Markup and Cost Dispersion across Firms: Direct Evidence from Producer Surveys in Pakistan

摘要

How do markups differ across firms in the same narrow industry? The question is key to answering a number of bigger questions-How do firms respond to trade liberalization or industrial-policy interventions? How large are the welfare gains from trade? How do firms transmit international price shocks to local markets? How should productivity be measured?-but there is little consensus in the literature either about how to model markups or about how to estimate them. The most common theoretical approach is to assume that firms are monopolistically competitive and that representative consumers have constant elasticity of substitution (CES) preferences (e.g., Melitz 2003), with the result that markups are constant in multiplicative terms. Recently, there has been growing interest in more flexible approaches to modeling markups. Researchers have developed tractable models with variable markups by combining, for instance, CES preferences and other forms of imperfect competition (e.g., Atkeson and Burstein 2008; Edmond, Midrigan, and Xu 2015; Amiti, Itskhoki, and Konings 2014), or other demand systems and monopolistic competition (e.g., Melitz and Ottaviano 2008; Zhelobodko et al. 2012). But it is not clear how to discriminate among these models based on existing empirical evidence. Empirically, there are two main approaches to estimating markups. The most common is to estimate markups from the "demand side," which entails specifying functional forms for consumer utility and making assumptions on the market structure of firms (e.g., Berry, Levinsohn, and Pakes 1995; Goldberg 1995; Feenstra and Weinstein 2010). Alternatively, one can estimate markups through the "supply side" using an approach pioneered by Hall (1986) and more recently implemented by De Loecker and Warzynski (2012) and De Loecker et al. (2012). This approach does not require functional-form assumptions about preferences or the competitive environment, and instead relies on cost minimization and two additional assumptions: that input adjustment is costless and that firms in the same industry face different, exogenously given, input prices. Studies adopting either approach find that markups do in fact vary tremendously across firms. But while each approach to estimating markups has its merits, both require a measure of faith in their respective assumptions. Ideally, one would like to have direct measures of markups that do not depend on these theoretical assumptions. In this short paper, we take a different approach: we have directly asked firms about their markups. Since 2011, we have been studying a cluster of soccer-ball manufacturers located in Sialkot, Pakistan and have learned a great deal about the ins and outs of producing soccer balls. In Atkin et al. (2014), we study the adoption of a new cost-saving technology that we invented and randomly introduced to a subset of firms. In conjunction with that study, we have directly collected information on prices and profitability, as well as the unit costs of manufacturing a standardized ball. Asking firms directly about their markups is not a guarantee of accurate information. However, the fact that the producers use similar production techniques that we understand well and that we have been able to build up a modicum of trust with the firms over the course of our several-year project give us confidence that there is signal in the self-reported markups. As we discuss in more detail below, we have six main findings. First, dispersion in markups is greater than dispersion in costs, at least in proportional terms. Second, both costs and markups are positively related to firm size, arguably the best observable proxy for the entrepreneurial-ability parameter in many heterogeneous-firm models (e.g., the "Melitz draw," ∮). Third, the elasticity of markups with respect to firm size is significantly greater than the elasticity of costs. Fourth, larger firms have greater costs primarily because they use higher-quality inputs (which are costly). Fifth, larger firms charge higher markups on average both because they produce a greater share of higher-quality ball types, which carry higher markups, and because they charge higher markups for a given type of ball. Sixth, there is suggestive evidence that marketing efforts may play an important role is generating higher average markups: markups are more highly correlated with whether an entrepreneur attends an annual trade expo and sells to richer countries than with available measures of technical efficiency. In the conclusion we discuss what we consider to be the main implications of these findings for theoretical and empirical work on heterogeneous firms.