ارزیابی روشهای جدید تجربی سازمان صنعتی: موارد از پنج صنعت

The methods proposed in the new empirical industrial organization (NEIO) literature have made significant contributions to our understanding of competitive behavior. However, these methods have yet to be compared with each other for their performance in explaining and diagnosing competitive market conduct. This inter-method comparison is important because conclusions about competitive behavior based on these methods have significant strategic as well as policy implications for firms. Our objective in this paper is to examine the performance of these different NEIO methods in terms of their discriminatory power, ability to identify strategic variables, and robustness in estimation. For empirical demonstration, we use data from diverse industries such as microprocessors, personal computers, facial tissue, disposable diapers and automobiles. Our results suggest that two commonly used NEIO methods-conjectural variation and non-nested model comparison-exhibit quite good convergence with each other and are consistent with a traditional time series method. This suggests that simpler methods such as conjectural variations deserve more credit. We also find that using these methods in tandem provides valuable additional information that may not be available when using any one method alone. While the emphasis in this study is on comparing different methods of analyzing competitive interaction, the findings also reveal some substantive insights about each market studied.

There has been an increase in empirical research in both economics and business analysis that studies competitive conduct using what is often referred to as the new empirical industrial organization (NEIO) paradigm. A review of the NEIO and related methods is provided in Bresnahan (1989) and more recently in Kadiyali, Sudhir, and Rao (2001). Many of these NEIO studies have used the conjectural variations (CV) approach to infer competitive conduct, which assumes that each firm believes that its choice of price (or some other strategic variable) will affect the price selected by its rival, and that the rival's reaction can be captured by a single parameter (Iwata, 1974). While a ‘single coefficient’ approach has obvious appeal, it is not without its weaknesses. Corts (1999) and Kim and Knittel (2004) illustrate that there can be severe biases in estimates of mark-up levels in marginal cost models using the conjectural variations (conduct parameters) framework. Interpretation of parameters poses another problem. For example, the CV method assumes that a positive CV parameter (price raises met with price increases) indicates cooperation. But a positive CV also implies that price decreases are met with price decreases. Therefore, it is not clear if meeting a price decrease with a price decrease of similar magnitude implies cooperation or a non-cooperative tit-for-tat strategy. The main alternative to the CV method is the Menu Approach, also referred to as the Non-Nested Model Comparisons (NNMC) approach. This method requires the alternative competitive models to be developed and the solutions obtained under different assumptions about competing firms' behavior such as Nash, Stackelberg, etc. Assuming that the observed market data reflect the equilibrium corresponding to a particular mode of conduct (e.g., Nash or Stackelberg), the mode of conduct that provides best fit to the data is considered the most accurate description of the competitive structure of the market. The purpose of this study is to compare the performance of these two popular NEIO methods in terms of their discriminatory power among different competitive modes, ability to identify strategic variables (price or quantity), as well as robustness to time aggregation and the sample size. We do this comparison by applying the two methods to data from five different markets (microprocessors, personal computers, facial tissue, disposable diapers and automobiles) and compare the results for each market. We also compare the results from the CV and NNMC methods with those obtained by the Granger Causality (GC) test using the same performance criteria. A time-series based approach, the GC test has been widely used to study competitive interactions (Hanssens, 1980), and a comparison of the CV and NNMC methods with the GC test allows us to examine the extra value-added of the NEIO methods over traditional methods.

We applied three distinct methods to understand competitive conduct in five different product markets. In addition to demonstrating the applicability of these methods to real market data, our examples highlight certain strengths and weaknesses of each method. The overall results point to the strength of these methods because of the observed consistency. Each method, however has its own strength and weaknesses, and therefore should be used in tandem with the other methods. The current study is not free from several concerns. First, we did not include cost data in our study because we were not able to obtain data on relevant input factor prices for each of the industries we covered. It is possible to estimate unknown marginal costs within both the CV and NNMC frameworks and modeling the cost side explicitly may provide some additional insights. We hope this will be tackled in a related future study. Second, we used a linear demand function in our analysis. It may be argued that other forms of demand function (e.g., nonlinear demand function) are more appropriate for some markets. However, it is very difficult to specify consistent models of nonlinear demand function when they have price as the outcome and quantity as the control variable. For example, if one were to use a logit-type model, one might have to implicitly assume that quantity is not the strategic variable. A linear model is more flexible in this regard. In fact, the linear (or log-linear) type of demand function has been widely accepted in previous research based on its merit of analysis (Bresnahan, 1989 and Rubinovitz, 1993). In this study a linear demand model also provided excellent fit to the data. However, future research efforts with different demand specifications are warranted. Third, each of the markets we study consists of a number of competing brands. However, we focus on sub-markets within each of these markets and consider the two leading brands in each market. The competing firms are assumed to maximize profits from the brand in question and not the entire product line. As it turns out, all the markets we study do in fact have two dominant firms. A fourth, more general concern, is that our models might be subject to the Lucas critique (Van Heerde, Dekimpe, & Putsis, 2005) — firms might change their behavior and not act as they have been found to in the past, if there are changes in the structure of series relevant to decision makers. Any policy recommendations based on our analysis would have to be tempered by this realization. However, according to Van Heerde et al. (2005) the Lucas critique might not be as applicable to structural models such as ours, which focus on short-term decisions.