A Service of

Leibniz-Informationszentrum econstor Wirtschaft Leibniz Information Centre Make Your Publications Visible. zbw for Economics

Rose, Nancy L.

Article Industrial organization

NBER Reporter

Provided in Cooperation with: National Bureau of Economic Research (NBER), Cambridge, Mass.

Suggested Citation: Rose, Nancy L. (2012) : Industrial organization, NBER Reporter, National Bureau of Economic Research (NBER), Cambridge, MA, Iss. 4, pp. 1-6

This Version is available at: http://hdl.handle.net/10419/103232

Standard-Nutzungsbedingungen: Terms of use:

Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Documents in EconStor may be saved and copied for your Zwecken und zum Privatgebrauch gespeichert und kopiert werden. personal and scholarly purposes.

Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle You are not to copy documents for public or commercial Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich purposes, to exhibit the documents publicly, to make them machen, vertreiben oder anderweitig nutzen. publicly available on the internet, or to distribute or otherwise use the documents in public. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, If the documents have been made available under an Open gelten abweichend von diesen Nutzungsbedingungen die in der dort Content Licence (especially Creative Commons Licences), you genannten Lizenz gewährten Nutzungsrechte. may exercise further usage rights as specified in the indicated licence. www.econstor.eu NBER Reporter NATIONAL BUREAU OF ECONOMIC RESEARCH

Reporter OnLine at: www.nber.org/reporter 2012 Number 4

Program Report

IN THIS ISSUE Program Report Industrial Organization Industrial Organization 1

Research Summaries Nancy L. Rose* Cross Border Capital Flows 7 Kidney Exchange 10 … Productivity Differences 13 Bank Supervision … and Market Microstructure 16 The NBER’s Program on Industrial Organization (IO) begins its third decade with a core of 60 program members, including 15 whose NBER Profiles 19 primary affiliations are in another NBER Program. The Program’s meet- Conferences 21 ings attract submissions from a large and diverse set of researchers, and NBER News 27 are lively sessions with 75 to 90 scholars typically in attendance. The IO Program and Working Group Meetings 29 Program produces important applied research on a broad range of indus- Bureau Books 39 tries and topics, increasingly at the intersection with such other NBER Programs as Environmental and Energy Economics, Productivity, and Health Care. That commonality is recognized with frequent joint pro- gram meetings and contiguous Summer Institute sessions with other NBER groups. In 2012, IO Program members Aviv Nevo and Ariel Pakes delivered the annual Summer Institute Methods Lectures, focusing on the econometrics of demand estimation and related methodologies. This report describes work in just three of the Program’s areas: mod- eling consumer choice; the industrial organization of the digital econ- omy; and lessons for designing government auctions. Readers interested in exploring the broader range of NBER work in IO are encouraged to visit https://www.nber.org/papersbyprog/IO.html Consumer Choice

Empirical economists in the field of IO have devoted substantial attention to modeling the determinants of demand across a variety of set- tings. For some time, NBER researchers have been active in the design, innovation, and evaluation of methods to estimate demand based on neoclassical theories of consumer utility maximization. Nevo and Pakes discussed this in their 2012 Methods Lectures1 and dozens of NBER Working Papers have been published in this area.2 In recent years, empir-

*Rose directs the NBER’s Program on Industrial Organization and is a Professor of Economics at MIT. The numbers in parentheses throughout this report refer to NBER Working Papers.

NBER Reporter • 2012 Number 4 ical researchers increasingly have turned their attention to analyzing the underpinnings of NBER Reporter individual choice, for example characterizing the implications of deviations from standard neoclassical models of optimization behavior and the role of information in markets.

The National Bureau of Economic Research is a private, nonprofit research orga- nization founded in 1920 and devoted to objective quantitative analysis of the Consumer Behavior American economy. Its officers and board of directors are: The detailed microdata that are the main- President and Chief Executive Officer — James M. Poterba Controller — Kelly Horak stay of much empirical IO research have proved useful for identifying departures from con- BOARD OF DIRECTORS ventional models of consumer utility maxi- Chairman — Kathleen B. Cooper mization. A body of work in this area has Vice Chairman — Martin B. Zimmerman looked at automobile purchases, one of the Treasurer — Robert Mednick most significant consumer purchase decisions DIRECTORS AT LARGE for most households. Meghan Busse, Florian Peter Aldrich Mohamed El-Erian Michael H. Moskow Zettlemeyer, and co-author Duncan Simester Elizabeth E. Bailey Linda Ewing Alicia H. Munnell (13140) document consumer responses to John Herron Biggs Jacob A. Frenkel Robert T. Parry John S. Clarkeson Judith M. Gueron James M. Poterba “price cues” in the context of a Big Three Don R. Conlan Robert S. Hamada John S. Reed automaker “Employee Discount Pricing” pro- Kathleen B. Cooper Peter Blair Henry Marina v. N. Whitman motion in the summer of 2005. They find Charles H. Dallara Karen N. Horn Martin B. Zimmerman George C. Eads John Lipsky that consumers responded to this promotion Jessica P. Einhorn Laurence H. Meyer with unprecedented increases in new car pur- chases, even though prices during the promo- DIRECTORS BY UNIVERSITY APPOINTMENT tion were not substantially lower than imme- George Akerlof, California, Berkeley Mark Grinblatt, California, Los Angeles diately prior to it. Indeed, sales increased even Jagdish W. Bhagwati, Columbia Bruce Hansen, Wisconsin Timothy Bresnahan, Stanford Marjorie B. McElroy, Duke for some models with higher prices during Alan V. Deardorff, Michigan Joel Mokyr, Northwestern the promotion. While the researchers point Ray C. Fair, Yale Andrew Postlewaite, Pennsylvania out that this behavior can be consistent with Franklin Fisher, MIT Uwe E. Reinhardt, Princeton Edward Foster, Minnesota David B. Yoffie, Harvard rational reliance on (noisy) price signals, their John P. Gould, Chicago results are cautionary for those who would DIRECTORS BY APPOINTMENT OF OTHER ORGANIZATIONS model consumers as responding primarily to Christopher Carroll, American Statistical Association observed prices. In another paper on auto Jean Paul Chavas, Agricultural and Applied Economics Association purchases, Nicola Lacetera, Devin Pope, and Martin Gruber, American Finance Association Justin Sydnor (17030) look at heuristic infor- Ellen Hughes-Cromwick, National Association for Business Economics Thea Lee, American Federation of Labor and mation processing in used car purchases. They Congress of Industrial Organizations find that sale prices drop discontinuously at William W. Lewis, Committee for Economic Development exactly 10,000 mile odometer readings, con- Robert Mednick, American Institute of Certified Public Accountants Alan L. Olmstead, Economic History Association sistent with customers focusing on the left- Peter L. Rousseau, American Economic Association most digit of the odometer reading rather than Gregor W. Smith, Canadian Economics Association incorporating the full odometer reading into Bart van Ark, The Conference Board their valuation. They estimate $2.4 billion of The NBER depends on funding from individuals, corporations, and private foun- mispricing as a result. Busse and Pope and dations to maintain its independence and its flexibility in choosing its research their co-authors (18212) use a sample of 40 activities. Inquiries concerning contributions may be addressed to James M. Poterba, President & CEO, NBER 1050 Massachusetts Avenue, Cambridge, MA million vehicle purchases and 4 million house 02138-5398. All contributions to the NBER are tax deductible. purchases to explore the role of projection The Reporter is issued for informational purposes and has not been reviewed by bias — the tendency to over-predict the degree the Board of Directors of the NBER. It is not copyrighted and can be freely repro- to which one’s future tastes will resemble one’s duced with appropriate attribution of source. Please provide the NBER’s Public current tastes — in purchasing behavior. They Information Department with copies of anything reproduced. find that weather at the time of purchase overly Requests for subscriptions, changes of address, and cancellations should be sent influences purchase decisions for these major to Reporter, National Bureau of Economic Research, Inc., 1050 Massachusetts Avenue, Cambridge, MA 02138-5398. Please include the current mailing label. durables. They meticulously explore alternative explanations for this finding, and their results

