A Service of

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

National Bureau of Economic Research (NBER) (Ed.)

Periodical Part NBER Reporter Online, Volume 2003

NBER Reporter Online

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

Suggested Citation: National Bureau of Economic Research (NBER) (Ed.) (2003) : NBER Reporter Online, Volume 2003, NBER Reporter Online, National Bureau of Economic Research (NBER), Cambridge, MA

This Version is available at: http://hdl.handle.net/10419/61986

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 SUMMER 2003

IN THIS ISSUE Program Report Program Report: The Economics of Aging 1 The Economics of Aging Research Summaries: Randomized Trials and David A. Wise* Quasi-Experiments in Education Research 11 Inflation Dynamics: Combining Population aging, early retirement, limited but increasing retirement Measurement with Theory 15 saving, more expensive medical practice patterns, and an established The New Economics of Smoking 19 national entitlement to income and health care support after age 65 — Venture Capitalists As Economic all of these factors largely define the economic environment of the Principals 22 United States (and much of the world) at the beginning of the 21st cen- tury. Over the past 30 years, life expectancy has increased from age 71 to age 77, while the most common age of retirement has decreased from NBER Profiles 25 age 65 to age 62. Retiring at age 62, the typical American retiree today Conferences 28 faces another 20 years of living, consuming and, at one time or another, Bureau News 39 and in many cases regularly, needing expensive health care services. Bureau Books 49 These trends already have placed significant financial pressure on the Current Working Papers 50 public and employer-sponsored programs that provide income and health care support to older Americans. Meanwhile, the massive demographic bulge in the population — the baby boom generation — begins turning age 62 in 2008. Going for- ward, the number of Americans age 62 and older is projected to double from 40 million today to 80 million 30 years from now, while the work- ing age population is projected to grow by just 12 percent over the same period. Compounding the demographic situation is the continuing rise in medical costs. National health care expenditures have grown from $250 billion annually in 1980 to $1.4 trillion today, and show little sign of slow- ing down. The combination of economic, labor market, health and NBER Service Brings demographic trends points to any number of social and economic chal- You New Data for Free lenges in the decades ahead. Understanding the complexities of this sit- uation, and the relationships between demographics, policy, behavior, A new, free NBER email serv- economics, and health — this is the substantive aim of the NBER ice gives you daily email links to all Program on the Economics of Aging. U.S. government data releases, Begun in 1986, the Aging Program has developed primarily around including unemployment, trade, large, coordinated research projects that simultaneously address several interest rates, GDP, etc. We keep interrelated issues in the economics of aging. Extensive funding for the track of your preferences and program has been provided by the National Institute on Aging (NIA), email you the requested links both through multiple research grants and through a Center grant, which when they are released. To sign up provides centralized infrastructure support to the program effort. NIA for any or all of the government also has supported our efforts to engage new investigators in studying releases, visit www.nber.org/releases * and register your choices. IT’S Wise is Director of the NBER’s Program on the Economics of Aging and the Stambaugh Professor of Political Economy at Harvard University’s Kennedy School of Government. The numbers in brackets FREE!! throughout this report refer to NBER Working Papers. This report has been prepared with the intensive help of Richard Woodbury.

NBER Reporter Summer 2003 1. issues in aging, and at least a dozen graduate students and post-doctoral research fellows become engaged in the program each year NBER Reporter through NIA fellowships. Many more become engaged as research assistants on project NATIONAL BUREAU OF ECONOMIC RESEARCH grants. A number of smaller “exploratory” projects on issues in aging are supported The National Bureau of Economic Research is a private, nonprofit research through the Center, as well as projects that organization founded in 1920 and devoted to objective quantitative analysis of integrate related components of the overall re- the American economy. Its officers and board of directors are: search effort. More than 100 papers are com- President and Chief Executive Officer — Martin Feldstein pleted annually by participants in the NBER Vice President for Administration and Budget — Susan Colligan program. Some of these also appear in a series Controller — Kelly Horak of books published by the University of 1 BOARD OF DIRECTORS Chicago Press. Chairman — Michael H. Moskow Vice Chairman — Elizabeth E. Bailey Treasurer — Robert Mednick The Economic Circumstances of Older DIRECTORS AT LARGE Americans Peter Aldrich Jacob A. Frenkel Michael H. Moskow Elizabeth E. Bailey Judith M. Gueron Alicia Munnell John Herron Biggs Robert S. Hamada Rudolph A. Oswald Personal Retirement Accounts Andrew Brimmer George Hatsopoulos Robert T. Parry John S. Clarkeson Karen N. Horn Richard N. Rosett Don R. Conlan Judy C. Lewent Marina V. N. Whitman A major theme of the Aging Program George Eads John Lipsky Martin B. Zimmerman since its inception has been to better under- Martin Feldstein Laurence H. Meyer stand the economic circumstances of older Americans. A key fact from early project work DIRECTORS BY UNIVERSITY APPOINTMENT was the very small amount of financial assets George Akerlof, California, Berkeley Marjorie B. McElroy, Duke of most retirees in the United States in the Jagdish W. Bhagwati, Columbia Joel Mokyr, Northwestern early 1980s and the overwhelming reliance on William C. Brainard, Yale Andrew Postlewaite, Pennsylvania Michael J. Brennan, California, Los Angeles Uwe Reinhardt, Princeton Social Security, and in some cases firm pension Glen G. Cain, Wisconsin Nathan Rosenberg, Stanford plans, for financial support in retirement. But Franklin Fisher, MIT Craig Swan, Minnesota over the last two decades, there has been a dra- Saul H. Hymans, Michigan David B. Yoffie, Harvard Arnold Zellner, Chicago matic transformation in the magnitude of sav- ing that is taking place in personal retirement DIRECTORS BY APPOINTMENT OF OTHER ORGANIZATIONS accounts, such as IRAs and 401(k) plans. Now Mark Drabenstott, American Agricultural Economics Association about 85 percent of contributions to retire- Gail Fosler, The Conference Board ment plans are to personal accounts. In A. Ronald Gallant, American Statistical Association Richard C. Green, American Finance Association decomposing trends in plan eligibility, partici- Robert Mednick, American Institute of Certified Public Accountants pation given eligibility, participant contribu- Angelo Melino, Canadian Economics Association tions, and aggregate wealth accumulation in Richard D. Rippe, National Association for Business Economics John J. Siegfried, American Economic Association personal retirement accounts, and in projecting David A. Smith, American Federation of Labor and Congress of the future of these trends, James M. Poterba, Industrial Organizations Josh S. Weston, Committee for Economic Development Steven F. Venti and I have confirmed the Gavin Wright, Economic History Association potential of 401(k) plans to significantly alter the financial circumstances of individuals retir- The NBER depends on funding from individuals, corporations, and private foundations to maintain its independence and its flexibility in choosing its ing in the future [W8610, W6295, W5762]. research activities. Inquiries concerning contributions may be addressed to Because more individuals will have had more Martin Feldstein, President & CEO, NBER 1050 Massachusetts Avenue, years of participation in 401(k)-type savings Cambridge, MA 02138-5398. All contributions to the NBER are tax deductible. plans, they will reach retirement with increasing accumulations of financial assets. For many, The Reporter is issued for informational purposes and has not been reviewed by the Board of Directors of the NBER. It is not copyrighted and can be freely their personal retirement accounts will con- reproduced with appropriate attribution of source. Please provide the NBER’s tribute as much or more in retirement than Public Information Department with copies of anything reproduced. Social Security. The most recent in a long series Preparation of the NBER Reporter is under the editorial supervision of Donna of studies on this topic shows that the ratio of Zerwitz. retirement plan assets to wage and salary earn- ings has grown more than five fold since 1975. Requests for subscriptions, changes of address, and cancellations should be sent to Reporter, National Bureau of Economic Research, Inc., 1050 This represents a fundamental transition in the Massachusetts Avenue, Cambridge, MA 02138-5398. Please include the cur- composition of post-retirement financial sup- rent mailing label. port in the United States. 2. NBER Reporter Summer 2003 Despite these aggregate trends, is new saving that would not have only equities, and 22 percent hold only it’s clear that there is wide variation in occurred without these plans [W5599]. debt. These results provide a starting saving behavior across the population. Studies of the potential offsetting point from which to consider the mar- For example, Venti and I have found decline in defined benefit pension plan ket risk of personal accounts. that saving rates vary substantially at coverage conclude that the growth of all income levels, and that these varia- personal retirement accounts dwarfs tions lead to vastly different levels of any displacement of assets in tradi- Social Security, Housing, asset accumulation over a working tional plans [W8610]. And, studies of Annuities, and Bequests career [W7521]. What explains this alternative measures of saving empha- variation in behavior? Part of the size the implications of not counting As the primary source of retire- answer is simply choice: some people capital gains as part of saving in the ment income support in the United choose to save a lot, and others do NIPA personal saving rate, while the States, Social Security defines the eco- not. But that choice is made in the expenditure of money that has accu- nomic circumstances of many older context of one’s social and economic mulated in the form of capital gains is Americans. Many members of the environment, as well as the public and still counted as dis-saving (or spend- Aging Program are engaged in analyses employer policies that relate to indi- ing) [W8237, W8610]. As a result, the of Social Security, as well as on the vidual saving decisions. In a series of NIPA-measured saving rate can be potential design and implications of studies, James Choi, David Laibson, very low — about 2 percent now — various Social Security reform possibil- Brigitte Madrian, Andrew Metrick, while the retirement plan contribution ities. In this report, I have focused on and Dennis G. Shea find that the rate (as a percentage of wage and salary projects supported by the National “default” provisions of 401(k) plans earnings) has been over 8 percent. Institute on Aging as part of NIA make a huge difference in whether and Looking ahead, our research research grants. One such project, how much people save in their 401(k) agenda has evolved to focus on the directed by Jeffrey B. Liebman, looks at plans [W8655, W8651, W7682]. Many risks associated with different forms of how alternative Social Security reforms more employees participate when retirement saving, and how recent would be likely to affect lower-income there is automatic enrollment; and trends may have changed the risk households. Social Security is widely they contribute more to their plans, on exposure of individuals in providing recognized as a redistributive program, average, when the default contribution for their retirement. The different risk replacing a greater fraction of earnings rate of the program is higher. Recent characteristics of fixed assets (like for those at lower income levels. research by Esther Duflo and 401(k) accounts), as compared with Liebman’s research, however, points to Emmanuel Saez provides complemen- annuitized assets (like Social Security many other redistributive aspects of tary evidence of increased participa- and traditional pension plans) is one Social Security — from people with tion in retirement plans by individuals piece of this work. Another is the shorter life expectancies to people with who received a financial incentive to increased exposure to market fluctua- longer life expectancies, from single attend a program information session, tions that is associated with personal workers and dual-earning couples to as well as by individuals who did not retirement accounts. The extreme case one-earner couples, and from long- receive the incentive, but who worked of company stock as a retirement career workers to short-career workers. in departments where others received investment has been explored recently Because of these other factors, about the incentive [W8885, W7735]. Thus by Olivia S. Mitchell and Stephen P. 20 percent of Social Security partici- the influence of peer behavior was Utkus; they confirm the risk of such pants in the top income quintile receive found to be a significant factor in plan investments, and identify policy tools a larger net transfer from Social Security participation decisions. that might encourage greater diversifi- than the average transfer for people in Related studies from the Aging cation [W9250]. Poterba, with various the lowest income quintile [W8625]. Program have considered more specif- coauthors, also has looked at the allo- Liebman also finds that Social ic aspects of retirement saving. Studies cation of investments in personal Security reforms that blend the current of pre-retirement withdrawals from retirement accounts, focusing particu- system with an investment-based com- 401(k) plans find little effect on the larly on the mix between debt and ponent could give most Social Security total dollar accumulations in these equity holdings [W9268, W7991, beneficiaries higher expected benefits plans, as a large fraction of job chang- W7392]. Among other findings, his than the current system, and lower the ers don’t withdraw assets from their research shows that the aggregate allo- percentage of widowed, divorced, and personal accounts and a large fraction cation between stocks and bonds in never married women with benefits of assets that are withdrawn are individual accounts is broadly similar below the poverty line from 26 percent “rolled over” into other personal to the allocation one finds in defined to 9 percent [W7492]. Kathleen retirement accounts [W7314]. Studies benefit plans, which are managed by McGarry has also looked at poverty of whether the assets in retirement financial professionals. But there is among the elderly, noting that reforms accounts have replaced saving that much more variation across individual to the Supplemental Security Income would have otherwise taken place in retirement accounts. For instance, just Program could have a significant effect other forms conclude that the large over 45 percent of households with in reducing elderly poverty [W7574]. majority of personal retirement saving tax-deferred accounts appear to hold She also finds that a surprising number

NBER Reporter Summer 2003 3. of potential SSI recipients do not apply find a heavy concentration of transfer and services. He also finds that 56 per- for benefits. wealth among a relative smaller num- cent of full income was supported by Aside from Social Security (and ber of households. Focusing on the transfers from the young; compared for some, an employer-provided pen- allocation of bequests among siblings, with 44 percent from the elderly them- sion), housing equity is the major asset McGarry and Audrey Light find that selves. He also points to future trends of a large fraction of current retirees. 80 percent of older parents plan to which may stimulate the need for more Despite the value of these housing make equal bequests. Among those saving and more work prior to retire- assets, Venti and I find that housing planning to divide their estates ment. equity is rarely used to support general unequally, about half point to the non-housing consumption during inequality as compensation for a child Retirement Policies and retirement [W8608, W7882]. We find taking care of them in their old age, that in the absence of a precipitating and half point to differences in the Labor Market Behavior event (such as divorce, the death of a needs of different children. spouse, or the entry of a family mem- Aging Program research is also Retirement Policy in the United ber into a nursing home), families are exploring how assets are used at older States unlikely to sell their homes, downsize, ages, and how consumption changes at or remortgage. And even among those retirement. Recent work by Hurd and A second major theme of the experiencing major life changes, dis- Susann Rohwedder, for example, finds a Aging Program is the relationship continuing ownership is the exception reduction in both anticipated and actual between retirement policies and labor rather than the rule. A related study by spending at retirement [W9586]. Brown, market behavior, as well as the deter- Gary V. Engelhardt, Jonathan Gruber, Amy Finkelstein, Poterba, Mitchell, and minants of work and retirement deci- and Cynthia Perry finds that the living others are studying issues in annuity sions more generally. Over the years, arrangements of widows and divorcees pricing, the differences in mortality the strong relationship between the are more sensitive to economic circum- experience between annuitants and the economic incentives of retirement stances than the living arrangements of population at large, the potential for policies and the ages at which individ- couples [W8911]. They estimate that a good inflation-adjusted annuity prod- uals retire from the labor force has 10 percent cut in Social Security bene- ucts, the differing characteristics of indi- been confirmed in multiple studies, fits would lead more than 600,000 sin- vidual and joint-life annuity products, using multiple data sources, and apply- gle-resident elderly households to the differences in pricing between ing multiple research methodologies. move into shared living arrangements. mandatory and voluntary annuity prod- The most recent have been conducted The role of bequests in the finan- ucts, and the redistributive effects of by Courtney Coile and Gruber [W7830, cial circumstances of older Americans annuities from those with shorter to W7651, W7318], Alan L. Gustman and also has been considered in a number those with longer life expectancies Thomas L. Steinmeier [W9183, W8229, of recent studies. One study by [W9256, W8567, W8064, W8045, W7588], and Andrew A. Samwick and Michael D. Hurd and James Smith W7812, W7560, W7268, W7199, me.2 These studies point to the younger finds that the distribution of bequests W7193, W7191, W7168, W7005, retirement that occurs, when younger is highly skewed, so that the typical W6918]. As more assets are being accu- retirement is encouraged by the eco- baby-boom person will receive a very mulated in personal retirement accounts, nomic structure of the benefit pro- modest inheritance [W9142]. This is the question of whether, under what cir- grams. Recent work by Courtney Coile partly because of the skewed distribu- cumstances, and how much those assets has extended this line of analysis to the tion of wealth and partly because of might be annuitized is an increasingly joint retirement decisions of couples, the tendency of the wealthy to have important one. An important finding of finding for example that the economic fewer children. But it is also attributa- this work is the higher longevity of incentives in a wife’s retirement plan ble to anticipated dis-saving: it is esti- those who purchase annuities, relative to can have a significant effect on the mated that households aged 70-74 will the population as a whole, and the need retirement decisions of both husband bequeath just 39 percent of their to price annuities higher because of this and wife [W9496]. Gustman and wealth, consuming the rest before they differential. Steinmeier also have explored joint die. Jeffrey Brown and Scott Weisbenner Finally, a broad view of the finan- retirement decisions, and find much also have explored the magnitude of cial circumstances of older Americans stronger interdependence when the assets that are passed from one genera- is presented in a recent study by Victor spouses say they value spending time tion to another through bequests and R. Fuchs [W8236]. Fuchs takes what he together [W8772]. other intergenerational transfers calls “a holistic approach” to the finan- Many other studies have looked [W8753]. They find that about one-fifth cial problems of the elderly, focusing at other determinants of retirement. of current household wealth (on aver- simultaneously on expenditures that A recent study by McGarry finds that age) was obtained as a result of trans- are self-financed and those financed by changes in retirement expectations are fers and bequests, while four-fifths transfers from the young (under age driven to a much greater degree by resulted from individual decisions 65). He finds that about 35 percent of changes in health than by changes in about how much income to save. But, the elderly’s full income was devoted to income or wealth [W9317]. Gustman as in the Hurd and Smith study, they health care; 65 percent to other goods and Steinmeier find that the dramatic

4. NBER Reporter Summer 2003 stock market rise in the 1990s had an across countries between the social retirement; and the relationship between effect in increasing retirement, while it security incentives to retire and the age social security system provisions and lasted, but any continuing effects since at which older workers withdraw from the employment of the young. All of the decline in the market are much the labor force. The second phase of the new work is being done in the con- more modest [W9404]. A study by the project applied micro-data from text of an evolving retirement policy Hurd, James Smith, and Julie M. each country to estimate more formal- environment worldwide, a policy evo- Zissimopoulos finds that those who ly the relationship between social secu- lution driven in large part by a growing have worse survival expectations retire rity provisions and retirement in each awareness that social security programs sooner, and collect retirement benefits country. The models emphasized the are not sustainable under demographic sooner, although the majority of work- effect on retirement of the age of eli- trends which are compounded by pro- ers claim Social Security benefits as gibility for retirement programs, and gram inducements for older workers to soon as they are eligible, regardless of measures of the incentive to leave the leave the labor force at younger and their survival expectations [W9140]. labor force once a person is eligible. younger ages. Leora Friedberg finds that computer The model simulations confirmed very users retire later than non-users large effects of program provisions on [W8297]. And a review of the literature retirement decisions in every country. Socioeconomic by Gruber and Madrian (including a The third phase of the project Circumstances, Health, number of their own studies) concludes (just finished) has used the retirement and Health Disparities that health insurance availability is model estimates from phase two to another important determinant of describe the fiscal implications of var- retirement [W8817]. This collection of ious illustrative social security reforms. There is a dramatic and well doc- studies makes clear the complexity of The first simulation predicts the effect umented correlation between socioeco- interrelated health, economic, social, of delaying all program benefit eligibil- nomic measures, such as education, and job circumstances that contribute ity ages by three years. The second income, and wealth, on the one hand, to individual retirement decisions. reform is a common social security and health measures, such as self- system in all countries. The third reported health status and mortality, reform is an adjustment of each coun- on the other. What leads those with Social Security and Retirement try's program to an actuarially fair level more income, education, wealth, and Around the World with benefit payment rates adjusted on other measures of social status to be in an actuarial basis for earlier or later better health, on average; and what Beyond the analyses of retirement retirement. While the simulations were leads those in better health to have behavior in the United States, we also conducted for each of the countries in more economic resources? Many have been engaged in a major cross- the project, the character of the find- potential causal links have been national project on social security sys- ings can be illustrated by looking at the explored in NBER research, and the tems and retirement around the world. case of Germany, which has not (until subject is a key aspect of our research This project, now in its third phase, has recently) reduced retirement benefit agenda going forward. brought together a team of investiga- rates for those qualifying for early In a series of studies, Angus S. tors from Belgium, Canada, Denmark, retirement. The findings suggest that Deaton and Christina Paxson have Italy, France, Germany, Japan, the implementing actuarially fair adjust- explored the multiple facets of the rela- Netherlands, Spain, Sweden, the United ments to retirement benefits in tionship between health and economic Kingdom, and the United States. For Germany would lead to an increase in circumstances, using data from a num- each phase of the project, a set of par- the average retirement age of about 3 ber of countries, and making clear that allel studies has been conducted in each years, for both men and women. there is no simple causal relationship country; and these studies are then Combining the impact of such a [W8534, W8370, W8318, W8099, integrated together to allow compar- reform on retirement behavior, benefit W7141, W7140]. Indeed the nature of isons across countries. Both the indi- payments, contributions to the social the relationship may be quite different vidual studies [such as W9494, W9455, security system, and other taxes paid as across geographic regions with differ- W8658] and the integrated analyses are a result of retention in the labor force ent living standards, health conditions, published in a series of volumes by the — there is an estimated fiscal benefit racial composition, population density, University of Chicago Press.3 These to the illustrative policy change of over inequality, and other factors. Most volumes have been supplemented by 80,000 Euros per worker, or about 1.2 recently, the focus of this work has publications that describe the overall percent of GDP in Germany. been on economic inequality, inde- results [W9407, W8103].4 Continuing work on the project pendent from economic status, as a The first phase of the project will consider the relationship between potential determinant of inequalities in mapped out the detailed provision of social security system provisions and health. One approach of this investiga- social security programs into measures the well-being of the elderly, as well as tion has focused on long-term histori- of retirement incentives comparable the young; the more complicated rela- cal trends in income inequality in the across countries. The studies demon- tionships between health, functional United States and United Kingdom, strated a very strong correspondence disability, social security provisions, and finding that the long-term decline in

NBER Reporter Summer 2003 5. mortality rates over the last half cen- of the intergenerational transmission of — and indeed the cost containment tury occurred most rapidly when socio-economic status may work strategies in all health insurance plans income inequality was rising — just through the impact of parents’ long-run — is increased financial pressure on the opposite of what one might average income on children’s health. hospitals and associated limitations on expect if income inequality were an Fuchs, Mark B. McClellan, “charity” care. This has led to more important causal factor of poorer Jonathan S. Skinner, and others have division of patients across hospitals health. The project also has looked at studied the geographic, income, and according to their ability to pay. Low relative deprivation (low income rela- racial distribution of medical utiliza- or non-paying patients increasingly are tive to a reference group) as a poten- tion [W8628, W6910, W6013]. One directed toward safety net hospitals, tial influence on health, developing a study explores the flow of Medicare and fewer resources are available to relative deprivation model of health benefits to high-income and low- these hospitals to provide un-reim- determination, along with empirical income neighborhoods in 1990 and bursed care. In other words, hospitals analyses to test the model. In the 1995. The authors find that Medicare that have traditionally provided more empirical work, relative deprivation as spending per capita for the lowest care to uninsured patients are being a measure of inequality also appears income groups grew much more rapid- doubly affected: first by having respon- to have little effect on health out- ly than Medicare spending in either sibility for an even larger portion of comes. The general conclusion of this high income or middle income neigh- uninsured care; and second by getting line of research is that while the direc- borhoods. Thus, disparities in health less reimbursement from insured patients. tion of the causal relationship care access and health care utilization Many other components of between health and wealth may be in may be decreasing over time. NBER research in the Aging Program doubt, the relationship is not deter- Another set of studies looks at relate to the theme of socioeconomic mined by economic inequality. the potential impact of insurance cov- status, health and health disparities. Studies by Anne C. Case and Duflo erage as a determining factor in health Robert T. Jensen has been engaged in have looked at the relationship between care utilization and access to care. For a project on the effects of the macro socioeconomic status and health from example, a question addressed in sev- economy on health circumstances in a somewhat different angle, focusing eral studies is whether managed care Russia and Eastern Europe. McClellan on the health effects of the expanded has limited access to care. Paul has looked at the effects of adverse pension program in South Africa. This Heidenreich, McClellan, Craig Frances, health events on work and earnings. program dramatically elevated the eco- and Laurence C. Baker find that patients Daniel L. McFadden has developed a nomic resources of many extended in geographic settings with a high share new research project on the dynamics families in South Africa, and not just of managed care patients, following a of health and wealth, building on the individuals directly eligible for the heart attack, are more likely to received methodological work done in coopera- pension. Case finds a large causal appropriate treatment with beta-block- tion with Peter Adams, Hurd, Angela effect of income on health status in ers and aspirin, but less likely to under- Merrill, and Tiago Ribeiro.6 With Hurd poorer populations — an effect that go coronary angiography [W8065]. In and Merrill, McFadden also has looked works at least in part through sanita- a similar study, Daniel Altman, David at the quantitative importance of vari- tion and living standards, in part M. Cutler, and Richard Zeckhauser ous predictors of mortality, including through nutritional status, and in part find almost no difference between income, wealth and education [W7440]. through the reduction of depression treatment intensity among individuals All of these investigations relate to the and psychosocial stress [W8495]. enrolled in HMO plans and those causal links between socioeconomic cir- Duflo finds larger effects on the health enrolled in indemnity insurance plans cumstances and health; the Aging of family members when a woman is in the same health care markets Program plays a significant role in inte- the recipient of the pension than when [W7832].5 The reason HMOs cost less, grating them in a coordinated program- a man is the recipient, suggesting dif- they conclude, is not because HMOs matic way. ferences in how income is allocated provide less care. Instead, about half within households [W8061]. of HMO cost savings is attributable to Health, Health Care, Research by Case, Darren Lubotsky, lower incidence of disease among and Paxson has focused on the childhood HMO enrollees; and about half results and Health Policy origins of the relationship between from HMOs paying lower prices for socioeconomic status and health the same treatments. Health, Disability and [W8344]. Using numerous data Where managed care may make a Mortality sources, they find that health is posi- difference, however, is in the medical tively related to household income even practice patterns used in geographical Many have worried that increas- in childhood. The relationship between regions that have more or less man- ing longevity would create a new bur- household income and children’s health aged care penetration, and in the den of health and long-term care status becomes more pronounced as spillover effects of cost containment needs for an increasingly sick and dis- children grow older, as the adverse on the provision of unreimbursed abled elderly population. While popu- health effects of lower income accu- care. According to Cutler, one implica- lation aging has created additional mulate over children’s lives. Thus, part tion of the expansion of managed care health and long-term care needs, the

