Presentation on theme: "1 Econometric Tools for Analyzing Market Outcomes Focus on Production Functions Daniel Ackerberg, C. Lanier Benkard, Steven Berry, and Ariel Pakes. (2007)"— Presentation transcript:
1 Econometric Tools for Analyzing Market Outcomes Focus on Production Functions Daniel Ackerberg, C. Lanier Benkard, Steven Berry, and Ariel Pakes. (2007) Handbook of Econometrics
3 OLS? We have known since Marshak and Andrews (1944) that direct OLS estimation of production function is problematic Simultaneity and selection issues Two traditional solutions to these endogeneity problems are instrumental variables and fixed effects
6 The Olley and Pakes (1996) Approach One might also look at Wooldridge (2004), who presents a concise, one-step, formulation of the OP approach for which standard error derivations are more straightforward. This one-step approach may also be more efficient than the standard OP methodology.
13 Test of Olley and Pakes’ Assumptions ? Conditional on capital and age, there is a one to one mapping between investment and productivity There is the direct assumption that the choice of labor has no dynamic implications; i.e. that labor is not a state variable in the dynamic problem But assume instead that there are significant hiring or firing costs for labor, or that labor contracts are long term (as in, for example, unionizedindustries), then the labor has dynamic
15 Comparing the models with our data set OLSFEOPLP Age0.007***0.264***-0.604*-- Capital0.028***0.021***0.0020.008*** Materials0.016***0.009***0.0015** Unskilled Labour 0.558***0.458***0.419***0.374*** Skilled Labour 0.435***0.136***0.361***0.235*** In our case the estimated TFP values for OP and LP are highly correlated (0.9992)
16 TFP Will try to estimate TFP with GMM soon...!!
18 Concluding Remark The appropriateness of different models of how these decisions are made will undoubtedly depend on the environment being studied. There are undoubtedly institutional settings where alternative frameworks might be better to use. It is not the precise framework that is important, but rather the fact that productivity studies must take explicit account of the fact that changes in productivity (or, if one prefers, sales for a given amount of inputs) in large part determine how firms respond to the changes being studied, and these must be taken into account in the estimation procedure.