2 NBER Reporter • 2012 Number 4 rule out explanations grounded in neo- ing body of empirical work exploring mal search behavior, and are indicative classical utility maximization. For exam- how imperfect information affects mar- of buyer confusion, not only on how to ple, spring or fall days that are unusually kets. In one paper, Ginger Jin and David assess complex menus of broker charges warm and sunny induce additional con- Dranove (15644) review the theoretical but even about the potential benefits of vertible sales, which are not merely time- and empirical research on product qual- search among brokers. They conclude shifted. Moreover, the convertibles pur- ity disclosure and certification. Jin and that current disclosure policies have chased on such days are more likely to be her collaborators (14252) also explore done little to mitigate that confusion. traded in quickly, consistent with mis­ how information about the properties estimating future tastes. of a new pharmaceutical is diffused to Industrial Organization Justine Hastings and Jesse Shapiro doctors and patients, and they consider of the Digital Economy (18248) analyze “mental account- a range of information sources includ- ing” in household purchases of gaso- ing academic articles, advertising, media The digital economy has exploded in line. Their results consistently reject reports, FDA updates, and individual the two decades since the IO Program’s the null hypothesis that households patient experiences. How information January 2001 conference on e-com- treat spending on gasoline as fungible is presented, in addition to its content, merce,3 along with economic research with other income. Instead, when gas- can have substantial impacts on con- on its characteristics and the implica- oline prices rise, consumers dispropor- sumer responses: Hastings and various tions for firm strategies and traditional tionately substitute to (less expensive) collaborators have shown this in the con- retail markets. Jonathan Levin (16852) lower octane gasoline, far more than text of workers’ choice among pension examines the literature in this area and the substitution that occurs for simi- investment options in Mexico’s priva- describes the economic implications lar income effects from non-gasoline tized social security system (14538) and of key features of the digital economy: price sources; the converse is true when in parents’ decisions on public school an unusual combination of substantial gas prices rise. This complements work choice programs (12995). Similarly, economies of scale with customer per- that Hastings has done with other col- Phillip Leslie and Alan Sorensen’s work sonalization; the ability to collect large laborators (13614) on how households with Bryan Bollinger (15648) on posting volumes of detailed data about custom- adjust grocery purchases when gasoline calorie counts in restaurant chains dem- ers, their behavior, and preferences; and prices change. Andrei Shleifer and his onstrates the importance of how infor- the rapid pace of innovation facilitated collaborators (17947) develop a model mation is presented to consumers. by seller experimentation. He notes the of context-dependent consumer choice In many markets, search does not critical role of economic theory in the focused on “salient attributes” that is appear to be effective in matching con- design and analysis of these markets, and consistent with this mental accounting sumers to the lowest-price or highest the platform that these markets offer for behavior, and use their model to study match quality product. Glenn Ellison’s empirical research on the digital mar- discounts in a variety of settings. research with Alexander Wolitzky ketplace and as a setting in which to test Better understanding of consumer (15237) argues that this may in part models of imperfect competition. Below choice is an important input to modeling reflect actions taken by firms to impede are results from just three strands of IO firm decisions. Julio Rotemberg (13754) effective search. In their model of research that explore these and other models the implications for firms and “obfuscation,” firms selling homoge- themes in online markets: the design of policymakers of consumers who do not neous goods find it individually ratio- online strategies by firms; mechanisms to make effective use of price information, nal to invest in actions that make it address asymmetric information about and then suffer ex post regret or anger more difficult for consumers to learn online seller quality; and the implica- as a consequence. Hastings and Shapiro about their product and full product tions of digital distribution for produc- (18248) argue that supplier response to price, because this reduces competi- ers and “bricks and mortar” retailers. 4 the consumer octane adjustment behav- tion across firms and sustains higher ior they find in gasoline markets may mark-ups. Bruce Carlin and coauthor Designing Online Strategies contribute to an observed inverse rela- Florian Ederer (17895) model oligop- tionship between gasoline prices and olists’ product proliferation responses As Levin notes, digital marketplaces retailer markups over time. to the possibility of consumer search offer new challenges and new possibil- fatigue, the notion that search is not ities for firms. David Reiley and his The Role of Imperfect only costly but also tiring, potentially collaborators analyze online auctions Information leading consumers to break from search- in their chapter for the Handbook of ing in some periods. Robert Hall and Economics and Information Systems The rich theoretical literature on Susan Woodward (16007) argue that (12785), focusing on the theory, exper- markets with imperfect information mortgage broker decisions by borrowers imental research, and empirical analysis recently has been married to an increas- suggest substantial deviation from opti- of online retail auctions such as eBay.5

NBER Reporter • 2012 Number 4 3 This work describes the responsiveness metric information on seller quality, and tive borrowers. Iyer et al. find that lend- of bidder strategies to seller strategies, has provided new tools for its empir- ers respond to coarse information in the and its implication for optimal design of ical investigation. While online mar- Prosper.com profiles in order to infer online auction markets. It also addresses kets may reduce search costs and offer much of the information that would endogenous innovation in markets such greater apparent pricing transparency, have been accessible from (unreported) as eBay, highlighting the importance of their heightened anonymity of exchange individual-level credit scores. Freedman considering dynamic implications of auc- exacerbates the problem of asymmetric and Jin report that lenders on Prosper. tion design for the viability of platforms. information between buyers and sellers, com generally underestimate the credit One of the most active online mar- particularly with respect to seller quality risk of borrowers, but learn significantly kets involves “position auctions” which or trustworthiness. Seller reputation can from their own experiences on the site. are conducted by search engines such as mitigate asymmetric information, and Newer cohorts of lenders underestimate Google, Bing, or Yahoo to allocate to often is established online through buyer less, suggesting some diffusion of learn- advertisers the “sponsored link” positions feedback mechanisms, such as eBay’s ing across cohorts. Of particular interest on a search response page. Susan Athey well-studied feedback system. Third- is the convergence the researchers note and Ellison (15253) emphasize the two- party certification provides an alternative between online and more traditional sided market aspect of these auctions. to feedback or reputation mechanisms. offline sources of credit: as lower quality Bidders (advertisers) care about how con- Jin and her co-authors (17955) study subprime borrowers have been increas- sumers respond to advertising links, and the effectiveness of certification authori- ingly excluded from funding on Prosper. those responses in turn are affected by the ties for online pharmacies, used by many com, the site has competed more directly mechanism that sellers (search engines) consumers to reduce drug acquisition with conventional lending institutions use to allocate advertisers to positions. costs. For four of the five popular brand- such as banks. Competition between Enriching the analysis to include con- name drugs they ordered from online online and offline outlets is also the sub- sumer search behavior yields a number of pharmacies, drugs labeled as branded ject of other work by Freedman and Jin insights not present in conventional auc- were authentic versions for all deliver- on peer-to-peer lending, and is the topic tion models, such as the benefits of high ing pharmacies, whether certified or not, of a broad research agenda by other reserve prices to exclude lower match but prices at certified U.S. pharmacies NBER researchers, to which I turn next. quality ads and the informational ineffi- were roughly 50 percent higher than ciencies that can be induced by weighting were prices at non-U.S. certified pharma- Interactions with bids by customer click-through rates. cies. This suggests considerable cost to Offline Markets In online markets, experimentation consumers from complying with FDA is facilitated and rewarded. Search engine warnings to avoid all foreign websites, The rapid growth of the internet- firms rely on substantial experimentation, perhaps without concomitant consumer based economy over the past 15 years in addition to economy theory, to enhance benefit. For the fifth drug—Viagra—cer- has dramatic implications for both pro- profits through the design of their auc- tified pharmacy prices and quality were ducers and “bricks and mortar” retail- tions. Liran Einav, Levin, and their col- roughly identical regardless of country, ers. Early research in this area focused laborators (17385) document the activity while uncertified pharmacies offered on pricing impacts of online search and of eBay sellers to improve their strategies both a lower price and a lower proba- e-commerce. More recent research has through both active and passive experi- bility of receiving an authentic branded highlighted the impact of the internet mentation. The ease of experimentation product. This suggests a potential advan- on the allocation of sales across retailers, online is a boon to researchers as well. tage to buying only from certified web- and the entry/exit decisions of firms, and Reiley et al. (12785) and Levin (16852) sites; the authors’ online survey of over product choice decisions by producers. describe a number of academic studies 2500 consumers suggests that is what Einav, Levin, and their co-authors that have taken advantage of online plat- more than 40 percent of consumers who (18018) explore the impact of sales forms to construct field experiments to purchase drugs online do. Jin and collab- taxes on consumers’ choices of online investigate consumer behavior, pricing orators study the role of price signals and retailers, which is of considerable policy strategies, advertising effectiveness, and regulation across international pharma- interest. Their analysis of eBay customer the implications of auction design, some ceutical markets (16854 and 18073). responses to sales taxes suggests consid- of which are described below. Erzo Luttmer, Asim Khwaja, erable sensitivity: a single percentage Rajkamal Iyer, and Kelly Shue (15242), as point increase in a customer’s home state Asymmetric Information on the well as Jin and co-author Seth Freedman sales tax implies an increase of nearly Internet (16855), use data from the online peer- 2 percent in online purchasing from to-peer lending platform Prosper.com other states, and a decline of roughly 3 The growth in online markets has to investigate the ability of lenders to to 4 percent in online purchasing from elevated interest in the effect of asym- screen the creditworthiness of prospec- home state sellers. The authors also note