6. NBER Reporter Summer 2003 potential cost has been moderated — 1988 and predict even sharper declines the proportion of the population that at least in the short run — by a signif- in older age mortality between 1988 dies in the hospital, a trend that might icant long-term decline over time in and 2022. Another study finds that the have been expected to reduce costs. the functional limitations of older peo- immediate effects of reduced infec- Instead, while there has been a decline ple. This trend has been documented tious disease rates and reduced mortal- in hospitalization rates over the 1988- in numerous data sources. However, ity from acute disease account for 62 95 period, the use of intensive care and much less is known about the rate, percent of the 20th century increase in intensive inpatient procedures has con- acceleration, character, causes, and survival rates, and the long-term tinued to increase, offsetting any consequences of the decline. Thus the effects account for another 12 percent potential cost savings. study of disability has become a new of the increase [W8000]. This line of Another set of studies has looked high priority area of research in the research has provided a valuable his- at trends in the treatment of specific Aging Program. torical perspective on more recent health conditions, such as acute In some initial work on this topic, health and disability trends. myocardial infarction or liver disease Cutler has conducted an overview of Alan M. Garber, Jay Bhattacharya, or ventricular arrhythmia.9 Illustrative of the evidence on disability decline, and and Thomas E. MaCurdy also have ini- the findings, a study of implantable car- potential causal explanations.7 He tiated a new project on disability dioverter-defibrillators finds that their quantifies the rate of disability decline decline, focusing on medical care for use, at a cost of up to $50,000 for the at 1 percent or more per year for the the disabled elderly. This work will procedure and treatment of complica- past several decades. Among the explore the individual patterns of dis- tions, grew more than 10-fold among potential explanations of disability ability decline (and improvement), the Medicare beneficiaries between 1987 decline that Cutler cites are medical persistence of disability status of indi- and 1995. Increasing use of bypass sur- care improvements, improved health viduals over time, the inter-temporal gery and angioplasty, the more intensive behaviors, increasing use of assistive links between disabilities and chronic approaches to heart attack treatment, devices and improvements in living diseases, and the medical utilization also have increased rapidly, raising conditions that increase independence and expenses of individuals before and overall expenditure on heart attack with functional limitations, higher levels after the development of disability, and treatment, despite a drop in the cost of of education and improved socioeco- as individual functional ability evolves any particular approach to treatment. nomic status, reductions in disease over time. The increasing technological intensity exposure, and social supports. He and I of medicine dominates almost all other are engaged in an ongoing research sources of health care cost growth, effort to better understand disability Trends in Treatment, including demographics, decreasing decline, its causes, the potential of inter- Utilization, and Costs mortality, decreasing hospitalization ventions to extend it, and the cost impli- rates, decreasing chronic illness rates, cations of the decline for Medicare A diversity of health care and and other factors. expenditures. A related project deals health policy issues have been studied Complementing these studies of with the role of health behaviors as a as part of the Aging Program in recent treatment trends are a collection of determining factor of health and func- years. These include studies of trends studies on the incremental benefits of tional ability. in utilization, the composition of med- intensive treatment, and the changing Dora L. Costa also has been ical care costs in different health insur- productivity of medical care.10 A studying trends in chronic illness, func- ance programs, the persistence of indi- recent study of angioplasty by Cutler tional disability, and mortality. Her vidual health care expenditures over and Robert S. Huckman is representa- focus has been on longer-term histori- time, the role of technology change in tive of this work [W9311]. The use of cal trends, and the historical underpin- medical practices as a key factor in angioplasty has spread dramatically nings of more recent health trends. For medical expenditure growth, the over the 20 years since its introduction. example, Costa finds that functional increasingly disproportionate spending It has raised the cost of treatment for disability (disability in walking, difficul- on medical care at older ages, the coronary artery disease. However, ty in bending, paralysis, blindness, and effectiveness of medical technology in because it now sometimes substitutes deafness) in the United States has fall- treating various health conditions, the for bypass surgery, a more expensive en at an average annual rate of 0.6 per- effects of different policy provisions procedure, there is some offset in costs cent among men age 50 to 74 from the in containing health care costs, and the with the same or better health out- early twentieth century to the early differences in access to medical care comes as bypass surgery. And where 1990s [W7605]. Another recent study across the population. angioplasty has replaced non-intensive by Costa looks at the socioeconomic One line of research, involving medical management, the value of the and demographic determinants of Garber, MaCurdy, McClellan, and oth- technology is likely enough to justify frame size using a data set of Civil War ers, has focused on the one-third of the cost. Indeed many of the analyses soldiers [W8843]. Costa finds that Medicare expenditures that are spent of medical productivity by NBER changes in frame size explain about on patients near the end of life.8 In investigators find that the increasing three-fifths of the mortality decline recent years, hospice care and other use of intensive procedures has done a among white men between 1915 and out-of-hospital treatment has decreased lot to improve health, on average,

NBER Reporter Summer 2003 7. although the procedures may be less the MSA account are treated as long- Database Development effective in marginal patients. term retirement saving. A catastrophic insurance plan covers any expenditures in Aging above this deductible. Such schemes are Health Insurance, Hospital Many NBER investigators have Organization, and Health designed to provide consumers — and their health care providers — incentives been involved in the development of Policy Reform not to spend money on care which data on older people in the United offers only low marginal benefit. Our States and abroad. The Health and In large part as a response to ris- research has considered both the desir- Retirement Survey (HRS) began about ing costs, the health insurance and able incentive features of MSA plans a decade ago and now provides longi- health policy environment in the and the potential limitations that might tudinal information on the health and United States has continued to evolve. result for individuals with long-term economic circumstances of about Cutler has explored the decline in continuing health care expenditures. 25,000 older Americans from 1992 to health insurance coverage among We find that only a small fraction of the present. NBER affiliates Charles working families in the United States households have continuing medical Brown, Gustman, Hurd, and Mitchell in the 1990s, finding that coverage expenditures year after year at a level are members of the HRS management declined primarily because fewer work- that would overwhelm their MSA bal- team; David O. Meltzer, John Rust, and ers took up coverage when offered it, ances. Indeed our most recent work Skinner are on the HRS Steering not because fewer workers were on the project shows how MSA plans Committee; and Cutler, David I. offered insurance or were eligible for it might work in conjunction with other Laibson, and I are on the NIA Data [W9036].11 Project estimates suggest personal retirement accounts as a Monitoring Committee for the HRS. that increased costs to employees can means of financial preparation for The HRS project has developed and explain the entire decline in take-up retirement. Based on actual medical applied numerous survey innovations, rates in the 1990s. histories, our simulations suggest that including the use of bracketing ques- Kessler and McClellan have been more than half of plan enrollees par- tions to minimize non-response, the engaged in a project on the effects of ticipating in an equity-invested MSA use of experimental modules for con- hospital organizational form on med- program over a working career would tinuous database development, the ical productivity. A recent study from accumulate MSA balances of over inclusion of data on expectations, the this project finds that areas with a for- three times the amount contributed to linkage of survey data to administra- profit hospital have about 2.4 percent the plans. And very few would accu- tive records from the Social Security lower hospital expenditures, but virtu- mulate less than 100 percent of their Administration and the Centers for ally the same patient health outcomes contributions, even after paying for Medicaid and Medicare Services, and [W8537]. Kessler and McClellan attrib- non-catastrophic health care costs. the inclusion and coding of pension ute this to the likely spillover effects of Thus a key finding from the project is and health plan data obtained directly for-profit hospitals on their nonprofit that MSA plans may not only be effec- from firms. Many participants in the and public counterparts: the competi- tive in containing health care costs; Aging Program have assisted in devel- tion from for-profit hospitals may limit they may also work in conjunction with oping, analyzing, and improving these the ability of non-profits to behave other programs to increase pre-retire- innovative components. inefficiently. In another study, Kessler, ment saving. McFadden has been a leader in McClellan, and Henry Hansmann com- Cutler also has considered the the study of survey response bias and pare the responsiveness of hospitals to dynamics of international health poli- in the development of internet survey reductions in demand, finding that for- cy reform.13 One study finds that as methodologies. A major accomplish- profit hospitals respond most rapidly, new and expensive medical technolo- ment of this work has been to imple- followed in turn by public and reli- gies have developed over time, the pol- ment an experimental internet survey giously affiliated non-profit hospitals, icy commitment to equal access to care method for data collection, adminis- while secular non-profits are distinctly has become ever more expensive. tered through an internet virtual labora- the least responsive ownership form Historically, countries have dealt with tory, or IVLab, developed explicitly for [W8989]. rising costs by maintaining equal the project. The IVLab has confirmed Matthew Eichner, McClellan, and access and restricting total spending. the value of internet-based question- I have considered the incentives, char- Today, many countries are considering ing as a low-cost survey methodology acteristics, and potential limitations of a move away from spending controls that enables cost-effective experimen- medical savings accounts (MSAs) as an and toward incentive-based medical tation with questionnaire formatting alternative approach to firm health care reform — possibly inducing more and survey content — experimenta- insurance provision.12 Under these cost-effective health care decisions, but tion that is generally cost-prohibitive plans, a specified amount of money is also greater reductions in care among using other survey approaches. deposited each year in an employee those less able to afford the incentive- McFadden’s work also has been at the MSA, and these funds are available to based cost-sharing provisions of the cutting edge in exploring how the for- support non-catastrophic medical care plans. mat, sequencing, and context of ques- expenses as needed. Unused assets in tioning affects responses, and how to

8. NBER Reporter Summer 2003 effectively correct (or at least correctly research on the effects of the pension 3. J. Gruber and D.A. Wise, eds., Social interpret) biases that result from these system in South Africa. Aging Security and Retirement Around the survey limitations. In addition to the Program investigators also have been World, University of Chicago Press, 1999; J. methodological advances made through involved, peripherally at this point, in Gruber and D.A. Wise, eds., Social Security the IVLab, McFadden and colleagues discussions with a larger consortium of Programs and Retirement around the have completed both a pilot and a larg- research and data sites in developing World: Micro-Estimation, Chicago: er-scale version of an internet-based regions throughout Africa and parts of University of Chicago Press, forthcoming. survey of older Americans, called the Asia (the INDEPTH network). Duflo 4. J. Gruber and D.A. Wise, “Social Security Retirement Perspectives Survey (RPS). is involved in database development in and Retirement Around the World: The content of the RPS survey draws areas of poor health in various regions Introduction and Summary,” in S. Polacheck heavily on the asset and health compo- of India. And Jensen has extended his and J. Robst, eds., Research in Labor nents of HRS, but with substantially research and database development Economics, Vol 18, JAI Press Inc., 1999; more variation in questionnaire format, work on health and economic circum- J. Gruber and D.A. Wise, “Social Security, sequencing, cues and context. These stances in Russia and Eastern Europe Retirement Incentives, and Retirement experimental treatments have enabled to developing countries in Asia, includ- Behavior: An International Perspective,” in the research team to test order effects ing India and Pakistan. A.J.Auerbach and R.D. Lee, eds., and other framing effects on subject For almost two decades, the Demographic Change and Fiscal responses. The larger-scale RPS survey NBER Aging Program has focused Policy, Cambridge: Cambridge University also includes both mail-out and inter- widespread attention on population Press, 2001; J. Gruber and D.A. Wise, “An net versions of the survey, so investi- aging, and the health and economic International Perspective on Policies for an gators can begin to assess how both circumstances of individuals as they Aging Society,” in S. Altman and D. sample selection and response biases age. It has also worked well in integrat- Schactman, eds., Policies for an Aging differ between them, as well as ing a wide range of related research Society: Confronting the Economic between these experimental surveys projects into a cohesive program, and Political Challenges, Baltimore: Johns and the HRS itself. including regular interaction among Hopkins Press, 2002; J. Gruber and D.A. The analysis of survey response members of the research team, exten- Wise, “Different Approaches to Pension bias has continued to be a key element sive dissemination of re-search find- Reform from an Economic Point of View,” of this work. For example, a recent ings in both academic publications and in M. Feldstein and H. Siebert, eds., Social study by Li Gan, Hurd, and McFadden non-technical reports, external involve- Security Pension Reform in Europe, looks at people’s mortality expecta- ment in promoting aging-related Chicago: University of Chicago Press, 2002. tions, which are very close to actual research and data resource develop- 5. See also D.M. Cutler and J. Seinfeld, mortality risks on average, but which ment in aging, organizing international “Managed Care Enrollment and Care for the have problematic characteristics, such collaborations and cross national com- Poor,” working paper, 2003. as unrealistic focal point responses parisons of aging issues, sponsoring 6. P. Adams, M.D. Hurd, D.L. McFadden, [W9480]. The methodological research regular workshops and conferences on A. Merrill, and T. Ribeiro. “Healthy, being done by Hurd, McFadden, and the economics of aging, and inspiring Wealthy, and Wise? Tests for Direct Causal others is helping to explain these the collaborative engagement of both Paths between Health and Socioeconomic anomalies, correct for them whenever senior scholars and new investigators Status,” Journal of Econometrics, Vol. possible, and appropriately interpret in the study of aging issues. 112, (1) (January 2003), pp. 3-56. research that uses these data. 7. D.M. Cutler, “Declining Disability Among Yet another exciting area of data- 1 These volumes include The Economics of the Elderly,” Health Affairs, 20 (6) base development has been in the Aging (1989), Issues in the Economics (November/December 2001), pp. 11-27. development of international data on of Aging (1990), Topics in the 8. A.E. Barnato, A.M. Garber, C.R. Kagay, aging. Axel Börsch-Supan is the coor- Economics of Aging (1992), Studies in and M.B. McClellan, “Trends in the Use of dinator of the Survey of Health, the Economics of Aging (1994), Intensive Procedures at the End of Life,” in Aging, and Retirement in Europe Advances in the Economics of Aging A. M. Garber, ed., Frontiers in Health (SHARE). SHARE is a coordinated (1996), Inquiries in the Economics of Policy Research, Vol. 4, Cambridge, MA: data collection effort in Denmark, Aging (1998), Frontiers in the MIT Press, 2001; J. Geppert and M. B. France, Germany, Greece, Italy, the Economics of Aging (1998), Themes McClellan, “Trends in Medicare Spending Nether-lands, Spain, and Sweden. It in the Economics of Aging (2001), and near the End of Life,” in D. A. Wise, ed., evolved in part from the HRS, and from Perspectives on the Economics of Themes in the Economics of Aging, the NBER project on social security Aging (forthcoming). Chicago: University of Chicago Press, 2001; systems and retirement around the 2. A. Samwick and D.A. Wise, “Option A.M. Garber, T.E. MaCurdy and M.B. world, and a number of NBER inves- Value Estimation with Health and Retirement McClellan, “Medical Care at the End of tigators (in addition to Börsch-Supan) Survey Data,” in S. Ogura, T. Tachibanaki, Life: Diseases, Treatment Patterns, and are advisors to the SHARE project. In and D.A. Wise, eds., Labor Markets and Costs,” in A.M. Garber, ed., Frontiers in other parts of the world, Case, Deaton, Firm Benefit Policies in Japan and the Health Policy Research, Vol. 2, and others have been involved in United States, Chicago: University of Cambridge, MA: MIT Press, 1999. designing surveys and conducting Chicago Press, forthcoming. 9. J.H. Best, J. Geppert and D. L. Veenstra,

NBER Reporter Summer 2003 9. “Trends in Expenditures for Medicare Liver 2001. paper, 2003. Transplant Recipients,” Liver Transplan- 10. D.M. Cutler and M.B. McClellan, 12. M. Eichner, M.B. McClellan and D.A. tation, 2001; N. Every, A.M. Garber, P. “Productivity Change in Health Care,” Wise, “Individual Expenditures and Medical Heidenreich, M. Hlatky, D. P. Kessler, M.B. Amercian Economic Review, 2001; Savings Accounts: Can They Work?” in S. McClellan, J.P. Newhouse and O. Saynina, D.M. Cutler and M.B. McClellan, “Is Ogura, T. Tachibanaki, and D.A. Wise, “Technological Change in Heart Attack Care Technological Change in Medicine Worth It?” eds., Labor Markets and Firm Benefit in the United States: Causes and Health Affairs, 2001; J.M. Brooks, M.B. Policies in Japan and the United States, Consequences,” in M.B. McClellan and D.P. McClellan, and H.S. Wong, “The Marginal Chicago: University of Chicago Press, forth- Kessler, eds., Technological Change in Benefits of Invasive Treatments for Acute coming; M. Eichner and D.A. Wise, “Little Health Care: A Global Analysis of Myocardial Infarction: Does Insurance Saving and Too Much Medical Insurance: Heart Attack, Ann Arbor, MI: University Coverage Matter?” Inquiry, 2000; A. Medical Savings Accounts Could Help,” in of Michigan Press, 2002; P. Heidenreich and Brown and A.M. Garber, “A Concise D. A. Wise, ed., Personal Saving, Personal M.B. McClellan, “Trends in Heart Attack Review of the Cost-Effectiveness of Coronary Choice, Stanford: Hoover Institution Treatment and Outcomes, 1975-95: Heart Disease Prevention,” Medical Clinics Press, 1999. Literature Review and Synthesis,” in D.M. of North America, 2000; W. Browner, C. 13. D. M. Cutler, “Equality, Efficiency, and Cutler and E.R. Berndt, eds., Medical Care Frances, B. Massie, M.B. McClellan, H. Market Fundamentals: The Dynamics of Output and Productivity, Chicago: Noguchi, and M.G. Shlipak, “Comparison International Medical Care Reform,” University of Chicago Press, 2001; P. of the Effects of Angiotensin Converting- Journal of Economic Literature, 2002; Heidenreich and M.B. McClellan, “Trends in enzyme Inhibitors and Beta Blockers on D.M. Cutler, “The Third Wave in Health Treatment and Outcomes for Acute Myocardial Survival in Elderly Patients with Reduced Care Reform,” in S. Ogura, T. Tachibanaki Infarction: 1975-1995, American Journal Left Ventricular Function after Myocardial and D.A. Wise, eds., Aging Issues in the of Medicine, 2001; D. P. Kessler and M.B. Infarction,” American Journal of United States and Japan, Chicago: McClellan, “A Global Analysis of Medicine, 2001; C. Frances, M.G. University of Chicago Press, 2001; D.M. Technological Change in Health Care: The Shlipak, H. Noguchi, P.Heidenreich, and Cutler, “Supplementing Public Insurance Case of Heart Attacks,” Health Affairs, M.B. McClellan, “Does Physician Specialty Coverage with Private Coverage: Implications 1999; M.B. McClellan, “Biomedical Research Affect Survival of Elderly Patients with for Medical Care Systems,” in S. Ogura, T. and Then Some: the Causes of Technological Myocardial Infarction?” Health Services Tachibanaki, and D.A. Wise, eds., Labor Change in Heart Attack Care,” in K. Research, 2000; A.M. Garber, “Advances Markets and Firm Benefit Policies in Murphy and R.H. Topel, eds., The Value of in Cost-Effectiveness Analysis of Health Japan and the United States, Chicago: Biomedical Research, Chicago: University Interventions,” in A. Culyer and J.P. University of Chicago Press, forthcoming. of Chicago Press, forthcoming; A.M. Garber, Newhouse, eds., Handbook of Health M. Hlatky, M.B. McClellan, K. McDonald, Economics, North-Holland, 2000. and O. Saynina, “Utilization and Outcomes of 11. See also M. Chernew, D.M. Cutler and P. the Implantable Cardioverter Defribrillators: Keenan, “Rising Medical Costs and the 1987-1995”, American Heart Journal, Decline in Insurance Coverage,” working

*

10. NBER Reporter Summer 2003 Research Summaries

Randomized Trials and Quasi-Experiments in Education Research Joshua D. Angrist*

The 2001 No Child Left Behind coin). Quasi-experimental research that other researchers can reproduce (NCLB) Act promises a series of sig- designs are based on naturally occur- and interpret. This is what science is nificant reforms. The hope is that ring circumstances or institutions that all about. In contrast, U.S. education these reforms will jump-start under- (perhaps unintentionally) divide people policy has often relied on evidence that performing American schools. Most into treatment and control groups in a is fragmentary or anecdotal, uses sub- public discussion of the Act has manner akin to purposeful random jective outcomes, and, most impor- focused on the mandate for test-based assignment. tantly, fails to make rigorous compar- school accountability and the federal A reliance on control groups and isons of treatment and control groups. endorsements of charter schools and random assignment indeed would If successful, a shift to scientifical- other forms of school choice. Other mark a new direction for education ly based research will move the study important provisions include changes research. For example, an important of education much closer to medicine, in funding rules for states and a new question on the education research which has been experiencing a similar emphasis on reading instruction. The agenda is the role of technology in transition to scientifically based NCLB Act also repeatedly calls for schools. Most previous research on research over the last half-century. education policy to rely on a founda- the use of technology in the classroom NBER researchers have been in the tion of scientifically based research. (computer-aided instruction or CAI) vanguard of this transition to scientifi- Although this appears to be a bland relies on uncontrolled measurements, cally based research on education. We technical statement, it strikes me as such as the level of satisfaction experi- have used natural experiments — and potentially at least as significant as enced by technology users. Not sur- in some cases, actual randomized trials other components of the Act. prisingly, teachers and students typical- — to provide powerful evidence on ly report that they enjoy using new issues ranging from the effects of computer equipment (as shown in a compulsory attendance laws to chang- What is Scientifically recent study of laptops in Maine’s pub- ing class size. I describe some of this Based Research? lic schools). But this does not establish work below, focusing on my own that students who use the laptops are efforts. I have used quasi-experiments NCLB defines scientifically-based learning more, or that the expenditure — and in recent and ongoing projects, research as research using rigorous on computers meets a cost-benefit randomized trials — to make scientifi- methodological designs and tech- standard (after all, computer hardware cally grounded inferences regarding niques, including control groups and and software is expensive). the effects of achievement incentives random assignment. In a presentation Randomized trials provide the best and school choice, school resources, made shortly after President Bush scientific evidence on the effects of and macro education policy. signed NCLB into law, the deputy policies like educational technology, director of the Office of Research in changes in class size, or school vouch- School Incentives and the Department of Education put ers because differences between the studies involving randomized trials and treatment and control group can be School Choice quasi-experiments at the top of the attributed confidently to the treatment. methodological hierarchy. A good quasi- or natural experiment is The desire to help disadvantaged Randomized trials are experiments the next best thing to a real experi- teens get through high school is a in which the division into treatment ment. In some cases, quasi-experi- recurring theme of school reform pro- and control groups is determined at ments also involve random assign- posals. Most anti-dropout efforts random (for example, by tossing a ment, such as in the lotteries some- involve the provision of support serv- times used to distribute school vouch- ices to low-achieving students. But the ers. In addition to comparing apples results from recent demonstration * Angrist is a Research Associate in the to apples, randomized trials and natu- projects assessing services for at-risk NBER’s Program in the Economics of ral experiments also rely on assess- high-school students have been disap- Education and a Professor of Economics at ments by disinterested non-partici- pointing.1 Motivated by the economic MIT. His profile appears later in this issue. pants and on clearly defined outcomes view of education, which sees student

