATT = E Y 1i - Y 0i D i = 1 = E{ E Y 1i - Y 0i D i = 1 p X i } = E{ E Y 1i D i = 1 p X i - E Y 0i D i = 0 p X i D i = 1} Y 1i Y 0i p X Lian et al. 2011 19 2012 17 Nearest Neighbor Matching i i 1 T i i T N C i j C i w ij = 1 /N C i T N T Becker & Ichino 2002 ATT 3 3 ATT = 1 /N T lny T i - 1 /N T w j lny C i 4 i T j C T C 2012 2012 2 ~ 4 490 490 11% 456 / 2011 2221 / 2011 3163 / 2011 490 3. 1 1415. 9 11% 490 184 109 381 1 Radius Matching Kernel Matching 55
2015 5 Research on Economics and Management No. 5 2015 48. 29% 3. 3 2. 1 1. 1 75% 150 50% 50 1000 8 21121 9349 1. 26 2470 10428 785 6416 1 1 Stata11 1 2 F 2 2 Ⅲ Ⅳ Ⅶ Ⅷ Ⅺ Ⅻ 0. 069 ** 0. 060 * 0. 062 2. 18 1. 83 1. 39 0. 484 *** 0. 464 *** 0. 459 *** 0. 437 *** 0. 533 *** 0. 510 *** 7. 89 7. 51 7. 02 6. 61 6. 73 6. 24 0. 471 *** 0. 442 *** 0. 538 *** 0. 518 *** 0. 299 0. 264 3. 26 3. 07 3. 73 3. 61 1. 56 1. 39 0. 617 *** 0. 562 *** 0. 163 0. 095-0. 0512-0. 101 3. 66 3. 33 0. 92 0. 53-0. 23-0. 46-0. 196-0. 170 0. 055 0. 087 0. 499 ** 0. 533 ** - 1. 26-1. 10 0. 34 0. 55 2. 42 2. 58 0. 571 *** 0. 538 *** 0. 389 *** 0. 355 ** - 0. 147-0. 171 3. 89 3. 60 2. 65 2. 37-0. 73-0. 84-0. 529 *** - 0. 508 *** - 0. 027-0. 016-0. 358 * - 0. 358 * - 3. 61-3. 45-0. 19-0. 11-1. 86-1. 87 3. 829 *** 3. 908 *** 1. 235 ** 1. 358 ** 2. 655 *** 2. 789 *** Ⅳ 0. 069 5% 1% 0. 069% Ⅷ 0. 060 10% 1% 7. 13 7. 25 2. 22 2. 44 3. 93 4. 07 0. 06% 339 334 329 326 323 320 Ⅶ R 2 0. 352 0. 363 0. 252 0. 261 0. 198 0. 204 1 *** ** * 1% 5% 10% t 2 56
Research on Economics and Management No. 5 2015 2015 5 2006 42 2006 2200 2007 self - selection selection bias T 3 3 T 181 194 8. 73 8. 15 3. 67 *** 1. 64 1. 42 179 187 6. 16 5. 58 3. 87 *** 1. 52 1. 35 178 182 7. 70 7. 21 2. 57 ** 1. 85 1. 78 175 169 5. 63 4. 95 5. 10 *** 1. 18 1. 29 181 193 9. 01 8. 56 3. 46 *** 1. 23 1. 27 169 179 5. 21 4. 91 2. 27 ** 1. 32 1. 20 174 186 3. 51 3. 17 2. 70 *** 1. 25 1. 15 176 185 0. 74 0. 71 0. 94 0. 30 0. 30 *** ** * 1% 5% 10% Logit 16 13 20 Logit Logit 4 Logit 57
2015 5 Research on Economics and Management No. 5 2015 Pseudo - R 2 AUC 21 V Pseudo - R 2 AUC V 4 Logit Ⅰ Ⅱ Ⅲ Ⅳ Ⅴ 0. 512 *** 0. 461 *** 0. 431 *** 0. 407 *** 0. 402 *** 5. 06 4. 37 4. 13 3. 33 3. 45 1. 247 *** 1. 205 *** 1. 315 *** 1. 471 *** 1. 385 *** 0. 119 * 3. 03 2. 85 3. 09 3. 24 3. 06 1. 79 0. 267 *** 0. 257 *** 0. 224 ** 2. 98 2. 67 2. 27 0. 291 0. 95-0. 132-0. 46 0. 162 0. 55-0. 696 ** - 2. 07 0. 518 * 1. 68 0. 550 * 1. 80-3. 559 *** - 4. 112 *** - 4. 730 *** - 4. 854 *** - 4. 665 *** Logit PS 1 1 PS - 5. 12-4. 88-5. 51-5. 08-5. 04 Pseudo R 2 0. 071 0. 071 0. 085 0. 092 0. 107 AUC 0. 677 0. 677 0. 706 0. 714 0. 727 332 320 325 295 302 *** ** * 1% 5% 10% PS 5 1 PS 58
Research on Economics and Management No. 5 2015 2015 5 5% 5 t t - test p > t 5. 64 4. 99 4. 60 0. 000 0. 39 0. 39 0. 14 0. 886 5. 64 5. 58 0. 46 0. 644 0. 39 0. 47-1. 27 0. 204 0. 74 0. 69 1. 44 0. 152 0. 59 0. 45 2. 55 0. 011 0. 74 0. 78-1. 36 0. 175 0. 59 0. 60-0. 12 0. 907 6. 20 5. 60 3. 76 0. 000 0. 53 0. 33 3. 44 0. 001 6. 20 6. 33-0. 83 0. 406 0. 53 0. 48 0. 80 0. 424 t t - test p > t ATT ATT 6 1 ATT 5% ATT 6 ATT ATT s. e. t - value 8. 760 8. 202 0. 558 0. 176 3. 17 *** 8. 760 8. 608 0. 152 0. 272 0. 56 6. 214 5. 620 0. 593 0. 160 3. 72 *** 6. 214 6. 204 0. 010 0. 253 0. 04 7. 769 7. 220 0. 549 0. 213 2. 57 ** 7. 769 7. 853-0. 084 0. 318-0. 26 1 2 ATT t 3 *** ** * t 1% 5% 10% ATT 490 1 ATT 59