4 NBER Reporter • 2012 Number 4 increased density of sellers on the low- one to post a review, to those on Expedia. explore whether reduced album revenues tax side of state borders. com, which restrict reviews to consum- have led to reductions in the production Ali Hortaçsu and Chad Syverson ers who have made a booking at that of new albums, and he concludes that and their collaborators (14166) 6 exam- hotel through Expedia. They find that there is no discernible decline in quan- ine impacts on physical retailers for three on TripAdvisor.com, small independent tity or quality post-Napster. That is con- of the sectors they expect to be most hotels have more five-star reviews, and sistent with lower costs of bringing new affected by the internet: auto dealers, their neighboring hotels have more one- works to market and growth of indepen- bookstores, and travel agencies. They and two-star reviews, consistent with dent labels. Finally, Leslie and Sorensen find that online shopping has shifted the their predictions of ex ante incentives for find that the expansion of ticket resale distribution of revenues among phys- review manipulation. While these results markets for major rock concerts, facili- ical bookstore and auto dealers from suggest that review manipulation may be tated by online resale sites, improves the small retailers to larger retailers, and economically significant, the authors note allocation of tickets to high valuation that smaller retailers disproportionately that the overall level of manipulation is buyers. However, half of the gains are dis- have exited as the fraction of consum- relatively low, ensuring that the platform sipated through higher transactions costs, ers using online shopping increased. still communicates useful content. so resale buyers end up with little of the Travel agencies experienced the same The music and video industries have potential surplus (15476). reallocation away from small outlets, but been among those argued to be most for that industry the trend appeared to affected by the internet, in large part Designing Government Auctions be national, a function of changes in because of the producers’ greater dif- airlines’ distribution systems and not ficulty in enforcing intellectual prop- Governments and quasi-public dependent on local consumer online erty protection of their content online. agencies use auctions in a wide variety shopping patterns. In principle, unauthorized distribu- of settings, including: competitive pro- Igal Hendel, Nevo, and co-author tion of online content may have both curement; sales or leases of publicly- Francois Ortalo-Magne (13360) compare demand contraction effects (by substitut- owned assets, such as mineral and timber the impact on home sellers of using con- ing for purchased content) and demand rights on public land and spectrum allo- ventional versus online sales outlets in expansion effects (by increasing poten- cation; wholesale electricity purchases a study of the 2004 housing market in tial consumer awareness of the product, and sales; and the allocation of pollu- Madison, Wisconsin. They find no sales effectively advertising). Joel Waldfogel tion permits under some cap and trade price difference across houses sold through (13497) explores these twin effects on programs. Economic theory has made traditional realtors using the Multiple television viewing in a study of unau- fundamental contributions to the design Listing Service (MLS) and those sold by thorized (primarily YouTube) and autho- of many of these auctions, and empirical owners using the online FSBO.com web- rized (primarily network) web distribu- research has contributed to evaluation site. However, houses on the MLS are tion of television shows. Using a survey of their operation and guiding improve- both more likely to sell and are quicker to of university students, he finds that inter- ments in their execution. NBER research sell, conditional on a transaction, which net access induces a modest substitu- has played a role in both fronts, and is consistent with improved matching on tion away from traditional television the NBER Working Group on Market the still-larger MLS network. which is more than offset by a strong Design, led by Athey and Parag Pathak, The internet facilitates search not demand expansion effect: overall time focuses on these and related issues. only on price but also on provider and on network-controlled sites (television Since 1994, Federal Communication product quality. And, online review sys- and network websites) increases by 1.5 Commission spectrum auctions have tems allow consumers to register feedback hours/week. Julie Mortimer, Sorensen, been used to allocate billions of dol- on physical outlets. Jin and her collabo- and co-author Chris Nosko (16507) find lars in spectrum rights. Patrick Bajari rators (18567) analyze restaurant ratings that musical artists have reacted to the and his co-author Jungwon Yeo (14441) on Yelp.com, focusing on the optimal decline in album sales that is associated describe how FCC auction design has way for a review site such as Yelp to con- with unauthorized file-sharing by increas- evolved over time to mitigate concerns struct aggregate ratings from individual ing their live performances. Less well- about tacit collusion by bidders. The feedback. The usefulness of user feed- known or popular musicians among the researchers examine patterns in the bid- back depends on its credibility, though, more than 1800 artists they study expe- ding data from four large spectrum auc- and assessing credibility can be difficult. rience significant increases in concert tions and conclude that later auctions Judith Chevalier, Dina Mayzlin, and revenue in the post-Napster era, in part do, in fact, exhibit fewer examples of Yaniv Dover (18340) investigate the inci- offsetting the lost album revenues, and strategies most likely associated with dence of review manipulation by com- perhaps reflecting greater awareness of potential collusion. Their analysis gives paring the distribution of hotel reviews their music by potential fans. Waldfogel a flavor of the considerable complex- on Tripadvisor.com, which allows any- (16882) assembles a novel dataset to ity that is involved in bidding in spec-

NBER Reporter • 2012 Number 4 5 trum auctions. Jeremy Bulow, Levin, owned businesses, in procurement or Minnesota highway projects to explore and co-author Paul Milgrom (14765) natural resource auctions. Their analy- empirically the role of adaptation and describe the potential for economic and sis shows that restricting entry to small delay, and illustrate the impact of alter- game-theoretic modeling to help bid- businesses is associated with significant native incentive structures on outcomes. ders devise successful strategies in the revenue and efficiency costs; replacing face of that complexity. They illustrate the restriction with a bidder subsidy such potential by demonstrating how a would increase revenue, efficiency, and 1 Video and slides can be viewed at www. new entrant used it in the 2006 90MHz the profit of small bidders, with minimal nber.org/econometrics_minicourse_2012/ auction, which contributed to the firm’s impact on large firm profitability. 2 See for example, recent contributions success in purchasing nationwide spec- Highway construction procurement by program members Patrick Bajari, trum coverage at one-third the price contracts are a significant state level Steven Berry, Jean-Pierre Dube, Jeremy paid by incumbents, thus saving more activity, imposing substantial direct costs Fox, Philip Haile, Christopher Knittel, than a billion dollars. to finance road construction and repair, Julie Mortimer, Aviv Nevo, and Stephen Much of the empirical work in and substantial indirect costs on drivers Ryan, among others. government auctions done by NBER who are subject to delays and longer 3 Published as a special issue of the researchers has focused on U.S. Forest commutes while construction projects Journal of Industrial Economics, Vol. Service timber auctions, which can gen- are underway. Bajari and Greg Lewis 49, no. 4, December 2001. erate more than one billion dollars annu- have developed a research agenda that 4 Many IO program members are ally. James Roberts and Andrew Sweeting investigates how to design procurement active in the Productivity Program’s (17624) focus on when sellers should use contracts to more effectively align the new Economics of Digitization and auctions, comparing the expected rela- incentives of contractors with those of Copyright Initiative, which brings tive performance of a simultaneous bid the highway department and drivers. In together researchers from a diverse set of auction to a setting in which sellers invite one paper (14855), they evaluate scoring fields to study this sector of the economy. buyers to make offers sequentially. Athey auctions used by the California This report focuses on work in this area and Levin work with various collabora- Department of Transportation to pro- by members of the IO program; addi- tors to analyze the design and operation vide explicit time-to-completion incen- tional working papers can be found on of timber auctions. In one paper (14590) tives in contract awards. They estimate the NBER’s website. they compare performance under two substantial welfare gains from the incen- 5 In Handbook of Economics different auction formats: sealed bid auc- tive contracts, although direct outlays by and Information Systems, Terrence tions, which attract more small bidders, the Department of Transportation also Hendershott, ed., Elsevier Science, 2006. and open outcry auctions. Their cali- increase through their effect on the win- 6 Published as Maris Goldmanis, brated model suggests that sealed bid ning bid. Their model suggests even Ali Hortaçsu, Chad Syverson, and auctions generate greater expected reve- larger potential gains from an optimally Önsel Emre, “E-Commerce and the nue for the Forest Service, and it focuses designed policy. In a more recent paper Market Structure of Retail Industries,” attention on bidder competitiveness as (17647), they develop a model of con- NBER Working Paper No. 14166, July a critical choice in auction format. In a tractor adaptation to productivity 2008, and Economic Journal, Royal more recent paper (16851) they turn to shocks, incorporating time incentives in Economic Society, vol. 120(545), the set-asides and subsidies that the gov- an optimal contract design. They com- (2010), pp. 651–82. ernment frequently uses for preferenced bine this with day-level information on bidders, most often small or minority- work plans, progress, and delays for