NBER Reporter Summer 2003 11. effort in school as determined partly paper,4 my collaborators and I studied example, NCLB singles out the stu- by a comparison of the costs and ben- what appears to be the largest school dents in the worst schools as being eli- efits of effort devoted to schooling, voucher program to date. This pro- gible for choice. In particular, NCLB Victor Lavy and I developed a unique gram provided over 125,000 pupils requires districts to allow students in program that rewards Israeli high from poor neighborhoods in the coun- schools judged to be “failing” the school students who pass their high try of Colombia with vouchers that opportunity to change schools. school matriculation exams (some- covered approximately half the cost of Policymakers and parents in the schools thing like the New York Regents exam private secondary school. Colombia is that accept these students have won- or British A Levels) with cash pay- an especially interesting setting for dered what the consequences will be ments. Although this project was con- testing the voucher concept because for high-achieving children when low troversial in Israel (and eventually was private secondary schooling in achievers from poor areas choose to cancelled as a political liability), it is in Colombia is a widely available and attend their schools. Economists refer the spirit of a 1998 proposal by former often inexpensive alternative to crowd- to research on questions of this sort as Labor Secretary Reich, who suggested ed public schools. (In Bogota, over the measurement of peer effects. that students from low- income fami- half of secondary school students are Boston’s long-running Metco pro- lies in the United States be offered a in private schools.) Moreover, govern- gram provides a unique opportunity to $25,000 cash bonus for graduating ments in many poor countries are estimate peer effects in the classroom from high school. It is also similar to increasingly likely to experiment with using a quasi-experimental research the merit-based stipends common in demand-side education finance pro- design. Metco gives mostly black stu- higher education. grams, including vouchers. dents in the Boston public school dis- Perhaps most unusually, Lavy and I Although not a randomized trial, a trict the opportunity to attend schools implemented the Achievement Awards key feature of our Colombia study is in more affluent suburban districts. program as a school-based random- the exploitation of voucher lotteries as Kevin Lang and I focus on the impact ized trial.2 That is, we identified 40 of the basis for a quasi-experimental of Metco on the students in one of the lowest-achieving schools in Israel, research design. Because demand for the largest Metco-receiving districts.6 and randomly selected 20 of them for vouchers exceeded supply, the avail- Because Metco students have substan- participation in the program. Any stu- able vouchers were allocated by lottery tially lower test scores than local stu- dent from the 20 treatment schools in large cities. Our study compares dents, this inflow generates a signifi- who passed their exams was eligible voucher applicants who won a vouch- cant decline in average scores. Our for a $1500 payment, quite a large sum er in the lottery to those who lost. research shows that the overall decline in Israel, although small relative to the Since the lotteries used random assign- in scores is attributable to a composi- private and social costs of dropping ment, losers provide a good control tion effect, though, because we find no out. Only about 18 percent of stu- group for winners. A comparison of impact on average scores in a sample dents in the control group completed voucher winners and losers shows that limited to non-Metco students. This their matriculation exams. Students in three years after the lotteries were weighs against the hypothesis of sig- our control group were about 7 per- held, winners were 15 percentage nificant negative peer effects as a result centage points more likely to complete points more likely to have attended of school choice (although we do find their matriculation exams, a statistical- private school and were about 10 per- a short-lived negative effect on the ly significant difference with an eco- centage points more likely to have fin- scores of minority third graders in nomic benefit that easily outweighs the ished eighth grade, primarily because reading and language). Our research cost of bonus payments. they were less likely to repeat grades. on Metco exploits idiosyncratic fea- The Achievement Awards demonstra- Lottery winners also scored 0.2 stan- tures of the process used to allocate tion is the first of what we hope will be dard deviations higher on standardized Metco students to different schools a series of randomized trials designat- tests. A follow-up study in progress through what is known as the “regres- ed to test education incentive plans. In shows that voucher winners also were sion-discontinuity” method for analy- research in progress, Lavy and I are more likely to apply to college. On sis of quasi-experiments. evaluating a package of incentives that balance, our study provides some of provides awards for teachers as well as the strongest evidence to date for the for students. A unique feature of our possible benefits of demand-side School Resources ongoing work is that the new demon- financing of secondary schooling, at stration project includes a component least in a developing country setting.5 Another strand of my work uses specifically designed to explore the Research on vouchers naturally quasi-experiments to look at what interaction of student and teacher focuses on the question of whether economists call the education produc- incentives.3 We also plan a long-run voucher recipients benefit from the tion function. This research links follow up study of the Achievement opportunity to use vouchers. A relat- school resources, including computers Awards program. ed question that gets less attention and class size, with outcomes such as One of the most controversial arises from the fact that voucher recip- student achievement on standardized innovations highlighted by NCLB is ients and other school choice benefici- tests. The principal challenge in school choice. In a recently published aries are typically low-income. For research of this type, as in most empir-

12. NBER Reporter Summer 2003 ical research in economics, is in isolat- discusses rules for the determination of experiments to study legislative and ing cause and effect. Many factors class size and pupil-teacher ratios in other macro-level education policies. make the observed correlation between bible study. The great 12th century Here, experiments are harder to come school resources and student achieve- Rabbinic scholar, Maimonides, inter- by and research may have to rely on ment hard to interpret. Rich and poor prets the Talmud’s discussion of class simple policy shifts that affect some districts differ on many dimensions, size as follows: “Twenty-five children states and not others. Nevertheless, teachers sort students into classes of may be put in charge of one teacher. If this work follows the natural-experi- different size, and students and parents the number in the class exceeds twenty- ments model in that there is always a make systematic choices that are five but is not more than forty, he well-defined control group. For exam- reflected in the resources/achievement should have an assistant to help with ple, Jon Guryan and I recently looked relationship. the instruction. If there are more than at state changes in teacher certification The question of how technology forty, two teachers must be appointed.” requirements.11 We find that states that affects learning has been at the center In my first study of school introduced teacher tests (such as the of recent debates over educational resources, also joint with Lavy,9 we use national teachers examination) ended inputs. My most recent research on Maimonides’ rule capping class size at up paying higher teacher salaries with school resources, joint with Lavy,7 40 to construct a natural experiment to no measurable increase in teacher qual- exploits a natural experiment arising estimate the effects of class size on the ity. This suggests that tests are more from the fact that the Israeli State lot- scholastic achievement of Israeli of a barrier to entry than an effective tery, which uses lottery profits to spon- pupils. To see how this experiment quality screen. sor various social programs, funded a works, note that according to In earlier work, Alan B. Krueger large-scale computerization effort in Maimonides’ rule, class size increases and I looked at the effects of compul- many elementary and middle schools. one-for-one with enrollment until 40 sory attendance laws.12 This research Although lottery officials did not use pupils are enrolled, but when 41 stu- exploits the interaction between indi- random assignment to allocate the dents are enrolled, there will be a sharp viduals’ quarter of birth and state laws computers across towns and schools, drop in class size, to an average of 20.5 (children born earlier in the year are they used an idiosyncratic priority pupils. Similarly, when 80 pupils are allowed to drop out of school after scheme that appears to have an essen- enrolled, the average class size will having completed less schooling than tially random component. We used this again be 40, but when 81 pupils are those born later). More recently, to estimate the impact of computeriza- enrolled the average class size drops to Daron Acemoglu and I have used state tion on both the instructional use of 27. Our use of this variation is an compulsory attendance laws to esti- computers and pupil achievement. application of the quasi-experimental mate the social returns to education Results from a survey of Israeli regression-discontinuity method. (that is, an economic benefit beyond school-teachers show that the influx of Interestingly, the observed associa- that accruing to the more educated new computers increased teachers’ use tion between class size and student individuals themselves).13 I also have of CAI in the fourth grade, with a achievement in our data is always per- looked at natural experiments increas- smaller effect on CAI in eighth grade. verse (that is, students in larger classes ing the education infrastructure, for Perhaps surprisingly, CAI does not tend to do better). But this illustrates example a large-scale expansion of appear to have had educational bene- the importance of research using a higher education in the West Bank and fits that translated into higher test good experiment. Estimates of class Gaza Strip.14 Finally, Lavy and I studied scores. In fact, estimates for fourth size effects using Maimonides’ Rule the economic consequences of the graders show lower math scores in the suggest that reductions in class size change in language of instruction in group that was awarded computers, induce a significant and substantial Morocco’s secondary schools.15 with smaller (insignificant) negative increase in math and reading achieve- effects on language scores. These ment for fifth graders, and a modest results call into question the widely- increase in reading achievement for Conclusion held view that additional resources fourth graders. We gain confidence is should be devoted to CAI.8 this result (as opposed to the simple In addition to providing evidence Another central question in the correlation between class size and test on specific questions, I believe that an school resources debate is the impor- scores) because a randomized trial important overall contribution of my tance of class size. Although recent manipulating class size in Tennessee work on education has been to docu- years have seen renewed interest in the generated similar estimates.10 ment the feasibility and promise of class-size question, academic interest in both quasi- experimental methods and this topic is not simply a modern phe- randomized trials in education research. nomenon: the choice of class size has The Effects of Macro- Many other NBER researchers are also been of concern to scholars and teach- Education Policy involved in this work and I expect that ers for hundreds of years. One of the education research along these lines will earliest references on this topic is the The work discussed above focuses be a growth area for economists in the Babylonian Talmud, completed around on a micro-level analysis of students years to come. I am certainly looking the beginning of the 6th century, which and schools. I have also used natural forward to doing more of it.

NBER Reporter Summer 2003 13. 1 M. Dynarski and P. Gleason, How Can Economic Review, 90 (5) (December 11 J. D. Angrist and J. Guryan, “Does We Help? What Have We Learned 2000), pp. 1209-38 is a quasi-experimental Teacher Testing Raise Teacher Quality? from Evaluations of Federal Dropout- study of school choice. In work in progress, J. Evidence from State Certification Prevention Program, Mathematica Policy B. Cullen, S. D. Levitt, and B. A. Jacob are Requirements,” NBER Working Paper No. Research report 8014-140, Princeton, NJ, using lotteries to study school choice in the 9545, March 2003. June 1998. Chicago public schools in, “The Impact of 12 J. D. Angrist and A. B. Krueger, “Does 2 J. D. Angrist and V. Lavy, “The Effect of School Choice on Enrollment and Compulsory School Attendance Affect High School Matriculation Awards” Achievement: Evidence from over 1000 Schooling and Earnings?” Quarterly Evidence from Randomized Trials,’ NBER Lotteries,” manuscript, 2003 (forthcoming as Journal of Economics, 106 (November Working Paper No. 9389, December 2002. an NBER Working Paper). 1991), and “The Effect of Age at School 3 See P. Glewwe, N. Ilias, and M. Kremer, 6 J. D. Angrist and K. Lang, “How Entry on Educational Attainment: An “Teacher Incentives,” NBER Working Important are Classroom Peer Effects? Application of Instrumental Variables with Paper No. 9671, May 2003, for a recent Evidence from Boston’s METCO Program,” Moments from Two Samples,” Journal of randomized trial of teacher incentives in NBER Working Paper No. 9263, October the American Statistical Association, Kenya. 2002. (June 1992). 4 J. D. Angrist, E. P. Bettinger, E. Bloom, 7 J. D. Angrist and V. Lavy, “New Evidence 13 J. D. Angrist and D. Acemoglu, “How E. King, and M. Kremer, “Vouchers for on Classroom Computers and Pupil Large are the Social Returns to Education? Private Schooling in Colombia: Evidence Learning,” The Economic Journal, 112 Evidence from Compulsory Attendance from a Randomized Natural Experiment,” (October 2002), pp. 735-65. Laws,” NBER Macro Annual, 15, NBER Working Paper No. 8343, June 8 My MIT colleagues Abhijit Banerjee and Cambridge, MA: MIT Press, 2000. 2001, and in American Economic Esther Duflo are currently running a ran- 14 J. D. Angrist, “The Economic Returns to Review, 92 (5) (December 2002), pp. domized trial of CAI in India. Schooling in the West Bank and Gaza 1535-58. 9 J. D. Angrist and V. Lavy, “Using Strip,” American Economic Review, 85 5 Evidence on voucher effects for the United Maimonides’ Rule to Estimate the Effect of (5) (December 1995), pp. 1065-87. A relat- States has been more mixed. Two studies Class Size on Student Achievement,” ed and more recent study in this spirit is E. involving randomization are A. B. Krueger Quarterly Journal of Economics, 114 Duflo, “The Medium Run Effects of and P. Zhu, “Another Look at the New (2) (May 1999), pp. 533-75. Educational Expansion: Evidence from a York City Voucher Experiment,” NBER 10 A. B. Krueger, “Experimental Estimates Large School Construction Program in Working Paper No. 9418, January 2003 of Education Production Functions,” Indonesia,” NBER Working Paper No. and C. E. Rouse, “Private School Vouchers Quarterly Journal of Economics, 104 8710, January 2002. and Student Achievement: An Evaluation of (1999) pp. 497-532. But see also C. M. 15 J. D. Angrist and V. Lavy, “The Effect of the Milwaukee Parental Choice Program,” Hoxby, “Does Competition Among Public a Change in Language of Instruction on the The Quarterly Journal of Economics, Schools Benefit Students and Taxpayers?” Returns to Schooling in Morocco,” Journal 113 (2) (May 1998). C. M. Hoxby, “Does which finds little evidence of a class size effect of Labor Economics, 15 (January 1997) Competition Among Public Schools Benefit using quasi- experimental methods to analyze S48-S76. Students and Taxpayers?” American data from Connecticut.

*

14. NBER Reporter Summer 2003 Inflation Dynamics: Combining Measurement with Theory Jordi Galí and Mark Gertler*

Among the central issues in macro- shifts in trend productivity and individual firms that set prices for mul- economics is the nature of short-run changes in the monetary policy regime. tiple periods based on current and inflation dynamics. This matter is also anticipated future nominal marginal one of the most fiercely debated, with cost, and do so on a staggered basis. few definitive answers available after The New Keynesian The second key building block is an decades of investigation. At stake, Phillips Curve equation that has real marginal cost among other things, is the nature of vary proportionately with the output business fluctuations and the appropri- Our work builds on the optimiza- gap, where the latter is defined as the ate conduct of monetary policy. How a tion-based approach to modeling percent deviation of output from its central bank should go about engineer- short-run inflation dynamics that has natural (flexible price equilibrium) ing a disinflation, for example, depends been used increasingly in applied work level. This second relation holds critically on the extent to which: 1) in recent years. This literature, in turn, explicitly in the model under certain there may be a short-run tradeoff is an outgrowth of early theoretical auxiliary assumptions, including in par- between inflation and real activity and work by Fischer1, Taylor2, Calvo3 and ticular the assumption of competitive 2) expectations of future economic others that emphasized staggered labor markets. Intuitively, periods of activity affect current price setting nominal wage and price setting by for- excess demand (output above the nat- behavior. The issue is also highly rele- ward looking workers and firms. The ural level) are associated with marginal vant in the current era of low inflation: modern literature extends this earlier cost above average, and the reverse is how to manage monetary policy to work by casting the price setting deci- true for periods of excess supply. avoid deflation and potentially slipping sion within an explicit individual opti- Combining these fundamental rela- into a liquidity trap (of the type many mization problem. Aggregating over tions yields the baseline NKPC: infla- observers believe is happening in individual behavior then leads, typical- tion depends on the output gap and Japan) is similarly sensitive to how ly, to a relationship between inflation anticipated future inflation. Note that inflation is determined in the short run. in the short run and some measure of the NKPC has some of the flavor of a Our research over the past few overall real activity, in the spirit of the traditional Phillips curve in the sense years has focused on both theoretical traditional Phillips curve. The explicit that inflation varies positively in the and empirical analysis of inflation use of micro-foundations, of course, short run with the output gap. The dynamics. In contrast to much of the places additional structure on the rela- similarity stops there, however. The important traditional work on the tionship and further leads to some defining property of the baseline Phillips curve, which was largely important differences in detail. NKPC is the forward looking nature empirical in nature, we explicitly A canonical version of this kind of of inflation dynamics. The NKPC is employ economic theory to develop an Phillips curve — often referred to as effectively a first-order forward-look- econometric model of inflation. By the New Keynesian Phillips Curve ing difference equation in inflation tying the empirical analysis tightly to (NKPC) — is attributable to Calvo. with the output gap as the forcing vari- theory, we believe we are able to obtain This approach involves making able. Iterating this relation forward a deeper understanding of what deter- assumptions that greatly simplify the yields inflation equal to a discounted mines inflation in the short run than aggregation of individual price setting, sum of current and expected future would be the case from just examining but still retain the feature of non-syn- output gaps. Intuitively, the theory sug- statistical relationships. In addition, by chronized multi-period price setting. gests that price adjustment is based on estimating an explicit economic model, Because it results in a reasonably parsi- current and anticipated marginal cost. we can potentially understand how sig- monious aggregate relation for infla- Under certain conditions, the output nificant structural changes might affect tion, the model has gained widespread gap proxies movements in marginal inflation better than a mainly empirical attention. It also has generated some cost. Hence, inflation depends on the approach would permit. Examples of controversy, as we discuss below. expected future path of the output gap, significant structural changes include There are two basic building blocks as well as on its current value. to the Calvo variant of the NKPC. The This forward-looking behavior of first is an equation that relates current inflation contrasts sharply with the * Galí and Gertler are Research Associates in inflation to two factors: the percent traditional Phillips curve (TPC), the NBER’s Program on Monetary deviation of real marginal cost (aver- where inflation is an entirely back- Economics. Gali is a Professor of Economics aged across firms) from its steady state; ward looking phenomenon.4 With the at Universitat Pompeu Fabra; his profile and expected future inflation. This rela- TPC, inflation depends on the output appears later in this issue. Gertler is a Professor tion is obtained as a log-linear approxi- gap and an arbitrary number of lags of Economics at New York University. mation of the aggregated behavior of of inflation. In contrast to the NKPC,

NBER Reporter Summer 2003 15. the exact specification, of course, is not output gap tends to lead inflation.8 cal failure of the baseline model may explicitly guided by theory. Also, the One possibility is that detrended out- be that the output gap is not a good TPC rules out the possibility that put might not be a good proxy for the proxy for marginal cost, rather than beliefs about the future may affect cur- output gap, particularly if there is con- the canonical model of staggered for- rent price setting and, hence, inflation. siderable variation in the true natural ward price setting being incorrect. The extent to which the NKPC level of output. However, it appears This leads us to estimate the canonical captures reality has important policy that it would take a somewhat peculiar version of the model that links infla- implications. If expectations of the pattern of measurement error in the tion directly to real marginal cost.11 future matter, then current inflation natural level of output for the model Further, we allow for the possibility will depend not only on prevailing eco- to fully account for the timing evi- that price setting is not purely forward nomic conditions but also on beliefs dence on detrended output versus looking. In particular, in our paper we about future conditions, including the inflation observed in the data. introduced a fraction of firms in the future course of monetary policy. As a Mankiw9 emphasizes a closely relat- model that adjust prices using a simple consequence, establishing the credibil- ed criticism: The identified VAR litera- backward looking rule of thumb. The ity of intentions about the future ture suggests that an anticipated tight- rule has the property that it converges course of monetary policy becomes an ening of monetary policy produces a to the optimal forward looking pricing important dimension of its conduct. decline in output after two to three policy rule in the steady state. The net As recent literature has shown, with quarters, but no decline in inflation result is a hybrid of the NKPC that forward looking price setting, estab- until nearly a year after a shock. nests the pure forward looking model lishing a credible commitment to Simulations of models that use the as a special case.12 Under our hybrid maintaining price stability in the future NKPC may get the output response formulation, inflation depends on real reduces the cost of doing so in the correct, but produce too early and too marginal cost and a linear combination present. Roughly speaking, in this con- strong a response of inflation. Again, of expected future and lagged infla- text establishing credibility improves the problem is that inflation antici- tion.13 The model coefficients, further, the short-run output-inflation trade- pates the output gap in the NKPC but are explicit functions of the primitive off. Note that this gain from credibili- not in the data. More generally, the parameters, including the frequency of ty does not depend on having a central NKPC has difficulty accounting for price adjustment and the fraction of bank that wishes to push output above the apparent history dependence of firms that are backward looking. the natural level, as is critical in the inflation in the data. The TPC address- Lagged inflation disappears in the lim- early literature on credibility. Rather, it es this issue by simply adding lags of iting case where the fraction of rule-of- is a product of the forward-looking inflation to the right hand side of the thumb firms goes to zero and the nature of price setting.5 equation. Of course, there is no under- model converges to the pure forward The issue is also relevant to current lying theoretical motivation. looking NKPC. discussions of the liquidity trap. As We estimate our hybrid NKPC emphasized by Krugman,6 Eggertson using Generalized Method of and Woodford7 and others, with for- A Hybrid NKPC: Moments (GMM) with lagged vari- ward looking price setting, an econo- Development and ables as instruments (thus allowing for my constrained by the zero interest the possibility of measurement error.) rate bound may be able to nonetheless Structural Estimates Three principle findings emerge: 1) stimulate economic activity by com- the coefficient on real marginal cost is mitting to inflate in the future. Our work addresses the basic positive and statistically significant; 2) empirical shortcomings of the NKPC. the coefficient on lagged inflation is In Galí and Gertler10 we build on the positive and statistically significant, Early Criticisms of the basic Calvo model and develop a new implying that the pure forward looking NKPC econometric framework for estimating model is rejected by the data; 3) for- a structural equation describing the ward looking behavior is still dominant As is now well known, the baseline dynamics of inflation. We start from across a range of estimates. The coef- NKPC, which links inflation to the the view that the canonical version of ficients on expected future and lagged output gap and expected future infla- the model suggests that inflation inflation generally sum to one, with the tion, is at odds with the data. Estimates should be related to movements in real coefficient on lagged inflation ranging of this equation have the output gap marginal cost, and not to the output between 0.2 and 0.4. In subsequent enter either insignificantly or with the gap per se. It is only under the strong work with David López-Salido14,we wrong sign. The basic problem is that assumption of frictionless labor mar- broadly confirmed these estimates, the theory suggests that inflation kets that the output gap should vary although we have tightened the coeffi- should anticipate movements in the proportionately with marginal cost. cient estimates to roughly 0.35 for output gap. In the data, the reverse With labor market rigidities (for exam- lagged inflation and 0.65 for expected appears true, at least using ad hoc ple, stickiness in either nominal or real future inflation. In addition, we show measures of the output gap based on wages) this relation will not hold. Put that the simple theory-based model simple detrending methods, that is, the differently, one reason for the empiri- does a reasonable job of capturing the

16. NBER Reporter Summer 2003 movements in inflation over the past state-dependent model. with inflation. Given that firms are forty years, including: the rise to dou- As a check on whether this pricing based on current and anticipat- ble digit inflation in the 1970s; the dis- assumption is reasonable, we explored ed real unit labor costs, the sluggish- inflation of the early 1980s; and the the robustness of our estimates of the ness in this variable helps account for period of simultaneous high output degree of price rigidity to different the persistence in inflation. growth and low inflation during the sub-samples. We found the estimates Note too that this approach pro- latter half of the 1990s. to be reasonably stable, suggesting that vides a theoretically cogent way to cap- As some further confirmation on our assumption of time-dependent ture the influence of supply shocks on the empirical plausibility of our model, price setting may be plausible for a inflation. Here supply shocks (for we obtain sensible estimates of the country like the United States that has example, shift in oil prices or shift in degree of price rigidity. From our esti- experienced moderate variation in total factor productivity) affect infla- mates of the slope coefficients of the inflation. Clearly it would not be rea- tion by influencing the measure of hybrid NKPC, we can use the underly- sonable for a country that has experi- firms’ real marginal cost. In the Cobb- ing theory to back measures of how enced high and volatile inflation, such Douglas case, these shocks affect mar- long prices are fixed on average. Our as Brazil, or Argentina. ginal cost by changing average labor early work suggested a period of 4 to 5 Overall, our estimates of the productivity. For example, in the late quarters, which is high relative to the hybrid NKPC appear reasonably 1990s, the surge in productivity (in survey evidence. In subsequent work robust to different estimation strate- conjunction with only mild real wage we found that after relaxing some of gies.17 For example, our results are growth) led to a decline in unit labor the technological assumptions of our largely invariant to estimating the costs relative to trend. For this reason, model, our estimates of the average closed form of the model (obtained by our model is able to capture the low duration of a price being fixed fell to solving out for expected inflation). inflation of this period reasonably well. about two and a half quarters, clearly Nor are they the product of weak A complete story of short-run in the range of what recent survey evi- instruments. In addition, while we inflation dynamics requires modeling dence suggests.15 When applying the have used a single equation/instru- the evolution of real marginal cost. It same approach to euro area data, we mental variable approach, a number of appears that much of the persistence obtain similar results: the hybrid ver- other papers have obtained very simi- in real marginal cost (at least as meas- sion of the NKPC seems to fit the lar results to ours using a full blown ured by real unit labor cost) is associat- European data equally well, if not bet- systems approach.18 ed with sluggishness in the evolution ter. Some differences emerge: the esti- Overall, the clear message from our of real wages. Whether this sluggish- mates of average price duration work is that while the pure forward ness is attributable to stickiness in (between four and six quarters) are a looking version of the NKPC may be nominal or real wages is an open ques- bit higher than in the United States, rejected by the data, the hybrid variant tion. Several authors have shown that but the forward-looking component is, with a dominant role for forward look- extending models similar to ours to if anything, even more dominant than ing behavior does reasonably well. It is allow for staggered nominal wage set- in the United States. in this respect that the NKPC provides ting holds promise.19 Indeed a variation It is clearly a virtue of our useful insights into the nature of infla- of this model that allows for wages to approach that we are able to obtain tion dynamics and, along with it, useful be indexed to past inflation appears to estimates of the degree of price rigid- insights for the conduct of monetary do well empirically. Among other ity. One potential shortcoming, policy. things, this model of staggered wage though, is that we have to assume that and price setting captures very well the the average time interval over which response of inflation to a monetary firms keep price fixed remains con- The NKPC and policy shock, thus providing a direct stant over the sample. Put differently, Inflation Persistence response to the Mankiw critique. We our model assumes time dependent are currently exploring how well the price setting (where the time period How does the hybrid NKPC model captures the overall variation of remains constant) as opposed to state account for the apparent high degree postwar inflation. dependent pricing (where firms face of persistence of inflation in the data? 1 fixed costs of changing price but with Two factors are key: the relatively S. Fischer, “Long-Term Contracts, the time interval between adjustments Rational Expectations, and the Optimal 16 modest amount of lagged inflation endogenous.) Our implicit assump- (certainly as compared to the TPC) Money Supply,” Journal of Political tion is that variation in macroeconom- Economy, 85 (1) (1977), pp. 191-206. and the persistence of real marginal 2 ic conditions (particularly inflation) is cost. Regarding the latter: under the J. Taylor, “Aggregate Dynamics and not sufficiently large to systematically assumption of Cobb-Douglas technol- Staggered Contracts,” Journal of Political affect the interval over which firms Economy, 88 (1) (1980), pp. 1-24. ogy, real marginal cost corresponds to 3 change prices. In this case, the relative- real unit labor costs: the real wage G. A. Calvo, “Staggered Prices in a Utility ly simple time-dependent model may divided by average labor productivity. Maximizing Framework,” Journal of be thought of as a reduced form In the data, real unit labor costs are Monetary Economics, 12 (1983), pp. approximation of the more complex highly persistent and highly correlated 383-98.

