Comment Karen Helene Ulltveit-Moe, University of Oslo and CEPR I. Contribution of the Paper Recessions hit differently. They hit differently across sectors and regions. Focusing on the latter, Fogli, Hill, and Perri emphasize that recessions are not purely aggregate phenomena, but vary significantly across space. They document geographical features of the US business cycle over the last 30 years, and argue that—although usually not included in macro analysis—geography is important in explaining the transmission, amplification, and propagating of business cycles. Their paper starts by presenting the spatial patterns of unemployment in the United States during the Great Recession (2007 to 2009). They move on to analyzing spatial correlation and dispersion of unemployment during this last recession as well as the three previous recessions, and present three main findings: (a) at any point in time, unemployment across US counties is significantly spatially dispersed; (b) during most recessions, spatial correlation falls before and at the start of the recession, then increases sharply at the beginning of the recession and stabilizes toward the end; (c) spatial dispersion typically increases during the recession, then stabilizes and falls thereafter. Motivated by geographical differences in unemployment, the authors build a simple theoretical model of unemployment with many counties and spatial heterogeneity, as well as spatial transmission of countyspecific shocks. They calibrate the model, and find that the model can generate patterns of spatial correlation and spatial dispersion that are broadly consistent with the patterns observed under most recessions. Finally, the model is used to evaluate the importance of local geographical connections for the transmission of shocks and unemploy- © 2013 by the National Bureau of Economic Research. All rights reserved. 978-0-226-05327-1/2013/2012-0071$10.00 9121.indd 332 3/19/13 1:26 PM Comment 333 ment dynamics. It is found that local connections amplify large aggregate shocks, while the same local connections serve to mute the effect of small aggregate shocks. II. Location Matters Unemployment maps illustrating the geographical differences across US counties during the Great Recession work well to motivate the paper. Looking at deviation of unemployment within each county from its long term, we see that unemployment is not randomly distributed across space. Instead we observe that groups of counties that are located close to each other either experience higher or lower unemployment than the average county. In line with what we would expect from just eyeing the map, countyspecific unemployment is both significantly spatially correlated and spatially dispersed. This result forms an important basis for the analysis. However, the argument may be made that the spatial characteristics of unemployment that we observe have nothing to do with location as such, but rather to do with geographical specialization stretching across the borders of single counties. It is well recognized that the United States is characterized by significant industrial specialization and localization (see, e.g., Krugman 1991; Kim 1995). The presence of strong industrial specialization and localization implies that spatial correlation may not be the result of local transmission channels of shocks from one county to another, but instead can be explained by similar industrial structure in neighboring counties. To account for similarity in industrial structure, industry controls are introduced. It almost goes without saying that it is important that these controls are not based on too much of an aggregate industry classification, as this would hide patterns of specialization and make the controls ineffective. But nor does one want a too disaggregated classification, since this does not allow us to take into account the effect of industrial cluster and supplier networks stretching across county borders. Finally, it is important to recognize that industrial structures change over time, and thus use time-specific controls. In the present version of the paper, industry controls are time-specific, and based on NAICS three-digit code (which includes 86 sectors), while in an earlier version of the paper a much cruder industry classification was employed. Regardless of aggregation level, we see that accounting for industrial structure reduces spatial correlation as well as spatial dis- 9121.indd 333 3/19/13 1:26 PM 334 Ulltveit-Moe persion, but both statistics still take on values that are still significantly different from zero. Hence, there is reason to conclude that location does indeed matter, and that local transmission channels may be important for explaining the dynamics of county-specific unemployment. III. Are Neighbors Good or Bad? Fogli, Hill, and Perri build a simple model with two channels of local geographic connection between counties: (a) contemporaneous correlation of local conditions (e.g., due to intercounty trade); and (b) unemployment in a county depends on the lagged unemployment of neighboring counties (e.g., due to migration). Calibrating this model allows them to assess the importance of local geographic connections for aggregate unemployment dynamics and in particular to examine whether these connections serve to amplify or mute the effect of shocks. The authors find that the role of neighboring counties differs, and depends on the magnitude of the aggregate shock that hits the economy. If the economy is hit by a small aggregate shock, then local transmission serves as a means to mute the effect of the shock and entail less aggregate unemployment. On the other hand, if the economy is hit by a large aggregate shock, then neighbors have the opposite effect and amplify the effects of an aggregate shock. However, as the paper does not provide a structural estimation of the labor market, we are left with something like a black box when it comes to understanding the type, and working of, the local interactions that are responsible for (sometimes) amplifying and (sometimes) muting the effect of shocks. IV. Concluding Remarks The authors are convincing in their argument that geographical factors are important for understanding aggregate business cycle dynamics. But they also leave a set of questions open. They show that the statistics of spatial correlation and dispersion follow certain patterns during most recessions. Moreover, they provide evidence suggesting that having neighboring counties may help or hurt depending on the magnitude of the recession. However, the reason why spatial statistics follow specific patterns, and why neighbors mat- 9121.indd 334 3/19/13 1:26 PM Comment 335 ter for business cycle dynamics, should indeed be the subject of future research. In their seminal paper, Blanchard and Katz (1992) investigated how US states adjusted after having been affected by an adverse shock to employment. They found that labor mobility across states served as the most important mechanism of adjustment and helped to smooth out shocks. In light of their results, the more ambiguous results on the impact of neighboring counties presented here definitely ask for further investigation. Is there, for instance, a differential impact of neighbors depending on geographical aggregation? In other words, do local transmission channels work differently between counties than between states? It is well known that US labor mobility has gradually declined over the last two decades, and has more recently almost come to a halt. This suggests that the impact and importance of various local geographical connections may have changed over time. This time variation is obviously also an issue that research into the transmission channels and their workings should account for. Fogli, Hill, and Perri have reminded us that geography matters, also for aggregate macro phenomena, and that stabilization policy can benefit from the insights of spatial dynamics. Endnote For acknowledgments, sources of research support, and disclosure of the author’s material financial relationships, if any, please see http://www.nber.org/chapters/c12790 .ack. References Blanchard, O. J., and L. F. Katz. 1992. “Regional Evolutions.” Brookings Papers on Economic Activity 1992:1–75. Kim, S. 1995. “Expansion of Markets and the Geographic Distribution of Economic Activities: The Trends in US Regional Manufacturing Structure, 1860– 1987.” Quarterly Journal of Economics 110:881–908. Krugman, P. R. 1991. Geography and Trade. Cambridge, MA: MIT Press. 9121.indd 335 3/19/13 1:26 PM