6 NBER Reporter • 2012 Number 4 Research Summaries

Cross-Border Capital Flows, Fluctuations and Growth

Sebnem Kalemli-Ozcan *

What is the extent of international for the amount and direction of capital Our results show that in a sample of financial integration, and how does such flows and their implications for economic developed and developing countries, the integration affect economic fluctuations fluctuations and growth. The appropriate positive correlation between capital flows and growth? Does the effect differ dur- definition (FDI versus debt, public versus and GDP per capita (that is, the Lucas ing tranquil times versus times of finan- private, or net versus gross flows) must be paradox) during 1970–2000 goes away cial crisis? Does financial integration used depending on the question asked. once we account for the effect of insti- transmit shocks across the globe and For example, the neoclassical model tutional quality: rich countries receive lead to contagion? In recent research, predicts a large amount of capital flows more foreign investment because they together with my co-authors, I search for based on return differentials from capital- have better institutions. Exogenous vari- answers to these and other related ques- abundant rich countries to capital-scarce ation in institutional quality, measured tions, using both macro-level country poor ones. The lack of such flows in the by the historical determinants of institu- data and micro-level firm data. Figure 1: Net Capital Flows (% of GDP) by Type of Flow, Capital Flows: Where and Why? 1970-2006. All countries

One common definition of interna- 10 tional financial integration is the amount of cross-border capital flows. These flows 8 can take the form of foreign direct, port- 6 folio equity, and debt investment, consti- 4 tuting the financial account — the mirror image of current account in the balance- 2 of-payments statistics. Figure 1 plots the 0 average current account balance with reverse sign as a measure of total net capi- -2 tal flows from more than 150 countries, -4 together with different types of flows.1 -6 The black dashed line shows that the 1970 1974 1978 1982 1986 1990 1994 1998 2002 2006 world is running a current account deficit, roughly around 4 percent of GDP, imply- Net Debt Flows (Mean) Net FDI + Port. Flows (Mean) ing positive net capital flows on average Total Net Flows (Mean) since the 1970s.2 Since the 1990s, how- ever, countries seem to be net borrowers data is known as the Lucas paradox. The tions, is the most important determinant in FDI and equity investment and net recent period of global imbalances has of capital flows, causally explaining the lenders in debt instruments.3 This sim- seen a related paradox, where capital flows Lucas Paradox.5 ple plot hints that current account may in the reverse direction (when measured If capital is flowing to produc- not be informative in terms of testing the from current account), from “still poor tive places in the long run, where long- predictions of certain classes of models but growing fast” countries such as China run productivity is proxied by institu- * Kalemli-Ozcan is an NBER Research to “rich but not growing” countries such tional quality, then why do we worry Associate and a Professor of Economics at as the United States. Laura Alfaro, Vadym about capital flows from China to the the University of Maryland. Her Profile Volosovych, and I have investigated the United States, where the latter clearly has appears later in this issue. reasons for both of these phenomena.4 higher quality institutions? We worry

NBER Reporter • 2012 Number 4 7 because the standard models imply that might bring instability, especially short- put synchronization to banking integra- China must have received more capital term debt flows intermediated via banks. tion before the 2008 crisis we find a signif- flows than, say Zimbabwe, in a sample In fact, since the 2008 crisis, global banks icantly negative association. This implies of developing countries, given China’s have been seen as the “bad guys” who that in tranquil times, increases in bank- faster catch-up productivity growth to spread the crisis from the United States to ing integration within country-pairs over the United States. This does not seem to other financial systems. time are associated with more divergent be the case. The fast growing countries Does financial integration transmit output cycles. The negative association accumulate a large amount of reserves shocks? Academic research has not been between bilateral financial linkages and and export capital to slow growing coun- helpful in answering this question, because business cycle co-movement is in line tries, causing global imbalances. the theory suggests that financial integra- with the standard models summarized In our 2011 paper, we decompose tion may lead to a higher level of business above, where in the absence of financial international capital flows into public cycle synchronization but can also cause a shocks, banking integration causes diver- and private components (private debt, “decoupling” of business cycles between gence in output cycles. Yet, this negative FDI, equity flows versus public flows). inter-connected economies. The key issue correlation between financial integration We focus on a sample of developing coun- seems to be the nature of shocks.6 and output synchronization turned posi- tries since the 1980s and measure the Take for example the case of two inte- tive during 2008–10. This result is again amount of private and public capital flows grated economies, where one is hit by a in line with the above models showing these countries have received in relation negative shock. If the shock hits the bank- that during crisis, financial integration to their catch-up growth. It turns out that ing sector, then problems in one country facilitates co-movement via contagion. upstream flows and global imbalances will likely spread to the other, as banks These findings bridge two bodies are manifestations of the same under- operating in both countries pull funds of research in international macroeco- lying phenomenon: sovereign-to-sover- from the non-affected country, mak- nomics and finance on the implications eign flows in the form of government ing the business cycles of the interlinked of financial integration: one looks at its debt from official lenders, reserve accu- countries more synchronized. If, how- effect on international business cycles mulation, and aid. International capital ever, the negative shock hits the produc- and another focuses on financial con- flows net of government debt and/or tivity of firms in a country while banks tagion. The results imply that conduct aid are positively correlated with growth. stay healthy, then return to capital falls of monetary policy becomes significantly Government debt flows are negatively and banks pull funds out of the affected harder within financially integrated cur- correlated with growth only if govern- country, amplifying the initial shock; this rency areas. Financial integration magni- ment debt is financed by another sover- in turn makes the business cycles of finan- fies output fluctuations across countries, eign and not by private lenders, where cially interconnected economies diverge. thus making it difficult to conduct mon- government debt from private lenders In recent work, Elias Papaioannou, etary policy in all regions. This problem also flows in the right direction. Jose-Luis Peydro, Fabrizio Perri, and I is clearly illustrated nowadays in the euro As we document in these works, there examine the role of banking integra- area.8 The high degree of integration has is much more nuance to the direction of tion on business cycle synchronization amplified country-specific shocks, lead- capital flows than is commonly appre- since the 1970s for the OECD countries, ing to divergence in economic activity ciated. Standard model’s predictions explicitly allowing for the possibility that between countries of the south and the are best tested by using private capital its effect may differ in tranquil versus core; and, as global banks pull capital out flows, because government is outside this financial crisis times.7 Our results provide of the periphery, the low policy rate of the model. In relation to figure 1, China had first-time evidence for the above theory. ECB is not channeled to the south. a current account surplus and was a net We document that across country- lender overall. But this is because China pairs, there is a significant positive correla- Do Capital Flows Bring is a net lender in what we call “sovereign- tion between cross-border banking expo- Growth and Welfare? to-sovereign” flows. During the same sures and output synchronization. This period, China was a net borrower in should come as no surprise. The U.S. busi- The textbook case for supporting terms of private flows, as it should be ness cycles are both more synchronized international financial integration is well according to the neoclassical model. and more financially linked with Canada known. In spite of possible contagion than with France. There can be many rea- during major crisis, integration ultimately Do Capital Flows Transmit sons for such a result, including socioeco- will bring growth and welfare. The stan- Shocks? Tranquil Times nomic ties and less distance between cer- dard model implies that when capital versus Crisis Times tain pairs than others. flows from low return to high return In sharp contrast to the positive countries, the cost of borrowing will go Private capital flows go in the right cross-sectional correlation, in examining down, boosting investment and growth. direction to productive countries but still the within country-pair response of out- FDI will bring better technology and