NBER Reporter Summer 2003 17. 4 For examples of the traditional Phillips Mysterious Tradeoff Between Inflation and European Economic Review, Vol. curve, see R. J. Gordon, “Foundations of the Unemployment,” Economic Journal, 117 45(7) (2001), pp. 1237-70. Goldilocks Economy,” Brookings Papers (2001), pp. 1295-328. 15 See, for example, M. Bils and P. Klenow, on Economic Activity, Vol. 2 (1998), 10 J. Galí and M. Gertler, “Inflation “Some Evidence on the Importance of Sticky pp. 297-346, and J. H. Stock and M. W. Dynamics: A Structural Econometric Prices,” mimeo, 2002. Watson, “Forecasting Inflation,” Journal of Analysis.” 16 For models of state-dependent pricing, see Monetary Economics, Vol. 44 (2001) 11 See also A. Sbordone, “Prices and Unit A. Caplin and J. Leahy, “State-Dependent pp. 293-335. Labor Costs: Testing Models of Pricing Pricing and the Dynamics of Money and 5 R. C. Clarida, J. Galí, and M. Gertler, Behavior,” Journal of Monetary Output,” Quarterly Journal of “The Science of Monetary Policy: A New Economics, Vol. 45 (2) (2002), pp. 265- Economics, 106 (1991), pp. 683-708, Keynesian Perspective,” Journal of 92, which pursues a similar approach, though and M. Dotsey, R. G. King, and A. Economic Literature, Vol. 37 (1999), using a different econometric methodology. Wolman, “State-dependent Pricing and the pp. 1661-707; J. Galí, “New Perspectives on 12 As Christiano, Eichenbaum, and Evans General Equilibrium Dynamics of Money Monetary Policy, Inflation, and the Business observe, one can obtain an equivalent formu- and Output,” Quarterly Journal of Cycle,” NBER Working Paper No. 8767, lation by introducing lagged indexing, as Economics, 114 (2) (1999). February 2002, and in Advances in opposed to the rule of thumb price setting. See 17 J. Galí, M. Gertler, and D. López-Salido, Economics and Econometrics, Vol. L. Christiano, M. S. Eichenbaum, and C. “Robustness of The Estimates of the Hybrid III, M. Dewatripont, L. Hansen, and S. Evans, “Nominal Rigidities and the Effects Version of the New Keynesian Phillips Turnovsky, eds., Cambridge: Cambridge of a Shock to Monetary Policy,” mimeo, Curve.” University Press, 2003; M. Woodford, 2001. 18 L. Christiano, M. S. Eichenbaum, and C. “Optimal Monetary Policy Inertia,” NBER 13 Furhrer and Moore earlier proposed a Evans, “Nominal Rigidities and the Effects Working Paper No. 7261, July 1999. hybrid model. Ours differs by having real of a Shock to Monetary Policy;” P. N. 6 P. R. Krugman, “It’s Baaack! Japan’s marginal cost as the forcing variable and also Ireland, “Sticky Price Models of the Business Slump and the Return of the Liquidity having the slope coefficients derived as explicit Cycle: Specification and Stability,” Journal Trap,” Brookings Papers on Economic functions of the primitive parameters. Our of Monetary Economics, Vol. 47 (1) Activity, Vol. 2 (1998), pp. 137-87. formulation leads to a more important role for (2001), pp. 3-18; F. Smets and R. Wouters, 7 G. Eggertson and M. Woodford, “The forward-looking behavior. See J. Galí and M. “An Estimated Dynamic Stochastic General Zero Bound on Interest Rates and Optimal Gertler, “Inflation Dynamics: A Structural Equilibrium Model of the Euro Area,” Monetary Policy,” mimeo, 2003. Econometric Analysis;” J. Galí, M. Gertler, Journal of the European Economic 8 See, for example, J. C. Fuhrer and G. and D. López-Salido, “Robustness of The Association, forthcoming. Moore, “Inflation Persistence,” Quarterly Estimates of the Hybrid Version of the 19 For and alternative approach, based on Journal of Economics, Vol. 440 New Keynesian Phillips Curve,” mimeo, sticky information, see N. G. Mankiw and (February 1995), pp. 127-59, or J. Gali and 2003; and J. C. Fuhrer, “The (Un)Importance R. Reis, “Sticky Information versus Sticky M. Gertler, “Inflation Dynamics: A of Forward-Looking Behavior in Price Prices: A Proposal to Replace the New Structural Econometric Analysis,” Journal Setting,” Journal for Money, Credit and Keynesian Phillips Curve,” Quarterly of Monetary Economics, Vol. 44 (2) Banking, 29 (August 1997), pp. 338-50. Journal of Economics, 117 (November (1999), pp. 195-222. 14 J. Galí, M. Gertler, and D. López-Salido, 2002), pp. 1295-328. 9 N. G. Mankiw, “The Inexorable and “European Inflation Dynamics,”

18. NBER Reporter Summer 2003 The New Economics of Smoking Jonathan Gruber*

The past six years have seen an controls2. In this article, I describe this inability to carry out plans to quit. enormous change in the treatment of “new economics of smoking.” This formulation of preferences is smoking by both the public and policy one which is much more widely sup- makers. In 1995, federal and state ported by the large literature on exper- excise taxes on cigarettes were one- The New Approach imental evaluations of individual third lower, in real terms, than their choice over time. Experiments consis- peak level of the mid-1960s. But taxes The fundamental problem with the tently show that consumers are much have risen by over 50 percent since rational addiction model is that it does more patient when making decisions then, and now stand at over one dollar not account for the “self- control” about the future than when those same per pack. problems faced by smokers. There is decisions are made about today. From the traditional economic per- ample evidence that adults are unable Individuals are much more willing to spective, this shift in government poli- to quit smoking even if they have a declare that their diets will start tomor- cy is unwarranted. The traditional eco- desire to do so. Eight in ten smokers row than to start the diet today. The nomic model of smoking follows the in America express a desire to quit the problem is that when tomorrow standard approach to modeling any habit, but many fewer than that actual- comes, it is once again easier to push decision that involves tradeoffs over ly do quit. According to one study, off the date that the diet will begin. time1. Fully informed, forward-look- over 80 percent of smokers try to quit So, there is a conflict: you would ing, rational consumers make the deci- in a typical year, and the average smok- always like to start the diet tomorrow, sion of whether to smoke, weighing er tries to quit every eight and half but you never get to the point where the benefits of doing so in terms of months. Fifty-four percent of serious you are actually willing to make that smoking enjoyment against the costs quit attempts fail within one week. sacrifice. in terms of health and other risks. The These facts motivated Koszegi and The key implication of time incon- only call for intervention in such a me to develop an alternative formula- sistent preferences is that one’s future model is related to the externalities tion of the smoking decision which self would like to somehow constrain that smokers impose on others, such changes the traditional formulation in one’s current self to behave more as increased medical costs for our pub- just one critical way: by allowing smok- patiently (for example to somehow lic insurance programs. But such ers to be time inconsistent. This force you today to push away that extra externalities are in fact fairly small by approach, now widely used within the piece of cake). Thus, time inconsis- most measures, since these costs are new field of “behavioral economics,” tent consumers will have demand for offset by the savings from earlier mor- is one where there is conflict between commitment devices that can be used to tality of smokers, who pay a lifetime of what the smoker would like for himself induce more appropriate behavior in Social Security taxes but often don’t today and what he would like for him- the present. Indeed, the search for live long enough to collect their bene- self tomorrow. Today’s “self ” is impa- such commitment devices is the hall- fits. As a result, the traditional eco- tient. Faced with the tradeoff between mark of most recommended strategies nomic model would suggest that the the short- term pleasures of smoking for quitting smoking: people regularly “optimal” tax on cigarettes may be and the long- term health damages of set up socially managed incentives to below even its 1995 level. doing so, he will greatly discount the refrain from smoking by betting with My recent work has questioned the latter and decide to smoke. But tomor- others, telling others about the deci- validity of this traditional model. I row’s “self ” is much more patient. sion, and otherwise making it embar- have developed, in work with Botond That more patient self would prefer to rassing to smoke. Koszegi, an alternative model of the quit smoking. The problem, however, Unfortunately, such self-control smoking decision which has radically is that tomorrow never comes. The next devices are only imperfectly provided different implications for government day, the future self who was patient is by the private market. For every possi- policy, rationalizing large taxes on cig- now the current self who is impatient. ble device, there is another device that arettes and other types of regulatory So the smoking continues, to the long- can undo it. I can always cheat on my term regret of the smoker. This is in bets with others, or not go to my sup- contrast with the time consistent formu- port group meetings and smoke * Gruber is a Research Associate in the lation that is assumed by the tradition- instead. There is no way to truly com- NBER's Programs on Public Economics and al economics model. In that formula- mit oneself to not smoke or to not buy Health Care and is a professor of economics tion, today’s self and all future selves cigarettes through the private market. at MIT. His profile appears later in this are in agreement about the advisability But the government, on the other issue. of smoking, leading to no regret or hand, can provide an excellent com-

NBER Reporter Summer 2003 19. mitment device: cigarette taxation (or and the average cigarettes smoked minus one; that is, when cigarette legally-induced price rises). By raising over the smoker’s life, Koszegi and I prices rise, there is no net increase in the price of cigarettes, the government compute that the cost of smoking one cigarette spending for the lowest and courts can make smoking more pack of cigarettes, in terms of the income group. For higher income costly for today’s self, helping achieve value of life lost, is $35 per pack. This groups, the price sensitivity is only what the smoker’s own long- term self is an enormous figure which is on the about one-third as large. would desire by lowering smoking order of 100 times the typical estimate Koszegi and I show that, given today. There is a large literature which of the external damage done by smok- these differences, cigarette taxes are in documents that smoking falls as ciga- ing. Given this enormous damage that general not very regressive, since the rette prices rise; the best estimates sug- smokers do to themselves by smoking, larger self-control benefits for lower gest that each 10 percent rise in the any model which suggests that some income groups compensate for the price of cigarettes lowers the con- share of these “internalities” should higher taxes they pay as a share of sumption of cigarettes by 5 to 6 per- be reflected in government policy will income. Indeed, if self-control prob- cent. For youth smokers, price sensi- suggest very large optimal taxes on lems are large, then cigarette taxes can tivity is even higher. So higher taxes, cigarettes. be highly progressive under this alterna- and therefore higher prices, will signif- Koszegi and I show this in our tive approach. The point is that, with icantly reduce smoking today. work by considering the alternative a price elasticity of minus one, the formulation described above. We first poor as a group spend no more of consider a very modest degree of time their incomes on cigarettes after tax Implications of the inconsistency, much below that increases than they did before; the New Approach for assessed by most laboratory experi- higher spending among those who still ments. Even in that case, we find that smoke is offset by the savings among Government Policy the optimal tax on cigarettes, above those who quit. But, as a group, the and beyond any externality effects, is poor are much healthier as a result of While this new approach to model- $1 to $2. For more severe time incon- the fact that they have reduced their ing smoking changes the traditional sistency, which is consistent with labo- smoking. So, on net, they are better off model in only one way, it has dramatic ratory evidence on preferences, the tax from the higher prices if they wanted implications for government policy. In is much higher, on the order of $5 to to quit smoking, but could not because this model, the damage that smokers $10 per pack. Another common argu- of self-control problems. do to themselves is relevant, above and ment against cigarette taxation is on beyond external effects on others. distributional grounds. Smoking in the This is because, from their own long- United States is very socioeconomical- New Evidence run perspective, smokers are smoking ly concentrated. Expenditures on too much. Their long-term selves rec- tobacco products as a share of family It is important to highlight that this ognize this failure and would like to income fall from 3.2 percent in the alternative model is not a radical reduce smoking. But, their current bottom income quartile to only 0.4 departure from the traditional eco- selves are unable to do so. So the gov- percent in the top income quartile. nomic approach. This formulation ernment can do what the private sector This pattern raises a concern that continues to assume perfectly rational, cannot; they can make it more costly to increased cigarette taxes will be exces- forward-looking, fully informed con- smoke in a way that cannot be evaded, sively burdensome on those with the sumers. That is, in every respect but combating one’s short term impa- lowest incomes. one (time consistency), we retain the tience on behalf of one’s long term But this alternative approach to features of decisionmaking that econ- interests. modeling smoking also challenges the omists have used to model behaviors While the damage that smokers do standard perception that cigarette for years. As a result, this alternative to others is, on net, small, the damage taxes are highly regressive. This is model also generates many aspects of that smokers do to themselves is enor- because the self-control benefit of real world behavior that are predicted mous. There are many negative cigarette taxation is larger, the larger is by the traditional model. For example, impacts of smoking on individual the price sensitivity of smoking. That under both models, smokers react to health, but Koszegi and I focus on is, for groups who are particularly higher prices by smoking less. But the only one: the costs in terms of short- price sensitive, higher prices are a par- models do have one key differential ened lives. As noted above, on aver- ticularly effective self-control device, prediction. Under the traditional for- age smokers live about six fewer years since they will have more of the mulation, higher taxes on cigarettes than nonsmokers. Economists have desired effect of reducing their smok- make smokers worse off; the govern- spent years showing how we can use ing. And lower income groups are ment is constraining their choice of an individuals’ revealed preferences much more price sensitive than higher activity that they are pursuing rational- toward risk to value this type of lost income groups. Indeed, our estimates ly. But, under this alternative formula- life. Putting these estimates together suggest that the price elasticity of cig- tion, higher taxes on cigarettes make with information on the reduction in arette demand in the bottom quartile smokers better off: the government is years of life attributable to smoking of the income distribution is roughly helping them achieve the self-control

20. NBER Reporter Summer 2003 that they cannot achieve through the suicide? This was the question that My work on this new economics of private market. motivated me to organize the NBER smoking over the past few years has In a recent study, my coauthor Project on “Risky Behavior Among shown that there is an alternative for- Sendhil Mullainathan and I directly Youths: An Economic Analysis.”4 This mulation of the smoking decision tested this prediction3. We did so by project brought together ten teams of which is more consistent with the assessing whether the self-reported leading researchers to analyze the risky existing evidence on behavior, and well-being of smokers falls or rises decisions made by youths. The result- which has radically different implica- when cigarette taxes increase. Using ing set of papers on these diverse tions for government regulation of data from the General Social Survey in behaviors yielded several clear and this activity. But this is only one alter- the U.S., we find consistently strong consistent conclusions, as summarized native to the traditional model; for evidence that higher cigarette taxes are in my introductory chapter. First, eco- example, B. Douglas Bernheim and associated with higher levels of nomic incentives matter for youths in Antonio Rangel recently have devel- reported well-being among those with making these risky decisions; youth oped another alternative which shares a propensity to smoke. Moreover, the behavior in these areas is very sensitive many predictions with my model with effects are almost identical when we to prices, penalties, and other policy Koszegi6. Thus, there is clearly more replicate the exercise with Canadian levers that affect the costs of engaging work needed in distinguishing these data, and, in both countries, these in the activity. Second, these econom- alternatives to the traditional model. effects are present for cigarette taxes ic incentives nonetheless can explain And, more work is needed on develop- but not for any other excise taxes. only a small part of the time-series ing and testing alternative models of While this is not an ideal experimental movements in youth risk taking. For other potentially addictive behaviors; evaluation of these alternative models, example, the substantial decline in real David M. Cutler, Edward Glaeser, and it is much more consistent with the cigarette prices during the 1990s can Jesse Shapiro provide a nice example alternative formulation of the smok- explain at most one-quarter of the dra- of applying these models to the ing decision than it is with the tradi- matic rise in youth smoking over this important issue of obesity in the tional model. period. Third, macroeconomic condi- United States7. tions matter for youth decisionmaking; Moreover, there remains much dropping out of school goes up, but more to do in terms of understanding Youth Smoking having unprotected sex goes down, as the decisions by youths to start smok- the economy improves. ing, or to engage in other risky activi- An additional problem with both Finally, there is clear evidence that ties. While research in this area has of these models of smoking is that the the decisions made by youths have shown that economic incentives mat- decision to initially begin smoking is long lasting implications for their lives. ter, we have been unsuccessful thus far made primarily by youths, whose abili- For example, my analysis of youth in developing models that explain the ty to make fully-informed, appropri- smoking with Jonathan Zinman wide variation over time the propensi- ately forward looking decisions is showed that adults who lived in low ty of youths to engage in these behav- questioned by society in many contexts cigarette tax states as youths were iors. (for example,. minimum ages for much likelier to smoke as adults than drinking, driving, and voting). More were those who lived in higher ciga- 1 G. S. Becker and K. M. Murphy, “A Theory than three-quarters of smokers begin rette tax states5. Our estimates imply of Rational Addiction,” Journal of Political smoking before age 19. Moreover, that the rise in youth smoking of the Economy, 96 (1998), pp. 675-700. youths apparently initiate smoking 1990s will lead to 3.2 million fewer 2 J. Gruber and B. Koszegi, “Is Addiction without a full appreciation for the years of life for this youth cohort. ‘Rational’? Theory and Evidence,” NBER long-run implications of their actions. Other papers show that lower drinking Working Paper No. 7507, January 2000, Among high school seniors who ages lead to more adult binge drinking and Quarterly Journal of Economics, smoke more than one pack/day, the as youths age, and that those who drop (116) (4) (November 2001), pp. 1261- smoking rate five years later among out of school because of a good econ- 1303; and “A Theory of Government those who stated that they would not be omy never make up those years of lost Regulation of Addictive Bads: Optimal Tax smoking (74 percent) is actually higher schooling. Thus, the potential “mis- Levels and Tax Incidence for Cigarette than the smoking rate among those takes” made by youths, such as the Taxation,” NBER Working Paper No. who stated that they would be smoking understatement of the long-run addic- 8777, February 2002. (72 percent). tiveness of smoking noted above, can 3 J. Gruber and S. Mullainathan, “Do So how do youths make decisions have important future implications. Cigarette Taxes Make Smokers Happier?” to smoke, and to engage more general- NBER Working Paper No. 8872, April ly in the set of risky behaviors that 2002. they undertake, such as consumption What We Have Learned 4 J. Gruber, “Risky Behavior Among Youth: of alcohol and illicit drugs, criminal and Where We Need to An Economic Analysis, Introduction” in J. behavior, unprotected sex, over- or Go Next Gruber, ed., Risky Behavior Among under- eating, driving dangerously, Youth: An Economic Analysis, Chicago: dropping out of school, or attempting University of Chicago Press, 2001, pp. 1-28.

NBER Reporter Summer 2003 21. 5 J. Gruber and J. Zinman, “Youth Smoking Economic Analysis, Chicago: University November 2002. in the U.S.: Evidence and Implications,” of Chicago Press, 2001, pp. 69-120 7 D. M. Cutler, E. Glaeser, and J. Shapiro, NBER Working Paper No. 7780, July 6 B. D. Bernheim and A. Rangel, “Addiction “Why Have Americans Become More 2000, and in J. Gruber, ed., Risky and Cue-Conditioned Cognitive Processes,” Obese?” NBER Working Paper No. 9446, Behavior Among Youth: An NBER Working Paper No. 9329, January 2003.