2015 5 Research on Economics and Management No. 5 2015 T 50% 1. 2009 M. 2010. 2. J. 2005 2 33-39. 3. WTO D. 2006. 4. J. 2007 2 24-28. 60
Research on Economics and Management No. 5 2015 2015 5 5. J. 2009 9 91-108. 6. 2004-2007 J. 2010 4 100-106. 7. DEA 87 J. 2011 12 82-85. 8. J. 2004 10 33-40. 9. D. 2009. 10 ROSENBAUM P R RUBIN D B. The central role of the propensity score in observational studies for causal effects J. Biometrika 1983 70 1 41-55. 11 BRAND J E YU X. Who benefits most from college Evidence for negative selection in heterogeneous economic returns to higher education J. American Sociological Review 2010 75 2 273-302. 12 RUBIN D B. Estimating causal effects from large data sets using propensity scores J. Annals of Internal Medicine 1997 127 8 757-763. 13 MORGAN S L WINSHIP C. Counterfactuals and causal inference methods and principles for social research M. New York Cambridge University Press 2007. 14 WINSHIP C SOBEL M S. Causal inference in sociological studies A HARDY M A BRYMAN A. Handbook of Data Analysis C. CA Sage Publications 2004. 15 SOBEL M E. Causal inference in the social sciences J. Journal of the American Statistical Association 2000 95 450 647-651. 16 IRWIN E A KLENOW P J. High-tech R&D subsidies estimating the effects of sematech J. Journal of International Economics 1995 40 323-344. 17. J. 2012 4 70-81. 18 BECKER S ICHINO A. Estimation of average treatment effects based on propensity scores J. The Stata Journal 2002 4 358-377. 19 LIAN Y SU Z GU Y. Evaluating the effects of equity incentives using PSM evidence from China J. Frontiers of Business Research in China 2011 5 2 266-290. 20. J. 2012 1 221-246. 21 HOSMER D LEMESHOW S. Applied logistic regression M. New York John Wiley & Sons Inc. 2000. Effect of Public Financial Support for Agricultural Product Wholesale Market A Propensity Score Matching Analysis Based on the 490 Market Data in the Cooperative System ZENG Yinchu CHENG Xiaoping Renmin University of China Beijing 100872 Abstract Based on the investigation data of 490 markets in the cooperative system the effects on transaction scale employment and farmer development of public financial support for agricultural product wholesale market are analyzed by multiple regression models and the causal inference bias related to sample self selection is discussed by using propensity score matching method. The regression result show that the public financial support has significant effects on the transaction scale and the employment of the wholesale market but if the causal inference bias related to sample self selection is controlled these effects will not become significant which means the effects of public financial support will be over - evaluated by the ordinary regression model. Keywords agricultural product wholesale market public financial support policy effect propensity score matching 61