8 NBER Reporter • 2012 Number 4 know-how, together with financial stabil- shareholder/subsidiary links with sector investment and initial productivity are ity, because FDI tends to be long-term, and nationality of the investor. important for the realization of positive enhancing growth and welfare. Selection is a problem. Multi­ effects of FDI on productivity at the dis- Given the large increase in FDI and nationals are likely to buy local firms aggregated level. portfolio equity flows (Figure 1) during with high future growth potential. the last decade, it is important to study the Foreign firms may select themselves effect of such flows on both growth and into high productivity sectors and drive 1 Notes: The data is from IMF, IFS. Net volatility during tranquil times, for the weak domestic firms out of business. In capital flows represent average net flows of developed and emerging countries. Using such a case, domestic firms in the for- FDI and portfolio equity investment, and country-level data at the macro level is eign activity sector might become more debt (portfolio debt investment and other not appropriate for such a study, because productive on average, but not because investment) divided by nominal GDP in many policy changes occur simultane- any single firm has become productive. current dollars, based on WB and IMF ously with financial integration, growth, Since our data encompasses many coun- data and corresponding to the sum of the and volatility. That makes the identifica- tries and sectors, we can control selec- flows of assets (outflows) and liabilities tion of the individual effects harder. tion through the use of firm and sector- (inflows), because assets have a minus sign In joint work, Bent Sorensen, year effects. By exploiting the difference as BOP convention. Total capital flows are Volosovych, and I use a novel dataset between financial and industry invest- represented by the negative of total current of firm-level balance sheets and foreign ment as exogenous variation, we control account flows. The data for current account investment from 25 European countries dynamic selection at the firm level. is available for 186 countries; the data for for the period 1996–2006 (AMADEUS) Our results show that the positive cor- FDI and portfolio flows and debt flows is to study the effect of foreign direct invest- relation between foreign investment and available for 179 and 178 countries respec- ment and portfolio equity investment productivity growth in developed coun- tively, varying across years. FDI assets and on output volatility.9 Our dataset differs tries is driven by selection. Multinationals liabilities correspond respectively to Direct starkly from Compustat or Worldscope- target more productive companies but Investment Abroad (line 78bdd) and type data, because in our data 99 percent do not contribute to further increases. Direct Investment in Reporting Economy of the firms are privately held. We docu- In emerging markets, there is evidence (line 78bed). They include equity capital, ment a positive, highly robust, relation- of productivity enhancing effects of mul- reinvested earnings, other capital, and ship between firm-level foreign invest- tinationals on targets but the effects are financial derivatives associated with vari- ment and output (value added) volatility small. For domestic firms, in both devel- ous intercompany transactions between in cross-section and over time, both for oped and emerging countries there are affiliated enterprises. Portfolio Equity emerging and developed Europe. negative spillover effects from direct com- Investment assets and liabilities correspond One theory that can explain this styl- petitors. Only domestic firms with high to Equity Securities Assets (line 78bkd) and ized fact is as follows: foreign investment initial productivity and suppliers of for- Equity Securities Liabilities (line 78bmd). brings technology and know-how to the eign owned firms benefit from knowledge They include shares, stock participations, target firm and improves diversification spillovers. The effects are too small to con- and similar documents that usually denote of ownership of capital. After receiving tribute to aggregate productivity. ownership of equity. Debt assets and investments from multinationals, target liabilities include Debt Security Assets (line firms might choose to invest in risky but Summary 78bld) and Debt Security Liabilities (line high return projects that jointly increase 78bnd), which include bonds and money their productivity and output volatil- Overall, my research shows that the market or negotiable debt instruments; ity.10 In joint work, Chris Fons-Rosen, neoclassical model is alive and well in Other Investment Assets (line 78bhd); and Sorensen, Volosovych, Carolina Villegas- terms of predicting where and why capi- Other Investment Liabilities (line 78bid), Sanchez, and I ask whether firms that tal should flow if we measure capital flows which include all financial transactions receive foreign investment are becom- as private flows (FDI, portfolio equity not covered by direct investment, portfolio ing more productive and whether they and private debt). When it comes to the investment, financial derivatives, or other create spillovers for domestic firms.11 effects of capital flows in terms of further assets. The current account total corre- Notice that spillover effects from for- increases in productivity and knowledge sponds to the Current Account excluding eign to domestic firms are essential to spillovers, these are harder to find. Most Exceptional Financing (line 78ald). realizing any aggregate effects of foreign of the positive correlations between for- 2 This pattern of “borrowing from space” is investment, and they are the main reason eign investment and growth at the aggre- related to the fact that countries’ liabilities behind the big policy push for FDI over gate level initially are explained by the fact are better measured relative to their assets. last two decades. We extend our firm-level that foreign capital is attracted to high 3 If we divide the sample between rich and data to 60 countries worldwide (ORBIS), productivity firms, sectors, and countries. poor countries — where “rich” is defined as where we have more than 30 million Firm heterogeneity in terms of foreign GDP per capita higher than 15,000 USD

NBER Reporter • 2012 Number 4 9 in 2000 dollars on average throughout the Review, 91, (2001), pp. 1369–1401. Journal of International Economics. period — we see that poor countries receive 6 Most theoretical works focus on one 8 See R. Mundell, “A Theory of Optimum more capital flows than rich, with a 5 type of shock, that is, either tranquil times Currency Areas,” American Economic percent current account deficit relative to a or financial crises. For theoretical models Review, 51, 4, (1961), pp.657–65, for the zero balance for the rich, on average. But where both mechanisms are in place, see classical treatment of the issue. this is not true when we look at FDI and B. Holmsrom and J. Tirole, “Financial 9 S. Kalemli-Ozcan, B. Sorensen, and V. equity investment, where poor countries Intermediation, Loanable Funds, and Volosovych, “Deep Financial Integration receive less. Gross flows (sum of assets and the Real Sector,” Quarterly Journal of and Volatility,” NBER Working Paper No. liabilities) amount to 150 percent of GDP Economics, 112, 3, (1997), pp. 660–91; 15900, April 2010. for rich countries and 25 percent of GDP F. Perri and V. Q uadrini, “International 10 This argument is detailed in for poor countries, on average. Recessions,” NBER Working Paper No. M. Obstfeld, “Risk-Taking, Global 4 L. Alfaro, S. Kalemli-Ozcan, and V. 17201, July 2011, and S. Kalemli-Ozcan, Diversification and Growth,” American Volosovych, “Why does not Capital Flow E. Papaioannou, and F. Perri , “Global Economic Review, 84, 5, (1994), pp. from Rich to Poor Countries? An Empirical Banks and Crisis Transmission,” NBER 1310–29, and D. Acemoglu and F. Investigation,” NBER Working Paper Working Paper No. 18209, July 2012, Zilibotti, “Was Prometheus Unbound No. 11901, December 2005, published forthcoming in Journal of International by Chance? Risk, Diversification, and in Review of Economics and Statistics, Economics. Growth,” Journal of Political Economy, 90, 2008, pp. 347–68, and “Sovereigns, 7 S. Kalemli-Ozcan, E. Papaioannou, 105, 4, (1997), pp. 709–51, who show Upstream Capital Flows, and Global and J. Peydro , “Financial Globalization, how diversified capital ownership allows Imbalances,” NBER Working Paper No. Financial Regulation, and the firms to choose riskier projects leading to 17396, September 2011. Synchronization of Economic Activity,” higher growth and volatility. 5 We used settler mortality rates to account NBER Working Paper No. 14887, April 11 C. Fons-Rosen, S. Kalemli-Ozcan, B. for endogenous institutions, based on 2009, forthcoming in Journal of Finance, Sorensen, V. Volosovych, and C. Villegas- work by D. Acemoglu, S. Johnson, and 2012; S. Kalemli-Ozcan, E. Papaioannou, Sanchez, “Q uantifying Productivity Gains J. Robinson in “The Colonial Origins of and F. Perri, “Global Banks and Crisis from Foreign Investment,” forthcoming as Comparative Development: An Empirical Transmission,” NBER Working Paper No. an NBER Working Paper. Investigation,” American Economic 18209, July 2012, and forthcoming in