Venture Capitalists As Economic Principals Steven N. Kaplan and Per Strömberg*

There is a large academic literature contracts and actions of real world parts in financial contracting theory. on the principal-agent problem in principals to their counterparts in We do so by conducting a detailed financial contracting.1 This literature financial contracting theory has lagged study of 213 actual contracts between focuses on the conflicts of interest behind. In this paper, we describe VCs and entrepreneurs. We find first between an agent — an entrepreneur recent empirical work and its relation that VC financings allow VCs to sepa- with a venture that needs financing — to theory for one prominent class of rately allocate cash flow rights, voting and a principal — an investor with the such principals — venture capitalists rights, board rights, liquidation rights, funds to finance the venture. (VCs). In our view, VCs are real world and other control rights. The separa- According to these theories, there are a entities that closely approximate the tion of these rights is apparent, for number of ways that the investors of theory. VCs invest in example, in that VCs control roughly investor/principal can mitigate these entrepreneurs who need financing to half of the cash flow rights on aver- conflicts. First, the investor can collect fund a promising project or company. age, but have a majority of board seats information before deciding whether VCs have strong incentives to maxi- in only 25 percent of the investments. to invest, in order to screen out ex ante mize value, but at the same time Second, while convertible securities unprofitable projects and bad entre- receive few or no private benefits of are used most frequently, VCs also preneurs. Second, investors can struc- control. Although they are intermedi- implement the same set of rights using ture financial contracts — that is, the aries, VCs typically receive at least 20 combinations of multiple classes of allocation of cash flow, control and percent of the profits on their portfo- common stock and straight preferred liquidation rights — between them- lios. stock. We also point out that VCs use selves and entrepreneurs to provide In this article, we describe recent a variant of convertible preferred incentives for the entrepreneurs to empirical work — both ours and oth- called “participating preferred” in behave appropriately. And third, the ers’ — on the three things that VCs do: roughly 40 percent of the financings. investors can engage both in collecting contracting, screening, and monitoring. Participating preferred, under most information and in monitoring it once Unlike previous empirical work that circumstances, behaves like a position the project is under way. has relied largely on surveys, our work of straight preferred stock and com- Despite the large volume of theory, (and much of the work we describe) mon stock rather than like a position the empirical work in comparing the relies on detailed information collected of convertible preferred. from actual VC financings. Third, cash flow rights, voting * Kaplan is a Research Associate and rights, control rights, and future Stromberg is a Faculty Research Fellow in the Contracting financings are frequently contingent NBER’s Program on Corporate Finance. on observable measures of financial Both are on the faculty of the University of In a forthcoming article2, we com- and non-financial performance. These Chicago Graduate School of Business. Their pare the characteristics of real world state contingencies are more common profiles appear later in this issue. financial contracts to their counter- in the early stages of the VC-entrepre-

22. NBER Reporter Summer 2003 neur relationships (first VC rounds) Screening mixed predictions regarding external and in earlier stage investments. uncertainty. Traditional moral hazard Fourth, voting rights, board rights, Before making an investment and theories based on risk-sharing predict and liquidation rights are allocated designing the financial contracts, VCs performance-based pay decreases with such that if the company performs spend a significant amount of time external uncertainty. Alternatively, poorly, the VCs obtain full control. As and effort evaluating and screening the more recent theories10 suggest that company performance improves, the investment opportunity. In recent performance-based pay and direct entrepreneur retains/obtains more work8, we focus empirically on this monitoring are substitutes. As external control rights. If the company per- information collection and evaluation. uncertainty increases, direct monitor- forms very well and the VCs earn a To help the VC partnership evalu- ing becomes less effective and princi- sizable multiple of its investment, the ate an investment in a company, the pals make greater use of pay-perform- VCs retain their cash flow rights, but individual VC who is sponsoring the ance incentives. relinquish most of their control and investment often prepares a detailed Two types of theories make predic- liquidation rights through the automat- investment analysis or memorandum tions about execution or complexity ic conversion provision that is present for the other partners. In our 2002 risk. Theories of multitasking11 predict in virtually all our financings. paper, we analyze the investment that pay based on specific milestones Fifth, VCs typically include non- memoranda from eleven VC partner- should decrease as execution risk compete and vesting provisions that ships for investments in 67 portfolio increases because compensation based make it more expensive for the entre- companies. We complement that on a signal correlated with a particular preneur to leave the firm, thus mitigat- analysis with information from the action will lead the manager to put too ing the potential hold-up problem company business plans, as well as data much emphasis on that action. Hold- between the entrepreneur and the on financial contracts from our previ- up theories12 suggest that in highly investor. Vesting provisions are more ous study9. complex environments where the common in early stage financings The VC analyses that we describe entrepreneur’s human capital is partic- where it is more likely that the hold-up include a set of investment theses or ularly important, we should observe problem is more severe. rationales for making the investment contracts — such as vesting provisions Our results have a number of and a discussion of the concomitant — that make it costlier for the entre- implications. For example, cash flow risks. VCs explicitly consider the preneur to leave. rights matter in a way that is consistent attractiveness of the external environ- Consistent with the agency expla- with standard principal-agent theories ment — the market size, customer nation, internal uncertainty is signifi- such as Holmstrom (1979)3.VCs adoption, and competition — the fea- cantly related to many of the incentive change the entrepreneur’s equity com- sibility of the strategy and technology, and control mechanisms in the finan- pensation function, making it more the quality of the management team, cial contracts. Higher internal risk is sensitive to performance when incen- and the deal terms. VCs also explicitly associated with more VC control, tive and asymmetric information prob- delineate the risks involved in the more contingent compensation to the lems are more severe. investments. The risks typically relate entrepreneur, and more contingent Further, the allocation of control to the same characteristics that the financing in a given round. rights between the VC and the entre- VCs evaluate for attractiveness. Higher external risk, like internal preneur is a central feature of the We use these assessments to form risk, is associated with more VC con- financial contracts. This strongly sug- three different “risk measures”: inter- trol, and more contingent compensa- gests that, despite the prevalence of nal uncertainty — the relevant infor- tion. Higher external risk is also asso- contingent contracting, contracts are mation is internal to the firm and it is ciated with increases in the strength of inherently incomplete. This finding more likely that the VC is less VC liquidation rights, and tighter stag- gives support to the incomplete con- informed than the entrepreneur; exter- ing, in the sense of a shorter period tracting approach pioneered by nal uncertainty — the relevant infor- between financing rounds. These find- Grossman and Hart4 and Hart and mation is external to the firm and it is ings are highly inconsistent with opti- Moore5. more likely that the VC and the entre- mal risk-sharing between risk-averse Cash flow rights and control rights preneur are equally informed; and dif- entrepreneurs and risk-neutral also can be separated and made con- ficulty of execution, different from investors, a common assumption in tingent on observable and verifiable both previous notions of risk, which standard agency models. measures of performance. This is captures the complexity of the task Risk related to difficulty of execu- most supportive of theories that pre- and the reliance on the entrepreneur’s tion shows a (weakly) negative relation dict shifts of control to investors in human capital. We compare these risk with many contractual mechanisms, different states, such as Aghion and measures to the financial contracts. such as contingent compensation and Bolton (1992)6. Finally, the widespread If agency conflicts arising from VC liquidation rights. These results use of non-compete and vesting provi- moral hazard and asymmetric informa- suggest that for highly complex envi- sions indicates that VCs care about the tion are important, then the financial ronments, where the manager’s human hold-up problem.7 contracts should be related to internal capital is particularly important, stan- uncertainty. The agency theories make dard incentive mechanisms are less

NBER Reporter Summer 2003 23. effective. Furthermore, execution risk ities in this area. 2 S. N. Kaplan and P. Strömberg, is significantly positively related to We also show that the actions the “Financial Contracting Theory Meets the founder vesting provisions. This result VCs expect to take are related to the Real World: An Empirical Analysis of is consistent with the multitasking and contracts. Consistent with control the- Venture Capital Contracts,” forthcoming in hold-up theories. ories, VCs are more likely to strength- the Review of Economic Studies. en management as VC control increas- 3 B. Holmström, “Moral Hazard and Monitoring es. These theories show that investor Observability,” Bell Journal of control is necessary to implement Economics, 10 (1979), pp. 74-91. Finally, several recent papers focus actions that reduce the entrepreneur’s 4 S. J. Grossman and O. D. Hart. “The on post-investment information col- private benefits, such as management Costs and Benefits of Ownership: A Theory lection, monitoring, and other actions interventions. of Vertical and Lateral Integration,” by the VC. Anecdotal accounts stress Consistent with theories like Journal of Political Economy, 94 an important role for VCs in monitor- Inderst and Muller’s17 that stress having (1986), pp. 691-719. ing management, finding management, incentives for the VC to provide value- 5 O. D. Hart and J. Moore, “Property Rights and providing advice. For example, added services, we also find that VC’s and the Nature of the Firm,” Journal of Lerner13 finds that VCs are more likely value-added services increase as the Political Economy, 98 (1990), pp. 1119 to join or be added to the boards of VC’s equity stake increases, but are not – 58; and “Default and Renegotiation: A private companies in periods when the related to VC control. Dynamic Model of Debt,” NBER Working chief executive officer (CEO) of the Paper No. 5907, January 1997, and in company changes. He interprets this as Implications and Quarterly Journal of Economics, 113 evidence of VC monitoring. (1) (1998), pp. 1-41. Hellman and Puri14 study a hand- Conclusion 6 P. Aghion and P. Bolton. “An Incomplete collected sample of 173 start-up firms Contracts Approach to Financial from California’s Silicon Valley. They The empirical studies of venture Contracting,” Review of Economic find that venture capital is associated capitalists indicate that they attempt to Studies, 77 (1992), pp. 338-401. with a significant reduction in the time mitigate principal-agent conflicts in 7 O. D. Hart and J. Moore. “A Theory of to bring a product to market. They the three ways suggested by theory: Debt Based on the Inalienability of Human provide some evidence that this associ- through sophisticated contracting, pre- Capital,” Quarterly Journal of ation holds after controlling for VC investment screening, and post-invest- Economics, 109 (4) (November 1994), ability to select a more successful com- ment monitoring and advising. The pp. 841-79. pany. evidence also suggests that contract- 8 S. N. Kaplan and P. Strömberg, Hellman and Puri15 also study ing, screening, and monitoring are “Characteristics, Contracts, and Actions: another aspect of the same dataset. closely interrelated. In the screening Evidence From Venture Capitalist They find that VC-financed firms are process, the VCs identify areas where Analyses,” NBER Working Paper No. more likely, and quicker, to profession- they can add value through monitoring 8764, February 2002. alize by adopting stock option plans and support. In the contracting stage, 9 See S. N. Kaplan and P. Strömberg, and hiring a vice president of sales, the VCs allocate rights in order to “Financial Contracting Theory Meets the and by bringing in CEOs from outside facilitate monitoring and minimize the Real World: An Empirical Analysis of the firm. impact of the identified risk factors, Venture Capital Contracts.” The three studies described in the for example by allocating more control 10 See, for example, C. J. Prendergast, “The previous paragraphs find indirect evi- to investors when management is Tenuous Tradeoff Between Risk and dence of post-financing VC actions. In weak, or making founder cash flow Incentives,” Journal of Political our 2002 paper16, we complement rights and release of funds contingent Economy, 110 (5) (2002) pp. 1071-1102. these studies by presenting direct evi- on management actions. Also, the allo- 11 See, for example, B. Holmström and P. dence on VC actions or monitoring. cations of equity to VCs provide Milgrom, “Multitask Principal Agent We use the investment analyses to incentives to engage in costly support Analyses: Incentive Contracts, Asset measure the actions that the VCs took activities that increase the upside value Ownership, and Job Design”, Journal of before investing and that the VCs of the venture, rather than just mini- Law, Economics, and Organization, 7 expect to undertake conditional on mizing potential losses. There is room (1991), pp. 24-92. investing. In at least half of the invest- for future empirical research to study 12 See O. D. Hart and J. Moore. “A Theory ments, the VC expects to play an these activities in greater detail, both of Debt Based on the Inalienability of important role in recruiting manage- for VCs and for other intermediaries, Human Capital.” ment. In more than one-third of the such as banks. 13 J. Lerner, “Venture Capitalists and the investments, the VC expects to provide Oversight of Private Firms,” Journal of value-added services, such as strategic 1 For a recent summary, see O. D. Hart, Finance, 50 (March 1995), pp. 301-18. advice or customer introductions. “Financial Contracting,” NBER Working 14 T. Hellman and M. Puri, “The Interaction Because the investment memoranda Paper No. 8285, May 2001, and in Between Product Market and Financial vary in the detail they provide, these Journal of Economic Literature, 39 Strategy: The Role of Venture Capital,” results likely understate the VCs’ activ- (2001), pp. 1079-1100. Review of Financial Studies, (Winter

24. NBER Reporter Summer 2003 2000), pp. 959-84. 16 See S. N. Kaplan and P. Strömberg, and Efficiency in the Market for Venture 15 T. Hellman and M. Puri. “Venture “Characteristics, Contracts, and Actions: Capital”, working paper, New York Capital and the Professionalization of Start- Evidence From Venture Capitalist University, 2001. up Firms: Empirical Evidence,” Journal of 17Analyses.” Finance, 57 (1) (2002), pp. 169-197. R. Inderst and H. Müller, “Competition

*

NBER Profile: Joshua Angrist

Joshua Angrist is a Research Associate regulation and institutions, and econo- in the NBER’s Programs on Children, metric methods for program and policy Education, and Labor Studies, and a evaluation. Although many of his papers Professor of Economics at MIT. A dual use data from other countries, he does not U.S. and Israeli citizen, he taught at the especially like to travel and prefers to get Hebrew University of Jerusalem before the data in the mail. He is also a Fellow of coming to MIT. He holds a B.A. from the Econometric Society and a Co-editor Oberlin College and also spent time as an of the Journal of Labor Economics. undergraduate studying at the London Angrist has a long-standing interest- School of Economics and as a Masters ing in public policy. In addition to his student at Hebrew University. He com- academic work, he has worked as a con- pleted his Ph.D. in Economics at sultant to the U.S. Social Security Princeton in 1989 and his first academic Administration, The Manpower Demon- job was as an Assistant Professor at stration Research Corporation, and for Harvard from 1989-91. the Israeli government after the Oslo Angrist’s research interests include the peace negotiations in 1994. He lives in effects of school inputs and organization Brookline with his wife Mira, and their on student achievement, the impact of two children, Adie and Noam. The education and social programs on the Angrist family enjoy activities like hiking, labor market, immigration, labor market skiing, skating, sailing, and eating.

NBER Reporter Summer 2003 25. NBER Profile: Jordi Galí

Jordi Galí is a Research Associate in the macro-econometrics. He is currently a part- NBER’s Program on Economic Fluctu- time Professor of Economics at Universitat ations and Growth and the Program on Pompeu Fabra, a visiting Scholar at the Monetary Economics. He is also a Senior Bank of Spain, and co-director of the Researcher and Director of the Center for International Macroeconomics Program of Research on International Economics the Centre for Economic Policy Research (CREI) at Universitat Pompeu Fabra (CEPR) in London. (Barcelona). Gali currently lives in Barcelona Galí earned his Ph.D. in Economics at (Catalonia). He is married to Ellen Caldwell, MIT in 1989 and has been a Professor of who works in the pharmaceutical industry. Economics at New York University and, They have two children, Arnau, 10, and prior to that, an Associate Professor at the Anna Carla, 9. In his leisure time, Galí Graduate School of Business at Columbia enjoys listening to jazz and classical music, University. His research interests include reading, skiing, and mountain biking. macroeconomics, monetary theory, and

NBER Profile: Jonathan Gruber

Jonathan Gruber is Director of the During the 1997-8 academic year, Gruber NBER’s Program on Children, an NBER was on leave as Deputy Assistant Secretary Research Associate, and a Professor of for Economic Policy at the Treasury Economics at MIT, where he has taught Department. since 1992. He is also a co-editor of the Gruber’s research focuses on public Journal of Public Economics, and an Associate finance and health economics. His recent Editor of the Journal of Health Economics. areas of particular interest include the eco- Gruber received his B.S. in Economics nomics of employer-provided health insur- from MIT and his Ph.D. in Economics from ance, the efficiency of our current system of Harvard University. He was the recipient of delivering health care to the indigent, the an Alfred P. Sloan Foundation Research effect of the Social Security program on Fellowship, a FIRST award from the retirement behavior, and the economics of National Institute on Aging, and the smoking. Kenneth Arrow Award for the Best Paper in Gruber lives in Lexington with his wife, Health Economics in 1994. He was also one Andrea, and three children, Sam (8), Jack (6), of 15 scientists nationwide to receive the and Ava (4). He spends his free time “keep- Presidential Faculty Fellow Award from the ing them busy!” National Science Foundation in 1995.

26. NBER Reporter Summer 2003 NBER Profile: Steven N. Kaplan

Steven N. Kaplan is an NBER Economics from Harvard College. Research Associate in the Corporate Between college and graduate school, he Finance Program and the Neubauer was an investment banker at Kidder Family Professor of Entrepreneurship and Peabody. Finance at the University of Chicago’s Kaplan’s research and teaching focuses Graduate School of Business (GSB). He is on issues in private equity and entrepre- also the faculty director of the GSB’s neurial finance, corporate governance, Polsky Entrepreneurship Center. Kaplan mergers and acquisitions, e-commerce, and has been a member of the GSB faculty corporate finance. He has published since 1988. papers in a number of academic and busi- He received his Ph.D. in Business ness journals. Economics from Harvard University and He lives in Chicago with his wife, Carol his A.B. in Applied Mathematics and Rubin, and their two sons, Sam and Alec.

NBER Profile: Per Strömberg

Per Strömberg is a Faculty Research areas of financial distress and venture Fellow in the NBER’s Program on capital finance. His work has been pub- Corporate Finance and an Associate lished in Journal of Finance, American Professor of Finance at the Graduate Economic Review, and Review of Economic School of Business (GSB) of the Studies, and has been presented at numer- University of Chicago. He has been a ous conferences and universities. His member of the GSB faculty since 1997, work on bankruptcy auctions won the and previously taught at the Stockholm 2001 Brattle Prize for best corporate School of Economics and at Carnegie finance paper published in the Journal of Mellon University. Finance. Strömberg received his Ph.D. in Strömberg is married to Lina, who is Financial Economics from Carnegie expecting their third child in June. In his Mellon University. He also holds an free time, apart from spending it with his MBA from the Stockholm School of family, he enjoys playing jazz piano and Economics. cooking. Strömberg has done research in the

NBER Reporter Summer 2003 27. Conferences

Eighteenth Annual Conference on Macroeconomics The NBER’s Eighteenth Annual NBER and New York University, “Disagreement About Inflation Conference on Macroeconomics, “On the Welfare Consequences of Expectations” organized by Mark Gertler, NBER the Increase in Inequality in the Discussants: Robert King, NBER and New York University, and United States” and Boston University, and John Kenneth Rogoff, NBER and Discussants: Steven J. Davis, NBER Williams, Federal Reserve Bank of Harvard University, was held in and University of Chicago, and San Francisco Cambridge on April 4 and 5. The Kjetil Storesletten, Institute for program was: International Economics Pierpaolo Benigno, New York University, and Michael Woodford, Susanto Basu, NBER and Annette Vissing-Jorgensen, NBER and Princeton University, University of Michigan; John G. NBER and Northwestern “Optimal Targeting Rules for Fernald, Federal Reserve Bank of University, “Behavioral Finance: An Monetary and Fiscal Policy” Chicago; and Nick Oulton and Impartial Assessment, or: Does Discussants: Stefania Albanesi, New Sylaja Srinivasan, Bank of Irrationality Disappear with Wealth? York University, and Marios England, “The Case of Missing Evidence from Expectations and Angeletos, NBER and MIT Productivity Growth, or: Why Has Actions” Productivity Accelerated in the Discussants: John Y. Campbell, Arminio Fraga, Ilan Goldfajn, United States but not the United NBER and Harvard University, and and Andre Minella, Central Bank Kingdom?” Owen Lamont, NBER and of Brazil, “Inflation Targeting in Discussants: Olivier J. Blanchard, University of Chicago Emerging Market Economics” NBER and MIT, and Gianluca Discussants: Frederic S. Mishkin, Violante, New York University N. Gregory Mankiw, NBER and NBER and Columbia University, Harvard University; Ricardo Reis, and Robert E. Hall, NBER and Dirk Krueger, NBER and Stanford Harvard University; and Justin Stanford University University, and Fabrizio Perri, Wolfers, Stanford University,

The United Kingdom experienced some evidence that the TFP accelera- information on wages, hours worked, an information and communications tion was larger in industries that earnings, and consumption for a large technology (ICT) investment boom in increased their ICT share in the pre- cross section of U.S. households. In the 1990s in parallel with the United ceding 15 years (especially in the early their sample, the cross-sectional varia- States, but measured total factor pro- 1980s). Furthermore, the TFP acceler- tion in wages and disposable earnings ductivity (TFP) there has decelerated ation appears negatively correlated has increased significantly, both within rather than accelerated in recent years. with increases in ICT usage in the late and between groups identified by edu- Basu, Fernald, Oulton, and 1990s. In the United Kingdom, by cation and sex. This trend potentially Srinivasan ask whether ICT can contrast, the increase in ICT-intensity implies large welfare losses, especially explain the divergent TFP perform- came later than in the United States. for the poorest groups in the U.S. pop- ance in the two countries. As a “gener- Given the long lags, these results sug- ulation. On the other hand, the disper- al purpose technology,” measured TFP gest, albeit tentatively, that the United sion in consumption and in hours should rise in ICT-using sectors Kingdom should see an acceleration in worked has not increased significantly, (reflecting either unobserved accumu- TFP over the next decade. suggesting that smoothing mecha- lation of intangible organizational cap- Krueger and Perri investigate the nisms (like credit markets) might have ital, spillovers, or both) with a long lag. welfare consequences of the stark reduced these losses. To better quanti- The authors indeed find that the accel- increase in wage and earnings inequal- fy the welfare consequences of the eration in U.S. TFP after the mid- ity in the United States for the last 30 recent changes in inequality, the 1990s was broadbased — located pri- years. Their data come from the authors estimate stochastic processes marily in ICT-using sectors rather than Consumer Expenditure Survey (CE), for income, consumption, and leisure ICT-producing sectors — and there is the only U.S. dataset that contains that are consistent with observed

28. NBER Reporter Summer 2003 cross-sectional variability and with substantial variation through time, rules from the point of view of the one-year mobility patterns from the moving with inflation, the absolute maximization of expected discounted CE. They insert these estimates into a value of the change in inflation, the utility in a dynamic stochastic general- standard lifetime-utility framework to output gap, and relative price variabili- equilibrium model, building on their obtain estimates of the welfare losses. ty. The authors argue that a satisfacto- work for the case of optimal monetary They find that, for commonly used ry model of economic dynamics must policy when lump-sum taxes are avail- specifications of the lifetime-utility speak to these important business cycle able. Finally, they do not content them- function, the welfare losses for a sub- moments. Nothing that most macro- selves with merely characterizing the stantial fraction of the U.S. population economic models do not endogenously optimal dynamic responses of their amount to between 2 and 3 percent of generate disagreement, they show that policy instruments (and other state lifetime consumption. a simple “sticky-information” model variables) to shocks under an optimal Vissing-Jorgensen discusses the broadly matches these facts. Moreover, policy, given one assumption or anoth- current state of the behavioral finance the sticky information model is consis- er about the nature and statistical literature. She argues that more direct tent with other observed departures of properties of the exogenous distur- evidence on investors’ actions and inflation expectations from full ration- bances to our model economy. Instead, expectations would make existing the- ality, including autocorrelated forecast they derive policy rules for the mone- ories more convincing to outsiders and errors and insufficient sensitivity to tary and fiscal authorities. In particular, would help sort among behavioral the- recent macroeconomic news. they seek to characterize optimal policy ories for a given asset pricing phenom- While by now two substantial liter- in terms of optimal targeting rules for enon. Furthermore, evidence on the atures seek to characterize optimal monetary and fiscal policy. The rules dependence of a given bias on investor monetary and fiscal policy respectively, are specified in terms of a target crite- wealth/sophistication would be useful, the two have developed largely in iso- rion for each authority; each authority first for determining whether the bias lation, and apparently on contradictory commits itself to use its policy instru- could be attributable to (fixed) infor- foundations. The modern literature on ment in each period in whatever way is mation or transactions costs or is like- dynamically optimal fiscal policy often necessary in order to allow it to project ly to require a behavioral explanation, abstracts from monetary aspects of an evolution of the economy consis- and second for determining which the economy, and thus implicitly allows tent with its target criterion. biases are likely to be most important for no useful role for monetary policy. Fraga, Goldfajn, and Minella for asset prices. The author analyzes a The literature on optimal monetary assess inflation targeting in emerging novel dataset on investor expectations policy instead has been concerned market economies (EMEs) and devel- and actions obtained from UBS/Gallup. mainly with quite distinct objectives op applied prescriptions for the con- The data suggest that an investor’s for monetary stabilization policy, duct of monetary policy and inflation- expectations about future market namely the minimization of the distor- targeting design in EMEs. They verify returns depends strongly on the tions that result from prices or wages that EMEs have faced more acute investor’s own investment experience, that do not adjust quickly enough to trade-offs — higher output and infla- and expectational variables do affect clear markets. At the same time, this tion volatility — and worse perform- portfolio choice. The dependence of literature typically ignores the fiscal ance than developed economies. These beliefs on own experience remains consequences of alternative monetary results stem from more pronounced strong for high-wealth investors, sug- policies; the characterizations of opti- external shocks, lower credibility, and gesting that information costs are not a mal monetary policy that are obtained lower levels of development of institu- likely explanation. She then reviews thus are strictly correct only for a tions in these countries. In order to evidence on the dependence of a series world in which lump-sum taxes are improve their performance, the of “irrational” investor behaviors on available. Here Benigno and Woodford authors recommend high levels of investor wealth and concludes that model price stickiness by assuming transparency and communication with many such behaviors diminish substan- staggered pricing of the kind intro- the public and the development of tially with wealth. As an example of duced by Calvo (1983). Perhaps more more stable institutions. At an opera- how one may approach a calculation of importantly, they obtain analytical tional level, the authors describe a pro- the costs needed to explain a particular results rather than purely numerical cedure that a central bank under infla- type of “irrational” behavior, she con- ones, by proposing a linear-quadratic tion targeting can apply and communi- siders the size of the costs needed to approach to the characterization of cate when facing strong supply shocks, explain why many households do not optimal monetary and fiscal policy that and present a monitoring structure for invest in the stock market. allows them to nest both conventional an inflation-targeting regime under an Analyzing 50 years of inflation analyses of optimal monetary policy IMF program. expectations data from several sources, and analyses of optimal tax-smoothing These papers will be published by Mankiw, Reis, and Wolfers docu- as special cases of their more general the MIT Press as NBER Macroeconomics ment substantial disagreement among framework. They show how a linear- Annual, Volume 18. They will also be consumers and professional econo- quadratic policy problem can be available at “Books in Progress” on the mists about expected future inflation. derived which yields a correct linear NBER website, www.nber.org. Moreover, this disagreement shows approximation to the optimal policy

NBER Reporter Summer 2003 29. Innovation Policy and the Economy The NBER’s fourth annual Defense R and D Policy in the Anti- Bronwyn H. Hall and David C. Conference on Innovation Policy Terrorist Era” Mowery, NBER and University of and the Economy, organized by California, Berkeley; Stuart J.H. Research Associates Adam B. Jaffe Adam B. Jaffe; Richard Newell, Graham, University of California, of Brandeis University, Joshua Resources for the Future; and Berkeley; and Dietmar Harhoff, Lerner of Harvard Business School, Robert Stavins, Harvard University, Ludwig-Mazimilians-Universität and Faculty Research Fellow Scott “Technology Policy for Energy and München, “Prospects for Improving Stern of MIT, took place in the Environment” U.S. Patent Quality via Post-grant Washington on April 15. The follow- Opposition” ing papers were discussed: Kathryn Shaw, NBER and Carnegie Mellon University, “The Jeremy Bulow, NBER and Manuel Trajtenberg, NBER and Human Resources Revolution: Is it Stanford University, “The Gaming Tel Aviv University, “Crafting a Productivity Driver?” of Pharmaceutical Patents”

