Resource misallocation—which occurs when economic resources are not put to their best use—can arise for many reasons, including market failures (e.g. monopolies curtailing production) or policy failures (e.g. India’s complex licensing system, China’s support of state-owned enterprises, France’s strict labor regulations).1 Structural reforms are often motivated by the hope that shifting capital and labor to “better use” in more productive firms will give a large boost to aggregate productivity and GDP.
But how large are the gains from reducing misallocation? Put another way, is misallocation as bad as we instinctively think it is? The increasing availability of micro data, at the firm level, alongside advances in computing power, allow for new approaches to answer this decades-old question. In a seminal paper, Chang-Tai Hsieh and Peter Klenow argue that the dispersion of revenue productivity across firms (sales per unit of capital and labor) points to very costly misallocation in India and China. Consider a simple example: firm A has 10 workers and makes $500,000 a year ($50,000 a worker), while firm B has 50 workers and makes $1 million a year ($20,000 a worker). If one worker from firm B moves to firm A, the total number of workers stays constant but total revenue increases by $30,000: less worker misallocation and greater aggregate productivity! Based on this novel “dispersion approach” Hsieh and Klenow found that misallocation could explain roughly half the aggregate productivity gap between India and China and the United States. This approach to measuring misallocation has been highly influential.2
However, a recent paper by Mark Bils, Peter Klenow, and IMF economist Cian Ruane documents a dramatic rise in revenue productivity dispersion in the United States since the 1970s, so large that misallocation in the 2000s appears worse in the United States than in India (Figure 1). Could market distortions really have increased so much in the United States? The authors suggest a different explanation: the standard approach to measuring misallocation is inflated due to measurement error in survey data. This measurement error could have many causes: managers may not want to disclose information about their firm, surveys may not be completed by those with the most information, data may be imputed, and certain inputs in production are inherently difficult to measure. The authors find that “true” misallocation is not nearly as bad in India and the United States as it appears based on the data. In addition, misallocation in the United States has not grown as much as it seemed at first.
Measurement problems have always been ubiquitous in survey data, with low response rates and poor data quality causing problems for statisticians and researchers. But survey data quality might in fact be deteriorating over time. An article in the Economist, summarizing recent research, documented declining survey response rates in Canada, the United Kingdom, and the United States over time. These surveys inform policy questions regarding labor force participation, unemployment, inequality, misallocation, and the like, which means that this is an area of concern for all policy-oriented macroeconomists.
Another IMF economist working on measurement error is Ippei Shibata, who in a recent IMF working paper finds that misreporting in household surveys affects our measures of aggregate unemployment. “Evidence suggests that households misreport their employment status. Once we correct for such measurement error, the US unemployment rate could be close to a percentage point higher on average,” says Shibata. But there is hope: “Fortunately, we can correct for misreporting in household surveys around the world provided that households report their employment status for at least three consecutive months.”
There is no silver bulletforthe problem of measurement error in survey data. But misallocation will continue to be relevant as policymakers struggle to make the most of their countries’ resources. There has been good progress in the measurement of misallocation, thanks to the increasing availability of micro data, butthere is a long road ahead when it comes to disentangling the sources of misallocation and its policy implications. Studying the impact of structural reforms and other policy “experiments” will be useful to gauge whether there are big gains from reducing misallocation.