Kidney Exchange

Alvin E. Roth*

More than 90,000 patients are on the healthy with only one, which also makes kidney exchange in the United States was U.S. waiting list for a kidney transplant it possible to receive a kidney from a living performed at the Rhode Island Hospital from a deceased donor, and only 11,000 donor — around 6,000 such transplants in 2000, when doctors there noticed two or so such transplants are accomplished were accomplished in 2011. Nevertheless, incompatible patient-donor pairs who each year. So, the waiting is long and someone who is healthy enough to donate could benefit from exchange. Shortly after costly, sometime fatally so. But healthy a kidney may be unable to donate to his or that, Tayfun Sonmez, Utku Unver, and people have two kidneys and can remain her intended recipient because of various I proposed a way to organize a multi- types of donor-recipient incompatibility. hospital kidney exchange clearinghouse1, * Roth is a Research Associate in the This is the origin of kidney exchange. and began discussions with Dr. Frank NBER’s Program on Labor Studies and a In the simplest case, two incompatible Delmonico of Harvard Medical School, visiting professor of economics at Stanford patient-donor pairs exchange kidneys, that soon led to the founding of the New University. His Profile appears later in this with each patient receiving a compatible England Program for Kidney Exchange.2 issue. kidney from the other’s donor. The first Together with Itai Ashlagi, we have since

10 NBER Reporter • 2012 Number 4 assisted in the formation and operation of cycle” algorithm for trading indivisible reported in that AJT paper was performed other kidney exchange networks operat- goods without money and showed that simultaneously (and hence involved six ing around the country. it produced an allocation in the core5, operating rooms and surgical teams), the In the United States and most of the and Roth (1982)6 showed that the top paper also proposed that chains that begin world it is illegal to buy or sell organs for trading cycle algorithm made it a domi- with a non-directed donor might not transplant.3 As Jevons (1876)4 noted, one nant strategy for traders to reveal their need to be performed simultaneously. The obstacle to two-way barter exchange is true preferences. Abdulkadiroglu and argument was a simple cost-benefit analy- the need to find a counterparty who has Sonmez (1999)7 extended this model sis. The reason that cyclic exchanges are what you want and also wants what you to deal with assignment of dormitory performed simultaneously is that if they have. One way to reduce the difficulty rooms when some students already had were not, some patient-donor pair would of finding these double coincidences is rooms, some did not, and some rooms have to give a kidney before getting one, to assemble a large database of interested might be vacant, so that assignment and if the cycle were to be broken subse- patient-donor pairs. Another is to con- would involve chains as well as cycles. quently, that pair would suffer a grievous sider a larger variety of exchanges than We observed that the efficient chains loss. The donor in the pair would have those between just two pairs: for example, and cycles in kidney exchange mostly undergone a nephrectomy that yielded a cycle of exchange among three pairs, or would be short but occasionally would no benefit to the recipient in the pair, and a chain that begins with a donation by a be long, which presented a logistical there would no longer be a kidney with non-directed donor (such as a deceased problem, since, for incentive reasons, which to participate in a future exchange. donor, or an altruistic living donor) to the all surgeries in a given exchange would Now consider a chain that begins patient in an incompatible patient donor be performed simultaneously (because with a non-directed donor, who donates pair, whose donor “passes it forward” to contracts can’t be written on kidneys). to some incompatible patient-donor pair another such pair or ends the chain with This means that even an exchange under the understanding that they will a donation to someone on the waiting list between two pairs requires four operat- subsequently donate to another, and so for a deceased donor (that is, the chain ing rooms and surgical teams, for the two on. Every pair in this chain will receive a ends when a donation is made to a patient nephrectomies (kidney removal from the kidney before they donate one. If the who does not have a willing but incom- donor) and two transplants. A three-way chain is broken, then the pair that was patible live donor). exchange would require six. When we scheduled but fails to receive a kidney Our 2003 paper proposed kidney presented this initial proposal to our sur- will be disappointed, but not grievously exchange that integrated cyclic exchanges gical colleagues, led by Frank Delmonico, harmed. They are not worse off than they of all sizes and chains beginning with they felt it was a critical problem—the were before the non-directed donor came a non-directed donor and ending with prospect of four simultaneous surgeries forward, and, in particular, they still have a donation to someone without a liv- was daunting enough. They asked us to a kidney with which to participate in ing donor. We focused on two kinds of present a proposal with the more modest some future exchange. Hence the cost of incentive issues that seemed likely to be aim of organizing exchanges involving a broken link in a chain initiated by a important in a mature system of kidney only two-way exchanges. non-directed donor is much less than exchange, both concerned with align- Our new, more limited proposal8and that of a broken link in an exchange ing incentives so as to make it safe and the accompanying software formed the among a cycle of patient-donor pairs. simple to participate. First, we showed basis for organizing the New England In 2007, Mike Rees, a pioneer of how exchanges could be arranged so Program for Kidney Exchange,9 and was kidney exchange and the founder of the that they would be in the core of the widely shared and explained and soon Alliance for Paired Donation, which is game, which means that no coalition of adapted for use elsewhere. Almost simul- one of the most active networks, began patient-donor pairs could go off on their taneously, we began exploring with our the first such non-simultaneous chain. own, or to a competing exchange, and surgical colleagues the possibilities of It was reported on in Rees et al. (2009), do better than to accept the proposed including larger exchanges and chains. at which point it had accomplished ten exchanges. Second, we showed how this 10,11,12 (It speaks volumes about the rela- transplants (and 20 surgeries), many could be accomplished in a way that tive publishing speed of Economics and more than could have been done simul- made it a dominant strategy for patients Medicine to note that the follow-up paper taneously.13 Since then, non-simulta- (and their surgeons) to reveal the med- which reported in the American Journal neous non-directed donor chains have ical information that determined the of Transplantation how longer exchanges become the fastest growing part of kid- desirability of each potential transplant. actually had been carried out was pub- ney exchange, even though the number It is worth noting that the tools we used lished a year later than the publication of of non-directed donors is small. In some built on theory that was initially pro- the original 2005 NBER Working Paper cases a non-directed donor has initiated a posed in a very abstract setting: Shapley analyzing such exchanges.) chain of more than 30 transplants. and Scarf (1974) studied a “top trading Although the three-way chain Ashlagi and I have worked to under-