Trajtenberg seeks to analyze the and, accordingly, R and D aimed at than is socially desirable, in the pres- nature of the terrorist threat following providing advanced technological ence of weak or nonexistent environ- 9/11, and to explore the implications means for intelligence is viewed as the mental policies that would otherwise for defense R and D policy. First he cornerstone of defense R and D. This foster such technology. Positive knowl- reviews the defining trends of defense entails developing computerized sen- edge and adoption spillovers and R and D since the cold war, and brings sory interfaces, and increasing the abil- information problems further weaken in pertinent empirical evidence: during ity to analyze vast amounts of data. innovation incentives. While environ- the 1990s, the United States accumu- Both have direct civilian applications, mental technology policy is fraught lated a defense R and D stock ten and therefore the required R and D is with difficulties, a long-term view sug- times larger than any other country’s, mostly “dual use”. Indeed, there is gests a strategy of experimenting with and almost thirty times larger than already a private market for these sys- different policy approaches, and sys- Russia’s. Big weapon systems, key dur- tems, with a large number of players. R tematically evaluating their success. ing the cold war but of dubious signif- and D programs designed to preserve Shaw assesses the empirical evi- icance since then, still figure promi- this diversity and to encourage further dence and policy issues associated with nently, commanding 30 percent of competition may prove beneficial both the human resources “revolution.” current defense R and D spending, vis- for the required R and D, and for the While managers and practitioners have a-vis just about 13 percent for intelli- economy at large. long emphasized the role of human gence and antiterrorism. Trajtenberg Jaffe, Newell, and Stavins analyze resource practices, economists and then examines the nature of the ter- the implications of the interaction of policymakers only recently have begun rorist threat, focusing on the role of market failures associated with pollu- to evaluate the impact of human uncertainty, the lack of deterrence, and tion and the environment with market resource policies on overall productiv- the extent to which security against failures associated with the develop- ity growth. This paper suggests that terrorism is (still) a public good. ment and the diffusion of new tech- advanced human resource practices Drawing from a formal model of ter- nology. These combined market fail- (ranging from team-based problem- rorism developed elsewhere, he ures imply a strong prima facie case for solving to incentive pay for training) explores these and related issues in public policy intervention to foster have facilitated the strong productivity detail. Two strategies to confront ter- environmentally beneficial technology. record experienced since the mid- rorism are considered: fighting terror- Both theory and empirical evidence 1990s, both directly and as a comple- ism at its source, and protecting indi- suggest that the rate and direction of ment to the intensive adoption of vidual targets, which entails a negative technological advance is influenced by information technology. Two implica- externality. Contrary to the traditional incentives from the market and from tions emerge from the analysis. First, case of national defense, security regulation. Environmental policy based the advantages to innovative human against terrorism becomes a mixed pri- on incentive-based approaches is more resource practices can only be realized vate/public good. A key result of the likely to foster cost-effective technolo- when America’s workforce possesses a model is that the government should gy innovation and diffusion than poli- strong human capital foundation. spend enough on fighting terrorism at cy based on command and control Second, although the private sector has its source so as to nullify the incentives approaches. In addition, society’s invest- invested intensively in advanced of private targets to invest in their own ments in the development and diffusion human resource practices, many of security. Intelligence emerges as the of new environmentally beneficial these investments have not been meas- key aspect of the war against terrorism technologies is very likely to be less ured in a consistent way or expensed

30. NBER Reporter Summer 2003 correctly as an accounting matter. The post-grant review process modeled on good but not complete success. Bulow lack of standards by which to measure the European opposition system to argues for per se illegality of settle- workplace organization implies that improve patent quality, reveal over- ments that include side payments or society finds it difficult to identify and looked prior art, and reduce subse- deals which are beneficial to the gener- diffuse productive practices as quickly quent litigation. The authors argue that ic. Further, he shows a number of as possible. the welfare gains to such a system may additional strategies beyond side pay- The recent surge in U.S. patenting be substantial. ments, some highly questionable from and expansion of patentable subject Paragraph IV of the Hatch- an antitrust perspective, that brands matter has increased patent office Waxman Act provides a mechanism have used to keep out generics. backlogs and raised concerns that in for the litigation of pharmaceutical These papers will appear in an some cases patents of insufficient patent infringement disputes. Many of annual volume published by the MIT quality or with inadequate search of these cases have been settled with Press. Its availability will be announced prior art are being issued. At the same “reverse payments” by the brand to in a future issue of the Reporter. They time patent litigation and its costs are the generic in return for delayed gener- can also be found at “Books in rising. Hall, Graham, Harhoff, and ic entry. The FTC has contested a Progress” on the NBER’s website. Mowery explore the potential of a number of these settlements with

*

NBER Reporter Summer 2003 31. Economics of Aging The NBER’s held another in a Thomas E. MaCurdy, NBER and Florian Heiss, University of series of conferences on the eco- Stanford University, “Explanations Mannheim; Michael D. Hurd, nomics of aging on May 2-4. for the Growth in Medicare Cost” RAND and NBER; and Axel Program Director David A. Wise of Discussant: Jonathan S. Skinner, Börsch-Supan, NBER and Harvard University organized this NBER and Dartmouth College University of Mannheim, “How the agenda: Health-Wealth Nexus is Moderated Jonathan S. Skinner, Elliot S. by Family Closeness” James M. Poterba, NBER and Fisher, and John W. Wennberg, Discussant: Steven F. Venti MIT; Joshua Rauh, MIT; Steven F. Dartmouth Medical School, “The Venti, NBER and Dartmouth Efficiency of Medicare?” Arie Kapteyn and Stan Panis, College; and David A. Wise, Discussant: Alan M. Garber RAND, “Individual Retirement “Utility Evaluation of Risk in Planning, Institutions, and Retirement Saving Accounts” James P. Smith, RAND, Outcomes” Discussant: Bob Willis, University “Consequences and Predictors of Discussant: Andrew Samwick, of Michigan New Health Events” NBER and Dartmouth College Discussants: David M. Cutler James Choi, Harvard University; Axel Börsch-Supan, and Lothar David Laibson, NBER and Anne C. Case, NBER and Essig, University of Mannheim, Harvard University; Brigitte Princeton University, and Angus S. “Household Saving in Germany — Madrian, NBER and University of Deaton, NBER and Princeton Results of the First SAVE Study” Chicago; and Andrew Metrick, University, “Broken Down by Work Discussant: Andrew Samwick, NBER and University of and Sex: How our Health Declines” NBER and Dartmouth College Pennsylvania, “Optimal Defaults” Discussant: Daniel L. McFadden, Discussant: Antonio Rangel, NBER NBER and University of California, Li Gan, NBER and University of and Stanford University Berkeley Texas; Michael D. Hurd; and Daniel McFadden, “Individual David M. Cutler, NBER and Robert T. Jensen, NBER and Subjective Survival Curves” Harvard University, “Intensive Harvard University, “Support for Discussant: Bob Willis Medical Technology and the Widows in Underdeveloped Reduction in Disability” Countries” Discussant: Alan M. Garber, NBER Discussant: Esther Duflo, NBER and Stanford University and MIT

Poterba, Rauh, Venti, and Wise findings also demonstrate the impor- find that it is sometimes optimal to set develop a stochastic simulation algo- tance of the equity premium in affect- extreme defaults, which are far away rithm to evaluate the effect of holding ing investors’ utility from different from the mean optimal savings rate. A a broadly diversified portfolio of com- retirement asset allocations. Viewed default that is far away from a con- mon stocks, or a portfolio of index from the beginning of a working sumer’s optimal savings rate may make bonds, on the distribution of 401(k) career, and given the historical patterns that consumer better off, because such account balances at retirement. They of returns on stocks and bonds, a a “bad” default will lead procrastinat- compare the alternative distributions household that does not have extreme- ing consumers to opt out of the of retirement wealth by showing the ly high risk aversion would achieve a default more quickly. The authors cal- empirical distribution of potential higher expected utility by holding a culate optimal defaults for employees wealth values and by computing the portfolio of stocks rather than bonds. at four different companies. Their expected utility of these outcomes Default options have an enormous work suggests that optimal defaults are under standard assumptions about the impact on household “choices.” likely to be at one of three savings structure of household preferences. Defaults matter, because opting out of rates: the minimum savings rate (that Their analysis highlights the critical a default is costly and these costs is, 0 percent); the match threshold role of other sources of wealth, such change over time, generating an option (typically 5 percent or 6 percent); or as Social Security, defined benefit pen- value of waiting. In addition, people the maximum savings rate. sion annuities, and saving outside have a tendency to procrastinate. Choi, Substantial recent evidence shows a retirement plans, in determining the Laibson, Madrian, and Metrick reduction in disability among the eld- expected utility cost of holding equi- develop a theory of optimal defaults erly in the United States, on the order ties in the retirement account. Their based on these considerations. They of 25 percent in the past two decades. 32. NBER Reporter Summer 2003 The major issue raised by these find- While high-cost episodes account for a Smith also examines the ability of dif- ings is why disability has declined. large proportion of total spending ferent measures of SES to predict the Cutler investigates the role of inten- during any year, the majority of elder- future onset of disease. He finds no sive medical technologies in the ly experience such episodes at some predictive effect of income or wealth, decline in disability. Using data from point over a decade, implying far less but education does predict future the National Long-Term Care Survey, concentration in expenditures when onset, even after controlling for cur- he documents that increased use of viewed over lifetimes. Three-fifths of rent health status. intensive procedures might be associ- beneficiaries experience at least one Self-reported health status (SRHS) ated with some reduction in disability, 95-percentile month in the decade, and allows examination of how health sta- but probably does not account for the two-fifths realize one or more 98-per- tus varies over the life course. The majority of the decline. centile months. Knowledge of the SRHS of both men and women deteri- MaCurdy reveals several valuable incidence of 95-percentile months orates with age. In the bottom quartile insights into the growth of Medicare alone explains nearly two-thirds of of income, SRHS declines more rapid- expenditures in recent years. For Medicare spending over the decade, ly with age, but only until retirement example, 20–30 percent of total and spending accumulated for those in age. These and related facts motivate a growth in Medicare program payments the 98-percentile months comprises study of the role of work, particularly from 1989-99, arose from an increase almost four-fifths of total decade manual work, in health decline with in the participation rate; 50–60 percent expenditures. age. The Grossman capital-stock from an increase in average program Technological advances in health model of health implies that declines payments per service recipient; and the care have been shown to yield large in health status are driven, not by the remainder from higher enrollment. In health benefits for the U.S. elderly pop- rate of deterioration of the health sharp contrast to the first half of the ulation. However, less is known about stock, but by the rate of increase of 1990s, total Medicare costs actually fell the marginal or incremental benefits of the rate of deterioration. Case and in the late 1990s. Whereas the lower health care spending at a point in time. Deaton show instead that people in percentiles of the expenditure distri- Skinner, Fisher, and Wennberg use manual occupations have worse SRHS, bution continued to increase through- Medicare claims data on 306 hospital and more rapidly declining SRHS, even out the period, expenditures actually referral regions in the United States to with a comprehensive set of controls fell for the highest-cost users of estimate how different dimensions of for income and education. They also Medicare services. Published statistics health care affect survival rates. They find that much of the differences in extending beyond this sample period find that measures of effective care — SRHS across the income distribution is suggest that the reversal in the total factors that have been well established driven by health-related absence from growth of Medicare expenditures to be efficacious in treatment — are the labor-force, which is a mechanism achieved in the late 1990s was only not associated with Medicare costs, but running from health to income, not temporary; starting in 2000, the overall are associated with greater survival in the reverse. growth in Medicare expenditures the Medicare population after control- Jensen explores two themes relat- reverted to its previous rate and may ling for a variety of health status meas- ing to the well-being of widows in have even accelerated. Annual ures. By contrast, the authors find that India. First, measuring individual wel- Medicare spending is also highly con- a large component of Medicare expen- fare via household income or expendi- centrated among a small segment of ditures — $26 billion in 1996 dollars, ture per person implicitly assumes that the beneficiary population, and shares or nearly 20 percent of total Medicare each person in the household receives of spending attributable to high-cost expenditures — is associated with care an equal share. However, if some indi- users have remained remarkably stable for chronically ill patients and appears viduals systematically receive less, over the 1989-99 decade, even though to provide no benefit in terms of sur- which is typically the case for women growth rates have varied considerably vival, nor is it likely that this extra in upper caste households, this will be during the period and across intensity spending improves the quality of life. a misleading measure of individual of use. Those beneficiaries classified in While secular trends in health care welfare. Jensen finds that while upper the top 2 percent of the annual expen- technology have delivered large health caste widows have significantly higher diture distribution alone account for benefits, variation in health care inten- levels of household expenditure per about one quarter of total expendi- sity at a point in time have not. capita, there is no difference in average tures, and those in the top 5 percent Smith examines the consequences Body Mass Index, a summary measure cover almost half of annual expendi- of new health events on a series of of individual nutritional status, tures. Considering spending by SES-related outcomes: out-of-pocket between upper and lower caste wid- months only reinforces this picture of labor supply, labor force activity, ows; wealthier, upper caste widows are concentration. The top 2 percent of household income, and wealth. For no better-off in terms of nutritional months with spending during a year each of these outcomes, new severe status than poorer but more equally account for around two-fifths of total health events have a significant effect, treated lower caste widows. Second, annual expenditures, and the top 5 per- although most of the impact on Jensen analyzes well-being more cent of months cover nearly two- income and wealth takes place through broadly though measures of practices thirds of yearly Medicare expenditures. labor supply and not medical expenses. and customs relating to widows. He

NBER Reporter Summer 2003 33. finds that the status of widows is bet- net worth and gross wealth should be households prefer to cut consumption ter in villages where women and the highest in Italy; Dutch households if ends do not meet. In particular, the elderly are able to make larger eco- should hold the lowest wealth levels at elderly do not like to use credit cards, nomic contributions, in particular retirement; and the ownership of risky and they eschew debt. The authors where less strength-intensive crops assets should be highest in the United suspect large cohorts differences and (such as rice) are grown. Following States. The authors investigate these will study them once further waves of their husbands’ deaths, widows in and other hypotheses at both the the SAVE panel become available. these areas are less likely to experience macro and micro level and find that the Testing life-cycle models and other declines in treatment by the husband’s data are generally consistent with the economic models of saving and con- family or others in the village, to lose hypotheses. sumption at the micro level requires control of land, or to feel unwelcome Germany is an interesting country knowledge of individuals’ subjective at social events. for the study of saving among older beliefs of their mortality risk. Previous Health, wealth, and where one lives households because nearly everyone studies have shown that individual are important, if not the three most — whether in the middle income responses regarding subjective survival important material living conditions. bracket or richer — saves substantial probabilities are generally consistent There are many mechanisms that sug- amounts in old age. Only households with life tables. However, survey gest that living arrangements and well- in the lowest quarter of the income responses suffer serious problems being derived from health and eco- distribution spend more between the caused by focal responses of zero and nomic status are closely related. Heiss, ages of 60 and 75 than they save. one. Gan, Hurd, and McFadden sug- Hurd, and Börsch-Supan investigate Börsch-Supan and Essig exploit gest using a Bayesian update model the joint evolution of the three condi- newly collected data, the first wave of that accounts for the problems tions, using a microeconometric the so-called SAVE panel, specifically encountered in focal responses. They approach similar to what is known as collected to understand economic, also propose models that help us to “vector autoregressions” in the psychological, and sociological deter- identify how much each individual macroeconomics literature. minants of saving. Overall, they find deviates from the life table in her sub- Kapteyn and Panis analyze retire- extraordinarily stable savings patterns. jective belief. The resulting individual ment saving and portfolio choice in the More than 40 percent of German subjective survival curves have consid- United States, Italy, and the households regularly save a fixed erable variations, can better predict Netherlands. While these countries amount. About 25 percent of German observed survival experience than life enjoy roughly the same standard of liv- households plan their savings and have tables, and are readily applicable in ing, they vary widely in their institu- a clearly defined savings target in testing economic models that require tional organization of retirement mind. Most German household saving individual subjective life expectancies. income provisions. Building on exten- is in the form of contractual saving, The proceedings of this confer- sions of the life cycle model, the such as saving plans, whole life insur- ence will be published as an NBER authors derive hypotheses on the impli- ance, and building society contracts. conference volume by the University cations of institutional differences for This makes the flow of saving rather of Chicago Press. Some of the papers wealth accumulation and portfolio unresponsive to economic fluctua- may also be found at “Books in composition: for example, the ratio of tions, such as income shocks. Most Progress” on the NBER’s website.

*

34. NBER Reporter Summer 2003 Corporate Governance An NBER-Universities Research Francisco Peréz-González, and Woochan Kim, KDI School of Conference on “Corporate Gover- Columbia University, “Inherited Public Policy and Management, nance” took place in Cambridge on Control and Firm Performance” “Does Corporate Governance May 9 and 10. The organizers were Discussant: Marianne Bertrand, Affect Firm Value? Evidence from Bengt Holmström, NBER and MIT, NBER and University of Chicago Korea” and Steven N. Kaplan, NBER and Discussant: Michael Weisbach, University of Chicago. These papers Renee B. Adams, Federal Reserve NBER and University of Illinois were discussed: Bank of New York, and Daniel Ferreira, Getulio Vargas R. Gaston Gelos, International Simi Kedia, Harvard University, Foundation, “A Theory of Friendly Monetary Fund, and Shang-Jin “Do Executive Stock Options Boards” Wei, NBER and Harvard Generate Incentives for Earnings Harley E. Ryan, Jr., Louisiana University, “Transparency and Management? Evidence from State University, and Roy A. International Investor Behavior” Accounting Restatements” Wiggins, III, Bentley College, (NBER Working Paper No. 9260) Narayanan Subramanian, Atreya “Who Is In Whose Pocket? Mariassunta Giannetti and Chakraborty, and Shahbaz Ali Director Compensation, Board Andrei Simonov, Stockholm Sheikh, Brandeis University, Independence, and Barriers to School of Economics, “Which “Performance Incentives, Effective Monitoring” Investors Fear Expropriation? Performance Pressure, and Discussant: George Baker, NBER Evidence from Investors’ Stock Executive Turnover” and Harvard University Picking” Discussant: Jeremy Stein, NBER Mihir A. Desai, NBER and and Harvard University Rajesh K. Aggarwal, Dartmouth Harvard University; Alexander College, and Andrew A. Samwick, Dyck, Harvard University; and Chris McNeil, Pennsylvania State NBER and Dartmouth College, Luigi Zingales, NBER and University, and Greg Niehaus and “Empire-Builders and Shirkers: University of Chicago, “The Eric Powers, University of South Investment, Firm Performance, and Protecting Hand: Taxation and Carolina, “Management Turnover in Managerial Incentives” Corporate Governance” Subsidiaries of Conglomerates Bernie Black, Stanford University; Discussant: Rene Stulz, NBER and Versus Stand Alone Firms” Hasung Jang, Korea University; Ohio State University

Kedia examines the option grants quintile, by option grants, is double that compare turnover of subsidiary man- and option exercises of top executives in the lowest quintile; this difference is agers inside conglomerate firms to of 224 firms that announce restate- significant at the 1 percent level. turnover of CEOs of comparable ments of their financial results because Subramanian, Chakraborty, and stand-alone firms. The authors find of accounting irregularities between Sheikh examine the relationship that subsidiary manager turnover is January 1997 and June 2002. He finds between the optimal incentive contract significantly more sensitive to changes that firms that announce large negative and the firm’s decision to fire a man- in performance and significantly more restatements grant about 50 percent ager for poor performance. They find likely following poor performance more stock options to their top execu- that CEOs with steeper compensation than turnover of CEOs is. Further, for tives in the years prior to the announce- contracts (that is, with greater incen- subsidiary managers, the relationship ment than members of a size-and- tives) are more likely to be fired fol- between turnover and performance is industry-matched control group. Top lowing poor firm performance. Logit significantly stronger when the sub- executives of firms with large negative estimations indicate that, among poor- sidiary operates in an industry that is and large positive restatements also ly performing firms, a CEO receiving related to the parent’s primary indus- exercise significantly more options in incentives at the 60th percentile level is try. These results suggest that boards the years prior to the announcement roughly 10 percent more likely to be of directors are relatively ineffective than members of a size-and-industry- fired than a CEO with incentives at the disciplinarians of CEOs, and that, matched control group. Higher option 40th percentile. Also, the performance despite their other apparent failings, exercises appear to be concentrated pressure was greater in the latter half conglomerate firms have relatively among firms that are subsequently sub- (1997-9) of the sample than in the first strict disciplining mechanisms for sub- ject to SEC enforcement actions, while (1993-6). Increased firing pressure sidiary managers. higher option grants are not concen- might have been one of the factors Peréz-González examines the trated in this group of extreme viola- contributing to the accounting shenani- impact of inherited control on firms’ tors. Further, Kedia finds that the per- gans of the late 1990s. performance. He uses data from man- centage of restating firms in the highest McNeil, Niehaus, and Powers agement successions where the depart-

NBER Reporter Summer 2003 35. ing CEO was a member of the con- fewest financial incentives to monitor. firm, rather than because better-gov- trolling family of the corporation. He Aggarwal and Samwick consider erned firms generate higher reported finds that firms where control is inher- the equilibrium relationships between earnings or pay higher dividends. ited undergo large declines in return incentives from compensation, invest- Unique features of Korea’s corporate on assets and market-to-book ratios ment, and firm performance. Using an governance rules allow the authors to that are not experienced by firms that optimal contracting model, the authors partially address two alternative expla- promote CEOs not related to the con- show that the relationship between nations for these results: signaling trolling family. Consistent with waste- firm performance and managerial (firms signal high quality by adopting ful nepotism, these declines are partic- incentives, in isolation, is insufficient to good governance rules) and endogene- ularly prominent in firms that appoint identify whether managers have private ity (firms with high Tobin’s q choose family CEOs who did not attend a benefits of investment, as in theories good governance rules). The authors selective college. Overall, the results of managerial entrenchment. The use both a two-stage (2SLS) and a strongly suggest that nepotism hurts authors then estimate the joint relation- three-stage (3SLS) least squares firms’ performance by limiting the ships between incentives and firm per- approach; the coefficients are larger scope of labor market competition. formance and between incentives and than estimated earlier and are highly Adams and Ferreira analyze the investment and show that investment significant. This is consistent with cau- consequences of the board’s dual role increases with incentives. Further, they sation running from the exogenous as an advisor and a monitor of man- find that firm performance increases component of corporate governance agement. Because of this dual role, the with incentives at all levels. Taken rules to higher Tobin’s q (and other per- manager faces a trade-off concerning together, these results are not consis- formance variables). the amount of information he disclos- tent with theories of overinvestment Does country transparency affect es to the board. On the one hand, if he based on managers having private ben- international portfolio investment? reveals his information, he gets better efits from investment but are consis- Gelos and Wei examine this and relat- advice. On the other hand, the board tent with managers having private costs ed questions using new measures of may change its opinion of his ability of investment and, more generally, transparency and a unique micro on the basis of his information. The with models of underinvestment. dataset on international portfolio authors’ model shows that the board Black, Jang, and Kim report that holdings of emerging market funds. may choose to pre-commit to reduce corporate governance is an important They distinguish between government its monitoring of the manager in order factor in explaining the market value of and corporate transparency. There is to encourage the manager to share his Korean public companies. The authors clear evidence that funds invest sys- information. Therefore, management- construct a corporate governance tematically less in less transparent friendly boards may be optimal gover- index for 526 companies based prima- countries. Herding among funds tends nance structures under certain circum- rily on responses to a Spring 2001 sur- to be more prevalent in less transpar- stances. The authors discuss some evi- vey of all listed companies by the ent markets. Funds seem to react less dence consistent with the new empiri- Korea Stock Exchange. The index is strongly to macroeconomic news cal implications of their theory. Using based on five subindexes for share- about opaque countries. There is also the insights from the model, they also holder rights, board and committee some evidence that during crises, analyze the differences between a sole structure, board and committee proce- funds flee non-transparent countries board system, such as in the United dures, disclosure to investors, and own- to a greater extent. States, and the dual board system, as in ership parity. In their ordinary least Using a dataset that provides various countries in Europe. squares (OLS) regressions, a moderate unprecedented details on individual Ryan and Wiggins examine the 10 point increase in the corporate gov- investors’ stockholdings, Giannetti and relationship between the structure of ernance index predicts a 6 percent Simonov analyze whether investors outside-director compensation and increase in Tobin’s q and a 14 percent take into account corporate governance board-of-director independence. They increase in market/book ratio. A worst- when they select stocks. After control- consider the relationship between the to-best change in the index predicts a ling for the supply effect via free float structure of director compensation 42 percent increase in Tobin’s q and an and other firm characteristics, the and board size, the percentage of out- 87 percent increase in market/book authors find that all categories of side directors on the board, CEO ratio. This effect is statistically strong investors who generally enjoy only tenure, and CEO/Chair duality. The and robust to choice of performance security benefits (domestic and for- evidence indicates a positive relation- variable (Tobin’s q, market/book, and eign; institutional and small individual ship between characteristics associated market/sales) and to specification of investors) are reluctant to invest in with an independent board and direc- the corporate governance index. Each companies with bad corporate gover- tor compensation that is tied to stock- subindex is a significant predictor of nance. In contrast, individuals who price performance. If barriers to mon- higher Tobin’s q (and other perform- have strong connections with the local itoring are prevalent in a firm, then ance variables). This value effect financial community because they are director compensation provides weaker appears to exist primarily because board members or hold large blocks of incentives. Directors of firms in which investors value the same reported earn- at least some listed companies behave the CEO is entrenched receive the ings more highly for a better-governed differently. They do not care about the