NBER Reporter • 2012 Number 4 11 stand why long chains are so useful, long chains, and the presence of easy-to- Quarterly Journal of Economics, 119 (2) and how to structure them. As kidney match pairs allows more hard-to-match (May 2004), pp. 457–88. exchange has grown and become a stan- pairs to be included. 2 A.E. Roth, T. Sonmez, and M. Utku dard tool of transplantation, hospitals are Despite the growing success that Unver, “A Kidney Exchange Clearinghouse more able to do some exchanges among kidney exchange has had in facilitat- in New England,” American Economic their own patients. This means the play- ing transplants from living donors, the Review, Papers and Proceedings, 95 (2) ers in the kidney exchange game have list of people waiting for kidney trans- (May 2005), pp. 376–80. changed: where it used to be enough to plants from deceased donors continues to 3 A.E. Roth, “Repugnance as a think of the incentives of patients and grow. Deceased donor organs are a scarce Constraint on Markets,” NBER Working donors and their surgeons, now the direc- resource of an unusual kind, because their Paper No. 12702, November 2006, and tors of transplant centers are players, and supply depends on decisions to donate Journal of Economic Perspectives, 21 (3) they see many patient-donor pairs. Their made by potential donors (while still liv- (Summer 2007), pp. 37–58. strategy sets now include which pairs ing) and their next of kin (immediately 4 W.S.Jevons, Money and the Mechanism to show to a centralized exchange. The afterwards). Consequently there are mar- of Exchange, New York: D. Appleton and present organization of kidney exchanges ket design issues associated with how Company, 1876. gives them some incentives to withhold donations are solicited, and how organs 5 L. Shapley and H. Scarf, “On Cores and their easy-to-match pairs. This could be are allocated, both of which may influ- Indivisibility,” Journal of Mathematical fixed by taking account of which hospitals ence the donation decision and hence the Economics, 1(1) (1974) pp. 23–37. enrolled easy-to-match pairs and using supply. Judd Kessler and I have begun to 6 A.E. Roth, “Incentive Compatibility in a this information (in a sort of “frequent investigate this:16 we begin with an exper- Market with Indivisible Goods,” Economics flier program”) to give some increased imental investigation motivated by a pri- Letters, Vol. 9 (1982) pp.127–32. probability of matching to patients at ority allocation scheme just put into place 7 A. Abdulkadiroglu and T. Sonmez, those hospitals.14 But this faces impor- in Israel, in which people who have regis- “House Allocation with Existing Tenants,” tant political obstacles and has so far not tered as donors will be given some priority Journal of Economic Theory, LXXXVIII been adopted. Partly as a result of the in case they need to receive an organ for (1999), pp.233–60. withholding of easy-to-match pairs, the transplant, and so will members of their 8 A.E. Roth, T. Sonmez, and M. U. percentage of patients enrolled in kid- immediate family. Unver, “Pairwise Kidney Exchange,” ney exchange networks that are hard to While it is natural that economists NBER Working Paper No. 10698, August match, even to a blood-type compatible should investigate institutions that facil- 2004, and Journal of Economic Theory, donor, has skyrocketed. itate exchange, many people (including 125 (2) (December 2005) pp.151–88. We can organize patient and donor some economists) find it surprising that 9 A.E. Roth, T. Sonmez, and M. data in a compatibility graph, in which economists should be helping to design U. Unver, “A Kidney Exchange each node represents a patient and her the institutions of kidney exchange. This Clearinghouse in New England”. incompatible donor(s), and an edge goes is a natural outgrowth, however, of two 10 S. L. Saidman, A. E. Roth, T. from one node to another whenever the strands in modern economics: market Sönmez, M. U. Ünver, and F. L. donor in the first node is compatible design in general17, and the study of Delmonico, “Increasing the Opportunity with the patient in the second node. As matching markets. Matching markets are of Live Kidney Donation By Matching patients have become harder to match, the those in which price does not do all the for Two and Three Way Exchanges,” compatibility graphs have become sparser, work of determining who gets what, and Transplantation,81(5) (March 15, 2006) that is, they contain fewer edges. When they include some of the important pas- pp. 773–82. we look at the data of the kidney exchange sages in our lives, from school choice and 11 A. E. Roth, T. Sönmez, M. U. Ünver, networks with which we work, there is college admissions to marriage and labor F. L. Delmonico, and S. L. Saidman, a densely connected sub-graph of the markets. In none of these can you simply ‘’Utilizing List Exchange and Undirected relatively few fairly easy-to-match pairs, choose what you want — you also have to Good Samaritan Donation through and a sparse sub-graph of many hard-to- be chosen. In some of these, economists ‘Chain’ Paired Kidney Donations,” match pairs (this is joint work with David have begun to help design the matching American Journal of Transplantation, 6, Gamarnik and Mike Rees). Within the institutions. (11) (November 2006) pp. 2694–705. easy-to-match sub-graph, many patients Economists should welcome oppor- 12 A. E. Roth, T. Sonme, and M. U. could be transplanted with the aid of two- tunities to learn how to be engineers.18 Unver, “Efficient Kidney Exchange: way or three-way exchanges, but within Coincidence of Wants in Markets with the sub-graph of hard-to-match pairs, Compatibility-Based Preferences,” NBER only long chains offer the chance of trans- 1 A.E. Roth, T. Sonmez, and M. Utku Working Paper No. 11402, June 2005, planting many patients.15 Non-directed Unver, “Kidney Exchange,” NBER Working and American Economic Review, 97 (3) donors have a chance of starting those Paper No.10002, September 2003, and (June 2007) pp. 828–51.

12 NBER Reporter • 2012 Number 4 13 M. A. Rees, J. E. Kopke, R. P. Pelletier, Exchange,” NBER Working Paper No. 17 A.E. Roth, “What Have We Learned D. L. Segev, M. E. Rutter, A. J. Fabrega, 16720, January 2011. from Market Design?” NBER Working J. Rogers, O. G. Pankewycz, J. Hiller, A. 15 I. Ashlagi, D. Gamarnik, M. A. Rees, Paper No. 13530, October 2007, and E. Roth, T. Sandholm, M.U. Ünver, and and A. E. Roth, “The Need for (long) Economic Journal, 118 (March 2008) R. A. Montgomery, ‘’A Non-Simultaneous Chains in Kidney Exchange,” NBER pp.285–310. Extended Altruistic Donor Chain,” New Working Paper No.18202, July 2012. 18 A.E. Roth, “The Economist as Engineer: England Journal of Medicine, 360(11) 16 J. B. Kessler and A. E. Roth, “Organ Game Theory, Experimentation, (March 12, 2009) pp. 1096–1101. Allocation Policy and the Decision to and Computation as Tools for Design 14 I. Ashlagi and A.E. Roth, “Individual Donate,” NBER Working Paper No. 17324, Economics,” Econometrica, 70 (4) (July Rationality and Participation in Large August 2011, and American Economic 2002) pp. 1341–78. Scale, Multi-Hospital Kidney Review, 102 (August 2012). pp. 2018–47.

The Sources and Consequences of Productivity Differences

Chad Syverson*

Economists have consistently found business and those that are woefully that are external to the firm: elements both large and persistent differences in lacking in such knowledge. Far more of the industry or market environment measured productivity across producers, than bragging rights are at stake, because that can induce productivity changes or even within narrowly defined industries. higher productivity producers are more support productivity dispersion. I have The size of these differences is striking: likely to survive than their less efficient researched factors in both categories. for instance, within U.S. 4-digit SIC man- industry competitors. ufacturing industries (such as saw blade The discovery of these ubiquitous, Levers that Influence manufacturing), the plant at the 90th large, and persistent productivity dif- Productivity percentile of the industry’s productivity ferences has shaped research agendas in distribution typically obtains almost twice a number of fields, including (but not On the “lever” side of the ledger, as much output with the same measured limited to) macroeconomics, corporate Steven Levitt, John List, and I look at inputs as the plant at the 10th percen- finance, industrial organization, labor, the mechanisms that underlie learning tile of productivity. (These figures, and and trade. I have studied various aspects of by doing — productivity gains achieved all those described below, use total factor the sources and consequences of produc- through the very act of producing.2 Using productivity measures. They reflect the tivity dispersion as a part of my research extremely detailed data from an assem- amount of output that a producer obtains agenda; this essay summarizes that work. bly plant of a major auto producer, we from a given combination of labor, capi- find that productivity gains from learn- tal, and intermediate inputs.) And U.S. Two Sources of Productivity ing arrive quickly and in force. Defects manufacturing is not exceptional in this Differences per vehicle fall by more than 80 percent regard; in fact, researchers have docu- in the first eight weeks of production. mented even larger dispersion in other In a recent survey article, I review Interestingly, when the plant’s second shift sectors and countries. the research over the past decade that has comes on line at this point, the learn- The observed persistence of pro- sought to explain the sources of observed ing process does not begin again. Instead, ducers’ productivity levels indicates that productivity differences.1 I split the expla- the second shift actually comes on line at industries typically contain both firms nations into two categories. One includes defect rates lower than the first shift’s con- that appear to have figured out their factors that operate within the plant or temporaneous rates, despite the first shift’s * Syverson is an NBER Research Associate firm and which directly affect produc- two month head start in production. And, and a Professor of Economics at the tivity at the producer level. These are while worker absenteeism statistically University of Chicago’s Booth School of the “levers” that management or others affects defect rates, its impact is econom- Business. His profile appears later in this potentially can use to influence produc- ically small. Furthermore, the hundreds issue. tivity. The second category includes forces of assembly processes on the line have

NBER Reporter • 2012 Number 4 13 highly correlated defect rates across shifts, roles of both competition and regula- be in nonattainment with the Act’s pollu- even though the workers completing these tions in influencing producer productiv- tion limits, a determination that subjects tasks are different. Taken together, these ity levels. Most models of competition those plants to command-and-control- patterns illustrate one of our main find- among heterogeneous-productivity pro- style abatement mandates. We compare ings about the learning mechanisms at the ducers share a prediction that a greater productivity growth at these plants to plant: rather than remaining with work- ability or willingness of consumers to sub- their industry cohorts located in coun- ers, much of what is learned very quickly stitute across producers either will induce ties that are in attainment with the Act’s becomes embodied in the plant’s physical low productivity suppliers to improve provisions, and to plants in non-polluting or organizational capital. This finding is their efficiency or will force them to exit. industries that are free from regulation consistent with the institutional processes Either effect truncates the market’s equi- in all counties. We find that for surviv- that plant management puts in place to librium productivity distribution from ing plants in heavily polluting industries, encourage knowledge dissemination. below, thereby raising average productiv- a nonattainment designation and its asso- In a series of papers, Enghin Atalay, ity and reducing productivity dispersion. ciated abatement mandates result in an Ali Hortaçsu, and I examine the connec- I test this prediction in studies look- average 4.8 percent decline in the plants’ tions between firms’ vertical structures ing both across industries and across total factor productivity. In plain lan- and their plants’ productivity levels.3 We markets within an industry. The across- guage, this means the amount of output find that vertically integrated plants have industry analysis uses producer-level data that the plants are able to produce from higher productivity levels than their non- from 443 U.S. manufacturing industries a given amount of inputs (that is, labor, integrated industry cohorts. However, the and finds that industries with more sub- capital, and materials) is 4.8 percent lower evidence suggests that little of this differ- stitutable output — measured in sev- than before the abatement mandates. This ence is related to the firms’ vertical struc- eral ways, including aspects of spatial, output loss corresponds to an annual eco- tures per se, but rather to other factors physical, and brand-driven differentia- nomic cost from the regulation of manu- correlated with integration status. In fact tion — have less productivity disper- facturing plants of roughly $21 billion in these productivity differences — and the sion and higher median productivity 2010 dollars, about 8.8 percent of average firm’s decisions about whether to have a levels.4 The within-industry investiga- annual manufacturing sector profits over vertical structure in the first place — are tion focuses on the ready-mixed concrete the sample period. not usually related to the movement of industry.5 The industry’s homogeneous goods along the production chain. Using product and very high transport costs Productivity vs. Demand detailed shipment-level data on the flow make the density of concrete produc- of goods throughout the economy, we find ers in a market a primary determinant While productivity is typically that vertically integrated firms’ upstream of the intensity of competition (that thought of as a feature of the produc- plants ship a surprisingly small amount is, substitutability). There too, the pre- tion technology, as actually measured in to downstream plants in their firm (that dicted truncation effect of substitutabil- producer micro-data it generally reflects is, small relative to both the firms’ total ity is observed in the data. Markets with more than just supply-side forces. Much upstream production and their down- denser construction activity (an exog- of the work I’ve just described, and most stream needs). Almost half of upstream enous shifter of concrete producer den- of the broader literature investigating pro- plants report no shipments to downstream sity) have higher lower-bound produc- ductivity differences among businesses, units inside their firm. About 90 percent tivity levels, higher average productivity, uses revenue to measure output because of upstream plants ship less than a third of and less productivity dispersion. In fol- business-level price indexes are rarely their output internally. These patterns sug- low-up work, I demonstrate that these available. This means that within-indus- gest that vertical ownership is not usually patterns of competition-driven selection try price differences are embodied in out- about moderating goods transfers along on costs also are reflected in ready-mixed put and productivity measures. If prices production chains. We propose and find prices.6 reflect in part idiosyncratic demand shifts suggestive evidence that the primary pur- My recent work with Michael or market power variation across pro- pose of integration instead is to facili- Greenstone and John List considers reg- ducers — a distinct likelihood in many tate within-firm transfers of intangible ulation’s effect on plants’ productivity industries — then high “productivity” inputs (for example, managerial oversight levels.7 We use detailed production data businesses may not be especially techno- or intellectual capital). from nearly 1.2 million plant observa- logically efficient. tions from the 1972–93 Annual Survey A new strand of research has begun External Factors that of Manufactures to measure the economic to extend the productivity literature to Influence Productivity costs of the Clean Air Act Amendments. also explicitly account for such idiosyn- We track productivity growth at plants cratic demand effects. Lucia Foster, John My research on the external factors from heavily polluting industries that are Haltiwanger, and I have been active in shaping productivity has looked at the located in counties declared by the EPA to this area. We take advantage of the avail-

14 NBER Reporter • 2012 Number 4 ability of physical output data for a select lished industry competitors, and that this 250–301, and E. Atalay A. Hortaçsu, set of “commodity-like” product indus- size gap closes slowly.9 We show that these and C. Syverson, “Why Do Firms Own tries (for example, cardboard boxes, white patterns are not a result of physical pro- Production Chains?” NBER Working pan bread, and sugar). This lets us mea- ductivity gaps, but instead reflect differ- Paper No. 18020, April 2012. sure not just the standard revenue-based ences in demand. Even though new pro- 4 C. Syverson, “Product Substitutability productivity metric, but also its two com- ducers are technically more efficient, they and Productivity Dispersion,” NBER ponents: physical-quantity-based produc- sell only a fraction of the output of their Working Paper No. 10049, October tivity (number of units of output per unit more established competitors. Estimating 2003, and Review of Economics and input, reflecting more closely the pure a dynamic model of plant expansion in Statistics, 86(2), May 2004, pp. 534–50. supply-side concept of productivity) and the presence of a demand accumulation 5 C. Syverson, “Market Structure and average unit price. We show that there are process (for example, building a customer Productivity: A Concrete Example,” important differences between revenue base), we find that this accumulation NBER Working Paper No. 10501, May and physical productivity. results mostly through businesses’ active 2004, and Journal of Political Economy, In one paper, we consider the sepa- investments in building demand, rather 112(6) (2004), pp. 1181–1222. rate roles that supply- and demand-side than through passive processes tied simply 6 C. Syverson, “Prices, Spatial fundamentals play in driving selection to the passage of time. We also show that Competition, and Heterogeneous and survival in heterogeneous-producer within-firm demand spillovers, like those Producers: An Empirical Test,” NBER industries.8 We show that physical pro- conferred by established firms on their Working Paper No. 12231, May 2006, ductivity is inversely correlated with price new plants, affect plants’ initial demand and Journal of Industrial Economics, while revenue productivity is positively levels but not their growth. 55(2) (2007), pp.197–222. correlated with price. This means that 7 M. Greenstone, J. A. List, and C. previous work linking (revenue-based) Syverson, “The Effects of Environmental productivity to survival has confounded 1 C. Syverson, “What Determines Regulation on the Competitiveness of the separate and opposing effects of tech- Productivity?,” NBER Working Paper U.S. Manufacturing,” NBER Working nical efficiency and demand on survival, No. 15712, January 2010, and Journal Paper No. 18392, September 2012. understating the true impacts of both. of Economic Literature, 49(2) (2011), 8 L. Foster, J. C. Haltiwanger, and Perhaps most strikingly, we find that even pp. 326–65. C. Syverson, “Reallocation, Firm in these near-commodity industries, a 2 S. D. Levitt, J. A. List, and C. Turnover, and Efficiency: Selection on producer’s demand is particularly impor- Syverson, “Toward an Understanding Productivity or Profitability?” NBER tant for its survival prospects. A given- of Learning by Doing: Evidence from Working Paper No. 11555, August sized shift in a producer’s demand level an Automobile Assembly Plant,” NBER 2005, and American Economic has four times the effect on its likelihood Working Paper No. 18017, April 2012. Review, 98(1) (2008), pp. 394–425. of surviving as does the same-sized shift 3 A. Hortaçsu and C. Syverson, 9 L. Foster, J. C. Haltiwanger, and C. in its physical productivity. “Cementing Relationships: Vertical Syverson, “The Slow Growth of New A second paper looks at the role Integration, Foreclosure, Productivity, Plants: Learning about Demand?” of demand in explaining the well docu- and Prices,” NBER Working Paper No. NBER Working Paper No. 17853, mented fact that new businesses on aver- 12894, February 2007, and Journal of February 2012. age are much smaller than their estab- Political Economy, 115(2) (2007), pp.

NBER Reporter • 2012 Number 4 15