36. NBER Reporter Summer 2003 expected extraction of private benefits Desai, Dyck, and Zingales ana- of the Laffer curve. Finally, the or even prefer to invest in firms where lyze the previously unexplored rela- authors show that an increase in tax there is more room for it. The effect of tionship between corporate gover- enforcement can increase (rather than corporate governance on portfolio nance and corporate taxation. They decrease) the stock market value of a decisions is more pronounced for show that a higher tax rate increases company. They find that corporate tax small and medium-size companies. the level of managerial diversion, while rate changes have smaller (in fact, neg- These findings shed new light on the stronger tax enforcement reduces it. ative) effects on revenues when owner- determinants of investor behavior, They also show that when the corpo- ship is more concentrated and corpo- and suggest that in order to under- rate governance system is ineffective rate governance is worse. This corpo- stand portfolio choices it is important (that is, when it is easy to divert rate governance role of corporate to distinguish between investors who income), or when ownership concen- taxes provides a new rationale for the enjoy private benefits or access private tration levels are high, an increase in existence of a low, well-enforced tax at information and investors who enjoy the tax rate actually can reduce tax rev- the corporate level. only security benefits. enues, generating a corporate version

International Seminar on Macroeconomics The NBER’s 26th Annual Inter- Lars Ljungqvist, Stockholm School Watson, NBER and Princeton national Seminar on Macroeconomics of Economics University, “Understanding Changes (ISOM), organized by Zvi Eckstein, in International Business Cycle Boston University, and Kenneth Danny Quah, London School of Dynamics” West, NBER and University of Economics, and Helen Simpson, Discussants: Denise Osborn, Madison, Wisconsin, was held on Institute for Fiscal Studies, “Spatial University of Manchester, and June 13-14 at the Universitat Pompeu Cluster Empirics” Lucrezia Reichlin, Université Libre Fabra in Barcelona. Jeffrey A. Frankel, Discussants: Albert Marcet, de Bruxelles NBER and Harvard University, and Universitat Pompeu Fabra Francesco Giavazzi, NBER and Neville Francis, Lehigh University, Bocconi University, serve as co-chairs Adi Brender, Bank of Israel, and and Valerie A. Ramey, NBER and of ISOM. The following papers were Allan Drazen, NBER and Tel Aviv University of California, San Diego, discussed at the conference: University, “Where Does the “The Source of Historical Political Fiscal Cycle Really Come Economic Fluctuations: An Analysis Richard Blundell and Ian From?” using Long-Run Restrictions” Preston, University College Discussants: Alessandra Casella, Discussants: Susanto Basu, NBER London, and Luigi Pistaferri, NBER and Columbia University, and University of Michigan, and Stanford University, “Consumption and Kenneth S. Rogoff, Harald Uhlig, Humboldt University Inequality and Partial Insurance” International Monetary Fund Discussants: John Kennan, NBER Andrew B. Abel, NBER and and University of Wisconsin, and Kosuke Aoki, Universitat Pompeu University of Pennsylvania, and Jose-Victor Rios-Rull, NBER and Fabra, and Kalin Nikolov, Bank of Janice C. Eberly, NBER and University of Pennsylvania England, “Rule-Based Monetary Northwestern University, Policy under Central Bank “Investment and Firm Valuation Andreas Hornstein, Federal Learning” with Growth Options” Reserve Bank of Richmond; Per Discussants: Fabio Canova, Discussants: John V. Leahy, NBER Krusell, University of Rochester; Universitat Pompeu Fabra, and V.V. and New York University, and and Giovanni L. Violante, New Chari, NBER and University of Plutarchos Sakellaris, University of York University, “Upgrading Minnesota Maryland Vintage Capital” Discussants: Philippe Aghion, James H. Stock, NBER and NBER and Harvard University, and Harvard University, and Mark W.

Blundell, Pistaferri, and Preston into consumption inequality. Their by contrasting shifts in the distribution use panel data on household consump- framework nests the special cases of of income growth with shifts in the tion and income inequality to evaluate self-insurance and the complete mar- distribution of consumption growth, the degree of insurance against kets assumption. The authors assess and analyzing the way these two meas- income shocks. They aim to describe the degree of insurance, over and ures of household welfare correlate the transmission of income inequality above self-insurance through savings, over time. Combining panel data on

NBER Reporter Summer 2003 37. income with consumption data, they gated data on the U.K. manufacturing ticular the emergence of two groups find some partial insurance but reject sector. — one consisting of Euro-zone coun- the complete markets restriction. They Whereas a political fiscal cycle once tries and the other of English-speak- also find a greater degree of insurance was though to be a phenomenon of ing countries — within which correla- for transitory shocks and differences less developed economies, some recent tions have increased and across which in the degree of insurance over time. studies find such a cycle in a large correlations have decreased. Further, Finally, they document the importance cross section of both developed and cyclical movements in the United of durables and of taxes and transfers developing countries. Brender and Kingdom became less correlated with as a means of insurance. Drazen re-examine these empirical Euro-zone countries and more corre- Hornstein, Krusell, and Violante results and show that they are not lated with North American countries argue that certain ingredients are likely robust to the choice of countries and in the period they study. Second, com- to be important for understanding time periods. The authors show that mon international shocks have been how wage distributions and employ- the finding of a political fiscal cycle is smaller in the 1980s and 1990s than ment respond to technology. First, it driven by the experience of “new they were in the 1960s and 1970s. This matters whether there are labor market democracies,” in which fiscal manipu- declining volatility of common G-7 frictions. Matching frictions, and asso- lation may “work” because of lack of shocks is the source of much of the ciated departures from marginal-prod- experience with electoral politics or observed moderation in individual uct wage determination, are important. lack of information that voters in country business cycles. Moreover, this Second, the form of technological more established democracies use. The moderation of common G-7 shocks is change matters. Capital embodiment is strong fiscal cycle in those countries responsible, in a mechanical sense, for a key characteristic of technological accounts for the finding of a fiscal the failure of business cycles to change. Third, how new capital is cycle in larger samples that include become more synchronous as one introduced into the economy — what these countries. Once these countries might expect given the large increase firms acquire it, and who decides on are removed from the larger sample, in trade over this period: had world the purchase – is important. Fourth, the political fiscal cycle disappears. shocks been as large in the 1980s and labor market institutions are highly rel- These findings also reconcile two con- 1990s as they were in the 1960s and evant for both wage distribution and tradictory views of pre-electoral 1970s, then international cyclical cor- unemployment, especially in analyzing manipulation, one arguing it is a useful relations would have increased consid- the effects of technological change. instrument to gain voter support and a erably. Third, the Japanese experience Thus this paper analyzes a matching widespread empirical phenomenon, in many ways is exceptional. For the model with vintage capital and capital- the other arguing that voters punish other G-7 countries, volatility general- embodied technological change. The rather than reward fiscal manipulation. ly decreased or at least stayed constant authors consider the importance of Aoki and Nikolov evaluate the in the 1990s, but it increased in Japan where and how new capital enters, and performance of three kinds of rule- in the 1990s. During the 1980s and contribute to the literature on based monetary policies under central 1990s, cyclical fluctuations in Japanese “Europe-United States comparisons” bank-learning about the parameter val- GDP almost became detached from by emphasizing how the institutional ues of a simple New Keynesian model. the other G-7 economies, with domes- setting can matter for the effects of The central bank and the private sector tic shocks explaining almost all of the technology in general and for techno- learn the slopes of the IS and the cyclical movements in Japanese GDP. logical revolutions, such as the new Phillips curves by recursive least Fourth, however measured, persist- “IT era,” in particular. squares estimation, and form expecta- ence of disturbances has increased in A useful model of economic geog- tions and set policy based on their esti- Canada, France, and the United raphy should determine not just how mated model. The three policies the Kingdom. In those countries, a shock much economic activity occurs at a authors evaluate are: 1) the optimal of a given magnitude would result in given location but also where that loca- non-inertial rule; 2) the optimal histo- more cyclical volatility today than 30 tion is relative to others. In empirical ry-dependent rule; and 3) the years ago. studies, those latter features appear Wicksellian rule. They show that the Francis and Ramey investigate the nowhere in measures of concentration optimal Wicksellian rule delivers the source of historical fluctuations in or in cross-section and panel data highest welfare, by improving the annual U.S. and U.K. data extending regression. Indeed, they are absent in inflation-output gap variability trade- back to the nineteenth century. They many analytical models of economic off, without introducing undesirable use long-run identifying restrictions to geography. Quah and Simpson pro- feedback from past policy mistakes decompose shocks into technology vide a new econometric method for that are caused by imprecise parameter shocks and other shocks. For the U.S. studying clustering that explicitly takes estimates. data, a variety of models with differing into account such spatial relations, Stock and Watson find first that, auxiliary assumptions are investigated. motivating the analysis with a theoreti- although there has not been a general In the United States, the impact of cal model of dynamically evolving spa- increase in international synchroniza- technology shocks on labor input in tial distribution. They then apply their tion among G-7 business cycles, there the pre-WWII period is the opposite techniques to geographically disaggre- have been important changes, in par- of its impact in the post-WWII period

38. NBER Reporter Summer 2003 in most models. The U.K. data shows investment and Tobin’s Q, while main- and cash flow, consistent with that more sample stability, with the short- taining a positive correlation between found in empirical studies. Growth run impact of technology on labor investment and cash flow. By simulat- options also generate excess volatility being negative. The decomposition ing the model, Abel and Eberly show of firm value relative to cash flows. also reveals important changes in the that there is also a high frequency neg- These papers will be published in a volatility of shocks over time. ative relation between investment and special issue of the European Economic The presence of growth options in cash flow shocks that can reverse these Review. Many of them are also available the value of the firm generates vari- effects. However, time aggregation at “Books in Progress” on the NBER’s ability in firm value that is not driven restores the weak relationship between website. by current cash flows. This variability investment and Q, and a strong posi- diminishes the correlation between tive association between investment

Bureau News

President Bush Taps NBER Economists for CEA President Bush has nominated two and professor of economics at Princeton ed in May. She has also served as more NBER researchers to become University, was nominated in April. He deputy assistant secretary of the members of his Council of Economic holds a Ph. D. from Harvard University Treasury Department in the current Advisers. Harvey S. Rosen and Kristin and served as deputy assistant secretary Bush Administration. Forbes will be joining N. Gregory for tax analysis in the administration of NBER Research Associate Andrew Mankiw, a Harvard economics profes- George H. W. Bush. A. Samwick, a member of the eco- sor and NBER Research Associate, Forbes, an NBER Faculty Research nomics faculty at Dartmouth College, who is CEA Chairman. Fellow and professor at MIT's Sloan is the CEA’s Chief Economist. Rosen, an NBER Research Associate School of Management, was nominat-

Levitt Receives John Bates Clark Medal NBER Research Associate Steven Programs on Public Economics, Law MIT in 1994. D. Levitt, a professor of economics at and Economics, Children, and Previous Clark medalists among the University of Chicago, received the Education. He received the medal for NBER Research Associates were: Zvi American Economics Association’s his work on crime and corruption, Griliches, Daniel L. McFadden, Martin John Bates Clark Medal this year. The including studies of teacher cheating at S. Feldstein, Joseph E. Stiglitz, James J. award is given every two years to the Chicago schools, corruption in Heckman, Jerry A. Hausman, Sanford economist under the age of 40 who Japanese sumo wrestling circles, and J. Grossman, Paul R. Krugman, has made the most substantial contri- the economics of gang life. Levitt Lawrence H. Summers, David Card, bution to the field of economics. received his B.A. from Harvard Kevin M. Murphy, and Andrei Shleifer. Levitt is a member of the NBER’s University in 1989 and his Ph.D. from

NBER Reporter Summer 2003 39. NBER Researchers Lead American Economics Association Three NBER Research Associates current AEA President. NBER The AEA’s Executive Committee are the President, President-Elect, and President Martin Feldstein, a professor also has chosen Daniel L. McFadden as nominee for President-Elect, respec- of economics at Harvard University and the nominee for President-Elect, to tively, of the American Economics a member of the NBER’s Programs on succeed Feldstein. McFadden is a pro- Association (AEA). Peter A. Diamond, Aging, Health Care, Public Economics, fessor of economics at the University an MIT economics professor and and Economic Fluctuations and of California, Berkeley, and a member member of the NBER’s Programs on Growth, is President-Elect. He is slated of the NBER’s Program on Aging. Monetary Economics and on Eco- to become the AEA’s President in nomic Fluctuations and Growth, is the January, succeeding Diamond.

NBER Announces 2003-4 Nonprofit Fellowships NBER Research Associate James Bailey, Vanderbilt University, whose Technology Charter School, Boston”; M. Poterba has announced the recipi- topic is “Nonprofit Organizations Esther Duflo, MIT, who will focus on ents of NBER Fellowships for the Providing Family Planning Services”; “The Effectiveness of Nonprofit Study of Nonprofit Institutions for the Guy David, University of Chicago, who Organizations in Developing Coun- 2003-4 academic year. This NBER fel- is studying the “Performance of Not- tries”; Charles Phelps, University of lowship program is designed to for-profit versus For-Profit Hospitals”; Rochester, whose research is on encourage research on nonprofit insti- Dan Hungerman, Duke University, for “Governance in Not-for-Profit Hospi- tutions by NBER Research Associates “Behavior of Organized Religious tals and Universities”; David Reiley, and Faculty Research Fellows, and to Organizations”; and Peter Katuscak, University of Arizona, who will analyze support dissertation research on the University of Michigan, for “Managerial “Fundraising Mechanisms Used by same subject by graduate students in Pay in Nonprofit Organizations.” Charitable Organizations”; and Michael economics who work closely with Faculty grants were made to: Rothschild, Princeton University, who them. Joshua Angrist, MIT, and Kevin Lang, will ask “How Does California Produce The four graduate students who will Boston University, who will study “A First-Rate Research Universities?” receive fellowship support are: Martha Nonprofit Charter School: Media and

Thomas D. Flynn, Director Emeritus, Dead at 90 Thomas D. Flynn, who became a accounting from the Columbia named him to head the Municipal Director of the NBER in 1967 repre- Graduate School of Business in 1938. Assistance Corporation, an agency set senting the American Institute of After a short stint on the staff of the up to help manage New York City’s Certified Public Accountants as its Federal Trade Commission, Flynn troubled finances. He remained a mem- president, died on April 12 at age 90. joined the accounting firm Arthur ber of that Board until 1979. He has been a Director Emeritus of Young & Company in 1940. He rose to Flynn was also a trustee emeritus of the NBER since 1968. senior partner before being named vice Columbia University. He leaves his Flynn was a 1935 graduate of chairman of the firm’s management wife of 62 years, Harriett Howland Princeton University with a degree in committee. When he retired in 1975, Flynn, two daughters, one son, and six economics. He received his M.B.A. in Governor Hugh L. Carey of New York grandchildren.

2002 Japan Conference: A Summary of the Papers The NBER recently published a papers on the Japanese economy. This are available as NBER Working Papers booklet summarizing the papers and year, there was also a panel discussion and can be found at: discussions presented at a September on monetary policy in Japan, and a http://www.nber.org/papers/ 2002 conference in Tokyo. This con- luncheon address by Haruhiko An electronic version of this booklet is ference, which has been held annually Kuroda, at the time the Vice Minister also available at: for the past several years, presents of the Japanese Ministry of Finance. http://www.nber.org/2002japanconf/ some of the best English language The full versions of these papers

40. NBER Reporter Summer 2003 Labor Studies The NBER’s Program on Labor Norms Governing Exploding Distributional Effects of Welfare Studies met in Cambridge on April 4. Offers Affect Market Performance” Reform Experiments” Program Director Richard B. Freeman, and Lawrence F. Katz, Juan Botero, Yale University; Paul Gertler, NBER and University NBER and Harvard University, Simeon Djankov, World Bank; of California, Berkeley; Manisha organized the meeting. These papers Rafael LaPorta and Andrei Shah, University of California, were discussed: Shleifer, NBER and Harvard Berkeley; and Stefano Bertozzi, University; and Florencio Lopez- INSP, Mexico, “Risky Business: the Henry S. Farber, NBER and de-Silanes, NBER and Yale Market for Unprotected Princeton University, “The Labor University, “The Regulation of Commercial Sex” Supply of New York City Cab Labor” Drivers” Paul Glewwe, University of Marianne Bitler, RAND; Jonah B. Minnesota; Nauman Ilias, Muriel Niederle, NBER and Gelbach, University of Maryland; Competition Economics, Inc; and Stanford University, and Alvin E. and Hilary Hoynes, NBER and Michael Kremer, NBER and Roth, NBER and Harvard University of California, Davis, Harvard University, “ Teacher University, “Market Culture: How “What Mean Impacts Miss: Incentives”

Farber models the labor supply of Many markets have organizations collective bargaining laws, and social taxi drivers, suggesting that because that influence or try to establish norms security laws in 85 countries. They find income effects in response to tempo- concerning when offers can be made, that richer countries regulate labor less rary fluctuations in daily earnings accepted, and rejected. An examina- than poorer countries do, although the opportunities are likely to be small, tion of a dozen previously studied former have more generous social cumulative hours will be much more markets suggests that markets in which security systems. The authors’ meas- important than cumulative income in transactions are made far in advance ures of the political power of the left the decision to stop work on a given are those in which it is acceptable for are associated with more stringent day. However, if income effects are firms to make exploding offers, and labor regulations and more generous large because of very high discount where it is unacceptable for workers to social security systems, although the and interest rates, then labor supply renege on commitments they make, former result is not robust. Finally, functions could bend backward and, in however early. Still, markets differ in countries with socialist and French the extreme case where the wage elas- many ways other than norms concern- legal origin have sharply higher levels ticity of daily labor supply is minus ing offers. Laboratory experiments of labor regulation than do common one, drivers could be target earners. allow Niederle and Roth to isolate law countries. These results are diffi- Indeed, earlier studies by other the effects of exploding offers and cult to reconcile with straightforward researchers find that the daily wage binding acceptances. In a simple envi- versions of efficiency and with politi- elasticity of labor supply of New York ronment, where uncertainty about cal power theories of institutional City cab drivers is substantially nega- applicants’ quality is only resolved over choice, but are broadly consistent with tive; they conclude that cab drivers time, the authors find inefficient early legal theories, according to which probably are target earners. However, contracting when firms can make countries have pervasive regulatory based on new data, Farber concludes exploding offers and applicants’ styles inherited from the transplanta- that — when accounting for earnings acceptances are binding. Relaxing tion of legal systems. opportunities in a reduced form with either of these two conditions causes Bitler, Gelbach, and Hoynes use measures of clock hours, day of the matching to take place later, when experimental data to estimate quantile week, weather, and geographic loca- more information about applicants’ treatment effects (QTEs) of tion — cumulative hours worked on qualities is available, and consequently Connecticut’s Jobs First reform. They the shift is a primary determinant of results in higher efficiency and fewer find considerable evidence of system- the likelihood of stopping work while blocking pairs. This suggests that ele- atic heterogeneity, which is generally cumulative income earned on the shift ments of market culture may play an consistent with theoretical predictions. is related weakly, at best, to the likeli- important role in influencing market Moreover, they find that mean impacts hood of stopping work. This is consis- performance. computed for subgroups would not tent with the existence of inter-tem- Botero, Djankov, LaPorta, have uncovered a number of impor- poral substitution and is inconsistent Lopez-de-Silanes, and Shleifer tant findings. For example, earnings at with the hypothesis that taxi drivers are investigate the regulation of labor the very top of the distribution fall, as target earners. markets through employment laws, theory predicts should happen when

NBER Reporter Summer 2003 41. more generous policies encourage clients are willing to pay more to avoid primary school teachers in rural Kenya women to stay on or re-enter welfare. using condoms. Using a panel data set with incentives based on students’ test Also, Jobs First has essentially no from Mexico, the authors estimate that scores and dropout rates. Students in impact on earnings at the bottom of commercial sex workers received a 23 program schools had higher test the earnings distribution: income at percent premium for unprotected sex scores, significantly so on at least some the bottom deciles is constant before from clients who requested not to use exams, during the time the program time limits hit (contrary to large posi- a condom. However, this premium was in place. An examination of the tive mean impacts for a disadvantaged jumped to 46 percent if the sex work- channels through which this effect subgroup) and falls after time limits er was considered very attractive. took place, however, provides little evi- begin to take effect (again contrary to These results suggest that the current dence of more teacher effort aimed at a zero effect using means). These find- supply-side policies aimed at educating increasing long-run learning. Teacher ings suggest cause for concern about sex workers about risk and empower- attendance did not improve, home- the ability of programs like Jobs First ing them is not sufficient to signifi- work assignments did not increase, to end welfare dependence of signifi- cantly increase condom use. Rather, dropout rates did not fall, and peda- cant numbers of women. complementary interventions aimed at gogy did not change. However, there is Each day close to 20,000 people reducing the demand for not using evidence that teachers increased effort become infected with the HIV virus condoms are needed. at manipulating short-run test scores. worldwide; a large portion of them are Advocates of teacher incentive Conditional on being enrolled, stu- infected through unprotected sex with programs argue that they can strength- dents in treatment schools were more commercial sex workers. While con- en weak incentives, while opponents likely to take tests. Teachers in treat- doms are an effective defense against argue that they lead to “teaching to the ment schools scored higher than their the transmission of HIV and other test.” Glewwe, Ilias, and Kremer find counterparts in comparison schools sexually transmitted infections, large that existing teacher incentives in during the life of the program; they numbers of sex workers are not using Kenya are indeed weak, with teachers did not retain these gains after the end them with their clients. Gertler, Shah, absent 20 percent of the time. The of the program, consistent with a and Bertozzi argue that sex workers authors then report on a randomized model in which teachers focused on are willing to take the risk because evaluation of a program that provided manipulating short-run scores.

Monetary Economics The NBER’s Program on Measure of Monetary Shocks: Nash Inflation” Monetary Economics met in Derivation and Implications” Discussant: Tao Zha, Federal Cambridge on April 11. Susanto Discussant: Jeffrey Fuhrer, Federal Reserve Bank of Atlanta Basu, NBER and University of Reserve Bank of Boston Michigan, and Michael Woodford, Sylvain Leduc, Keith Sill, and NBER and Princeton University, Andreas Beyer, European Central Tom Stark, Federal Reserve Bank organized this program: Bank, and Roger E. A. Farmer, of Philadelphia, “Self-Fulfilling University of California, Los Expectations and the Inflation of Paul Beaudry, NBER and Angeles, “Identifying the Monetary the 1970s: Evidence from the University of British Columbia, and Transmission Mechanism using Livingston Survey” Franck Portier, University of Structural Breaks” Discussant: Athanasios Orphanides, Toulouse, “Stock Prices, News, and Discussant: Mark W. Watson, Federal Reserve Board Economic Fluctuations” NBER and Princeton University Discussant: Julio J. Rotemberg, Michelle Alexopoulos, University NBER and Harvard University Thomas J. Sargent, NBER and of Toronto, “A Monetary Business New York University, and Noah Cycle Model with Unemployment” Christina D. Romer and David H. Williams, NBER and Princeton Discussant: Garey Ramey, Romer, NBER and University of University, “Impacts of Priors on University of California, San Diego California, Berkeley, “A New Convergence and Escapes from

A common view in macroeconom- tions driven by unexpected changes in driven primarily by a shock that does ics is that business cycles can be technology. Beaudry and Portier pro- not affect productivity in the long run. decomposed meaningfully into fluctu- pose a means of evaluating this view The structural interpretation they sug- ations driven by demand shocks — and show that it is strongly at odds gest for this shock is that it represents shocks that have no short- or long-run with the data. They show instead that news about future technological oppor- effects on productivity — and fluctua- the data favor a view of business cycles tunities. They show that this shock

42. NBER Reporter Summer 2003 explains about 50 percent of business prior measures. This suggests that pre- tainability of Ramsey inflation cycle fluctuations and therefore de- vious measures of policy shocks are encountered in the analyses of Sims serves to be acknowledged and further significantly contaminated by forward- (1988) and Chung (1990), on the one understood by macroeconomists. looking Federal Reserve behavior and hand, and Cho, Williams, and Sargent Conventional measures of mone- endogeneity. (2002), on the other. tary policy, such as the federal funds Beyer and Farmer propose a Using survey data on expectations, rate, are surely influenced by forces method for estimating a subset of the Leduc, Sill, and Stark examine other than monetary policy. More parameters of a structural rational whether the post-war data are consis- importantly, central banks adjust poli- expectations model by exploiting tent with theories of a self-fulfilling cy in response to a wide range of changes in policy. They define a class inflation episode during the 1970s. information about future economic of models, midway between a vector Among commonly cited factors, oil developments. As a result, estimates of autoregression (VAR) and a structural and fiscal shocks do not appear to have the effects of monetary policy derived model, that they call the recoverable triggered an increase in expected infla- using conventional measures tend to structure. They provide an application tion that was subsequently validated by be biased. To address this problem, of their method by estimating the monetary policy. However, the evi- Romer and Romer develop a measure parameters of a three-equation model dence suggests that, prior to 1979, the of monetary policy shocks in the of the monetary transmission mecha- Fed accommodated temporary shocks United States from 1969 to 1996 that nism using data from 1970:Q1 to to expected inflation, which then led to is relatively free of endogenous and 1999: Q4. They first estimate a VAR permanent increases in actual inflation. anticipatory movements. To address and find that its parameters are unsta- The authors do not find this behavior the problem of forward-looking ble. However, using their proposed in the post-1979 data. behavior, the authors control for the identification method, they are able to To reproduce the high variation of Federal Reserve’s forecasts of output attribute instability in the parameters employment and low variation of real and inflation prepared for scheduled of the VAR solely to changes in the wages observed in the post-war U.S. FOMC meetings. They remove from parameters of the policy rule. They data, many business cycle models must their measure policy actions that are a recover parameter estimates of the assume that there are high price systematic response to the Federal recoverable structure and demonstrate markups and that agents have high Reserve’s anticipations of future devel- that these parameters are invariant to labor supply elasticities. The problem opments. To address the problem of changes in policy. Since the recover- with these assumptions is that micro- endogeneity, and to ensure that the able structure includes future expecta- economic evidence indicates that forecasts capture the main information tions as explanatory variables, their markups and labor supply elasticities the Federal Reserve had at the times parameter estimates are not subject to are generally low. To eliminate the need decisions were made, they consider the Lucas [24] critique of econometric for these assumptions, Alexopoulos only changes in the intended federal policy evaluation. introduces a labor market friction into funds rate around scheduled FOMC Recent papers have analyzed how a monetary general equilibrium model meetings. The series for intended adaptive agents may converge to and with limited participation. In the changes spans periods when the escape from self-confirming equilibri- model, firms imperfectly observe their Federal Reserve was not exclusively ums. All of these papers have imputed workers’ effort levels; detected shirkers targeting the funds rate. It is derived to agents a particular prior about drift- forgo an increase in their compensa- using information on the expected ing coefficients. In the context of a tion; and households make their deci- funds rate from the records of the model of monetary policy, Sargent sions about their level of monetary Open Market Manager and informa- and Williams analyze dynamics that deposits for the period before seeing tion on intentions from the narrative govern both convergence and escape the shocks to the economy. The esti- records of FOMC meetings. Estimates under a more general class of priors mated model is able to account for the of the effects of monetary policy for the government. They characterize observed variation in employment and obtained using the new measure indi- how the shape of the prior influences real wages over the business cycle, and cate that policy has large, relatively the dynamics in important ways. There is consistent with the existing evidence rapid, and statistically significant are priors for which the E-stability about the qualitative responses of the effects on both output and inflation. condition is not enough to assure local U.S. economy to fiscal and monetary The authors find that the effects using convergence to a self-confirming equi- policy shocks. the new measure are substantially librium. This analysis also tracks down stronger and quicker than those using the source of differences in the sus-

NBER Reporter Summer 2003 43. Higher Education The NBER’s Working Group on Richard A. Jensen, University of Malcolm Getz and John J. Higher Education met in Cambridge Notre Dame, and Marie Thursby, Siegfried, Vanderbilt University, on May 2. Charles T. Clotfelter, NBER and Georgia Institute of “Where Do the Children of NBER and Duke University, organ- Technology, “The Academic Effects Professors Attend College?” ized this program: of Patentable Research” Discussant: Christopher Avery, Discussant: Jerry R. Green, NBER NBER and Harvard University Julie B. Cullen, NBER and and Harvard University University of Michigan; Mark C. Larry D. Singell, Jr., and Joe A. Long, George Washington Sandy Baum, Skidmore College, Stone, University of Oregon, “For University; and Randall Reback, and Eben Goodstein, Lewis and Whom the Pell Tolls: Market Power, University of Michigan, “Jockeying Clark College, “Affirmative Action Tuition Discrimination, and the for Position: High School Student for Guys? The Consequences of Bennett Hypothesis” Mobility and Texas’ Top-Ten Gender Imbalance in College Discussant: Bridget T. Long, Percent Rule” Applications” Harvard University Discussant: Caroline M. Hoxby, Discussant: Linda Loury, Tufts NBER and Harvard University University

Beginning in 1998, all high school ity of each type of research and its likely to be dominated by men. students in the state of Texas who productivity in generating income and To ask whether the best-informed graduated in the top 10 percent of prestige. In equilibrium, if license consumers of higher education — the their class were guaranteed admission income is the same whether the faculty — make different choices from to any public higher education institu- researcher remains in the university or other similarly endowed consumers, tion, including the University of Texas. not, an increase in the share of income Siegfried and Getz compare the pat- While the goal of the policy was to leads to no change in optimal teaching tern of colleges chosen by 5,592 chil- improve access for disadvantaged and load or salary, or to a change in oppo- dren of college and university faculty minority students, the use of a school- site directions. A decrease in the por- with the pattern chosen by the chil- specific standard to determine eligibil- tion of patentable knowledge that can dren of non-faculty families of similar ity could have unintended conse- be used in education either has no socioeconomic status. The patterns are quences. Students may benefit from effect on the optimal teaching load or remarkably different. The children of switching schools near the end of their salary, or causes them to move in faculty are more likely to choose high school career in order to change opposite directions. Results on the research universities and even more their peer reference group and to quality of education are typically likely to choose selective liberal arts increase the chances of being in the ambiguous because changes in the colleges. This evidence is consistent top 10 percent. In their analysis of stu- teaching load induce changes in with the view that the level of infor- dent mobility patterns before and after research that have an opposite effect mation makes a difference in the the policy change, Cullen, Long, and on educational quality. choice of college. Reback find evidence that these Using data from 13 liberal arts col- Are federal Pell grants “appropriat- strategic moves did occur. leges, Baum and Goodstein test for ed” by universities through increases Jensen and Thursby examine con- affirmative action for men in the col- in tuition — consistent with what is cerns about whether recent changes in lege admissions process. Despite the known as the Bennett hypothesis? patent policy that increased the ability relative shortage of men in the appli- Based on a panel of 71 universities and the incentives for U.S. universities cant pools, except for some formerly from 1983 to 1996, Singell and Stone to patent inventions have been detri- female institutions, gender either has find little evidence of the Bennett mental to academic research and edu- no effect on the probability of admis- hypothesis among either public or cation. They analyze a model in which sion or there is a slight preference for lower-ranked private universities. For a researcher allocates time between women. The authors do find, however, top-ranked private universities, applied and basic research, given that the bottom quartile of both the though, increases in Pell grants appear administration choices of salary and applicant and acceptance pools, as to be more than matched by increases teaching load. The choice depends on measured by academic record, is dis- in net tuition. The behavior most con- a “composite” marginal rate of substi- proportionately male. As a result, even sistent with this result is price discrim- tution of applied-for-basic research with a gender-blind admissions policy, ination that is not purely redistributive which takes into account both the util- the lower tail of college classrooms is from wealthier to needier students.

44. NBER Reporter Summer 2003 International Trade and Organizations The NBER’s Working Group on London, “Competing for “Subcontractors for Tractors: International Trade and Ownership” Theory and Evidence on Flexible Organizations met in Cambridge on Specialization, Supplier Selection, May 3. Gordon Hanson, NBER Pol Antràs, NBER and MIT, and Contracting” and University of California, San “Incomplete Contracts and the Diego, organized the meeting. Product Cycle” Dalia Marin, University of These papers were discussed: Munich, and Thierry Verdier, Tahir Andrabi, Pomona College; DELTA, Paris, “Globalization and Patrick Legros, Princeton Maitreesh Ghatak, London School the Empowerment of Talent” University, and Andrew F. of Economics; and Asim Ijaz Newman, University College, Khwaja, Harvard University,

Legros and Newman show that product cycles — that is, incomplete their greater willingness to invest more vertical integration may be chosen by contracts — opens the door to a paral- in assets specific to the buyer, especial- managers to the detriment of con- lel analysis of the determinants of the ly in the presence of contracting prob- sumers, even in the absence of mode of organization. As a result, a lems and high uncertainty. Andrabi, monopoly power in either supply or new version of the product cycle Ghatak, and Khwaja examine a pri- product markets. This effect is most emerges, in which manufacturing mary dataset on contracts between the likely to occur when demand is initial- shifts to the South first within firm largest tractor assembler in Pakistan ly high and there is a negative supply boundaries, and only later to inde- and its suppliers and ask how the shock, or when demand is low and pendent firms in the South. Relative to extent of asset specificity and other there is a positive demand shock. a world with only arm’s length trans- supplier characteristics affect contrac- Therefore, there is a need for scrutiny acting, allowing for intrafirm produc- tual outcomes, such as prices and dis- of vertical mergers even in the absence tion transfer by multinational firms tribution of orders. They find that the of market power. This result is robust accelerates the shift of production more dedicated suppliers are indeed of to the introduction of active share- towards the South, while having an lower quality. holders who may oppose the merger ambiguous effect on relative wages. Globalization has been identified and to other extensions. Antràs also discusses several micro- by many experts as a new way for The incomplete nature of con- economic implications of the model firms to organize their activities and as tracts governing international transac- and relates them to the empirical liter- the emergence of talent as the new tions limits the extent to which the ature on the product cycle. stakeholder in the firm. Marin and production process can be fragmented Andrabi, Ghatak, and Khwaja Verdier examine the role of trade inte- across borders. In a dynamic, general- develop a simple model of flexible gration for the changing nature of the equilibrium Ricardian model of specialization under demand uncer- corporation. International trade leads North-South trade, the incomplete- tainty. A buyer faces multiple suppliers to a “war for talent.” This makes it ness of contracts leads to the emer- with heterogeneous quality who have more likely that in the integrated world gence of product cycles. Goods initial- the option of selling to other buyers. economy an organizational equilibri- ly are manufactured in the North, The more specific a seller’s assets are um emerges in which control is dele- where product development takes to the buyer, the lower is his flexibility gated to lower levels of the firms’ hier- place. As the good matures and to cater to the outside market, and this archy, empowering human capital. becomes more standardized, manufac- cost is greater for higher quality sup- Furthermore, trade integration leads to turing shifts to the South, benefiting pliers. Therefore, even if a buyer typi- waves of outsourcing and to conver- from lower wages. Following the prop- cally prefers high quality suppliers, gence in corporate cultures across erty-rights approach to the theory of some low quality suppliers might be countries. the firm, the same force that creates kept as marginal suppliers because of

NBER Reporter Summer 2003 45. Health Care Members of the NBER’s Ahmed W. Khwaja, Duke Nancy Beaulieu, NBER and Program on Health Care met in University, “Some Economic Harvard University, “Externalities in Cambridge on May 9. Program Consequences of Being Overlapping Supply Networks” Director Alan M. Garber, of (Medi)Careless” Stanford University, organized the Derek S. Brown, Duke University; meeting. These papers were dis- M. Kate Bundorf and Laurence Gabriel Picone, University of cussed: C. Baker, NBER and Stanford South Florida; and Frank A. Sloan, University, “The Market Level NBER and Duke University, “Do Jonathan S. Skinner and Douglas Effects of Health Plan Quality Routine Eye Exams Improve O. Staiger, NBER and Dartmouth Measurement and Reporting” Vision?” College, “Technological Innovation and Health Outcomes: A New Look”

There are a variety of models that Medicare on life-cycle health related information, independent of the seek to explain the process of techno- behaviors, that is, alcohol consump- extent to which it is used by con- logical innovation, but less is under- tion, smoking, and exercise. The simu- sumers, resulting in improvements in stood about the nature of innovation lations from the model show that, in the practice of medicine. Bundorf and and diffusion in health care. Most the absence of Medicare, the health Baker investigate the extent to which models of technological innovation outcomes fall marginally while the uti- reporting on quality among managed imply convergence; gradually, even the lization of medical care falls dramati- care plans within an area affects area- laggard adopters of new technology cally. Further, in the absence of level health care delivery patterns. catch up to the innovators. And, while Medicare alcohol consumption, smok- They study a nationally representative different hospitals or regions may ing, and exercise also increase. This sample of individuals with employer- develop their own strategies for treat- suggests that Medicare induces signifi- sponsored health insurance living in ing specific diseases, the productivity cant moral hazard — that is, over-con- metropolitan areas during 1996-9. The of their strategies — that is, their qual- sumption of medical care — with authors consider four services: breast ity-adjusted price — should exhibit small consequent effects in terms of cancer screening for women 52-69; convergence over time. Using improving health outcomes. Medicare cervical cancer screening for women Medicare data on 2.6 million heart also provides an incentive to reduce 21-64; eye exams for diabetics; and attack patients during 1989-2000, the risky habits of smoking and alco- annual check-ups for adults. They find Skinner and Staiger find first that the hol consumption. But the largest ben- that area-level reporting on health plan rapid technological gains occurring efit of Medicare is in terms of mitigat- quality is associated with area-level uti- during the late 1980s and early 1990s ing financial loss, especially for the eld- lization of some, but not all, measured have flattened by the late 1990s; indeed erly. The presence of Medicare services. Rates of screening for cervi- the quality-adjusted price has risen reduces out-of-pocket medical expen- cal cancer and eye exams for diabetics since 1995. Second, in considering ditures by large amounts. Thus, in sum were higher in markets with a greater regional differences in mortality costs the benefits of Medicare are primarily proportion of HMOs participating in and outcomes, the authors find no evi- in terms of providing financial protec- HEDIS (Health Plan Employer Data dence of convergence with respect to tion against illness. There is a small and Information Set) quality measure- productivity during this period. The benefit in terms of improving health ment. Mammogram and physical exam evidence favors long-term differences outcomes, because of better access to rates, in contrast, appear not to have across states in their propensity to care. But Medicare leads to over-con- been affected by HEDIS quality meas- adopt new technology; states likely to sumption of medical care. Finally, urement and reporting activity. The adopt hybrid corn in the 1930s and Medicare also seems to generate the implication is that health plan quality 1940s were most likely to adopt the right incentives vis-à-vis the health measurement and reporting activity use of highly effective and low cost related habits. may drive changes in health care deliv- beta-blockers in the 1990s. A number of studies suggest that ery, improving the quality of care Khwaja uses microdata on a longi- collecting and publicly reporting quali- throughout a market. This suggests tudinal sample of individuals from the ty data about health plans is of limited that collection of quality data can have Health and Retirement Study and con- value because consumers make little beneficial effects on health care even if ducts a computer- simulated thought use of the information in their health consumers do not actively use health experiment comparing the trends and plan choices. However, this does not plan quality information when making outcomes under the status quo to a address another possible beneficial health plan choices. However, some world without Medicare. Further, effect of collecting quality data: health measures of quality may be more Khwaja analyzes the effect of care providers may respond to the effective than others in stimulating

46. NBER Reporter Summer 2003 change in provider behavior. degree of physician network overlap Medicare claims linked to the National Beaulieu develops a model and for health plans competing in the same Long-Term Care Survey (NLTCS) to presents evidence of a spillover, or market. She confirms that MCOs with assess the productivity of routine eye free-riding externality, that may arise broader provider networks are less like- examinations. Although such exams when managed care organizations ly to invest in quality improvement are widely recommended by profes- (MCOs) have overlapping provider activities that might spillover to other sional organizations for certain popu- networks and compete on non-price MCO plans. Consistent with the find- lations, there is limited empirical evi- characteristics. With overlapping ing on health plan investment, a second dence on the productivity of these provider networks, quality improve- analysis shows that MCOs with tighter preventive services. The authors meas- ments initiated by one MCO may spill provider networks, and MCOs that ure two outcomes, the ability to con- over and improve the quality of other operate in markets with lower degrees tinue reading, and blindness or low MCOs sharing the same provider net- in network overlap, have higher per- vision. They find a statistically signifi- work. Since the quality of all products formance on HEDIS quality measures. cant and beneficial effect of routine potentially improves, less differentia- One implication of the model is that eye exams for both outcomes. The tion is achieved by the MCO making the degree of exclusivity of a MCOs marginal effects for reading ability are the initial investment. This free-riding provider network is a strategic organi- large but declining in the number of externality reduces the return on zational choice that interacts with the years with annual visits. Effects for investment in quality improvement by MCOs’ product choice. blindness are smaller for the general increasing the costs of differentiation. Picone, Brown, and Sloan use a population, but larger for diabetics. Beaulieu presents evidence on the longitudinal national sample of

Market Microstructure The NBER’s Working Group on Joel Hasbrouck, New York University of Notre Dame Market Microstructure met in University, “Trading Costs and Cambridge on May 16. The meeting Returns for U.S. Equities: The Ekkehart Boehmer, New York was organized by Bruce Lehmann, Evidence from Daily Data” Stock Exchange, and Gideon Saar NBER and University of California, Discussant: Kenneth Kavajecz, and Lei Yu, New York University, San Diego; Andrew Lo, NBER and University of Pennsylvania “Lifting the Veil: An Analysis of MIT; Matthew Spiegel, Yale University; Pre-Trade Transparency at the and Avanidhar Subrahmanyam, Ronnie Sadka, Northwestern NYSE” University of California, Los Angeles. University, “Momentum, Liquidity Discussant: Chester Spatt, Carnegie These papers were discussed: Risk, and Limits to Arbitrage” Mellon University Discussant: Jonathan Lewellen, Ronald L. Goettler, Christine A. NBER and MIT Ruediger Fahlenbrach and Patrik Parlour, and Uday Rajan, Sandas, University of Pennsylvania, Carnegie Mellon University, Ingrid M. Werner, Ohio State “Bid-Ask Spreads and Inventory “Welfare in a Dynamic Limit Order University, “Institutional Trading Risk: Evidence from the FTSE 100 Market” Costs on Nasdaq: Have They Been Index Options Market” Discussant: Michael Goldstein, Decimated?” Discussant: Amber Anand, Syracuse Babson College Discussant: Robert Battalio, University

Goettler, Parlour, and Rajan pro- geneity of order flow, the midpoint is Then he compares the daily-based esti- vide an algorithm for solving for equi- not always a good proxy for the true mates with corresponding estimates librium in a dynamic limit-order mar- price. The authors also find that the based on high-frequency TAQ data. ket. The authors formulate a limit- effective spread is negatively correlated The best daily-based estimate of effec- order market as a stochastic sequential with transactions costs and uncorrelat- tive cost is the Gibbs sampler estimate game and use a simulation technique ed with welfare. They explicitly deter- of the Roll model. The correlation based on Pakes and McGuire (2001) to mine investor welfare in their numeri- between the best daily-based estimate find a stationary equilibrium. Given cal solution. As one policy experiment, and the TAQ value is about 0.90 for the stationary equilibrium, they gener- they evaluate the effect of changing individual securities and about 0.98 for ate artificial time series and perform tick size. portfolios. Daily-based proxies for comparative dynamics. They then can Hasbrouck examines various price effect costs are more problemat- compare transaction prices to true or approaches for estimating effective ic, though. Among the proxies efficient prices. Because of the endo- costs and price effects using daily data. Hasbrouck considers, the illiquidity

NBER Reporter Summer 2003 47. measure (Amihud (2000)) appears to cost savings for one week’s worth of introduction of OpenBook, consistent be the best: its correlation with the institutional orders. The results are with the view that improvement in liq- TAQ-based price impact measure is robust to controlling for order, stock, uidity affects stock returns. 0.47 for individual stocks and 0.90 for trading, and dealer characteristics. Fahlenbrach and Sandas study portfolios. Hasbrouck then extends Moreover, fill rates have improved and the determinants of the bid-ask the Gibbs estimate to the full sample duration has not increased significant- spreads for index options using a sam- covered by the daily CRSP database ly. Werner also documents differences ple that consists of all trades and (beginning in 1962). The CRSP esti- in trading costs by order type: effective quotes for the European style options mates exhibit considerable cross-sec- half-spreads for market orders are now and the futures on the FTSE 100 stock tional variation — consistent with the 52 basis points and for marketable index from August 2001 to July 2002. presumption that trading costs vary limit orders 42 basis points, while limit They compute two measures that widely across firms — but only mod- orders enjoy a 46 basis point spread show that bid-ask spreads in their sam- est time-series variation. In specifica- gain on average. ple are economically large. First, they tions using Fama-French factors, the Boehmer, Saar, and Yu investigate show that the reservation spreads for a Gibbs effective cost estimates are an important feature of market design: liquidity provider who hedges the delta found to be positive determinants of pre-trade transparency, defined as the risk of a bought or sold index option expected returns. availability of information about pend- by trading index futures is on average Sadka demonstrates the impor- ing trading interest in the market. They only 47 percent of the observed tance of liquidity for asset pricing, espe- look at the way the NYSE’s introduc- spread. Then they show that, on aver- cially for understanding the momentum tion of OpenBook, which enables age, the spread for a synthetic index anomaly. First, systematic liquidity risk, traders off the exchange floor to futures contract is 8-14 times the rather than the absolute level of liquid- observe depth in the limit-order book spread for the actual index futures ity, is shown to be important in explain- in real time, affects the trading strate- contract, even after adjusting for dif- ing the cross-sectional variation of gies of investors and specialists, and ferences in trading volume and con- expected returns. Moreover, momen- influences informational efficiency, liq- tract size. The authors estimate fixed tum returns can be attributed partially uidity, and returns. The authors find effects regressions for a panel of to compensation for liquidity risk. that traders attempt to manage the options and show that 42 percent of Second, seemingly profitable momen- exposure of their limit orders: the can- the variation in the daily average tum strategies that earn superior risk- cellation rate increases, time-to-cancel- spreads is driven by order processing adjusted returns (in absolute value) in lation shortens, and smaller orders are and inventory costs. The results show fact are associated with low levels of submitted. The new information that that inventory risk is an important liquidity. Therefore, the liquidity level of OpenBook provides seems to cause determinant of spreads for index momentum portfolios suggests possi- traders to prefer to manage the trading options, but that the compensation for ble limits to arbitrage. process themselves, rather than to del- inventory risk together with order pro- Werner uses unique order data to egate the task to floor brokers. The cessing costs does not explain all sys- compare trading costs for institutional authors also show that specialists’ par- tematic variation in the observed orders routed to Nasdaq sell-side deal- ticipation rate in trading declines, and spreads. The authors reject the null ers pre- and post-decimals. Her results the depth they add to the quote is that all fixed effects are equal to zero show that average execution-value- reduced, consistent with a loss of their for both calls and puts. The fixed weighted trading costs have declined information advantage or with being effects vary systematically with option by 20 basis points compared to open “crowded out” by active limit-order characteristics; for example, out-of- mid-quotes, by 16 basis points com- strategies. There is an improvement in the-money options have wider spreads pared to the quotes at order arrival the informational efficiency of prices and in-the-money options have nar- (effective half-spreads), and by 1 basis after the introduction of OpenBook. rower spreads than predicted by these point compared to mid-quotes at the Greater pre-trade transparency leads regressions. One interpretation is that close (realized half-spreads). The to some improvement in displayed liq- the spreads include a mark-up above reduction in effective half-spreads rep- uidity in the book and a reduction in the marginal costs of providing liquid- resents a 22 percent reduction in aver- the execution costs of trades. The ity for at least some of the options in age institutional trading costs, which authors find that cumulative abnormal this sample. corresponds to $69 million in total returns are positive following the

48. NBER Reporter Summer 2003 Bureau Books

The following volumes may be ordered directly from the University of Chicago Press, Order Department, 11030 South Langley Avenue, Chicago, IL 60628-2215; 1-800-621-2736. Academic discounts of 10 percent for individual volumes and 20 percent for standing orders for all NBER books published by the University of Chicago Press are available to universi- ty faculty; orders must be sent on university stationery.

The Governance of Not-for-Profit Organizations

The Governance of Not-for-Profit it institutions, including art museums, operating not-for-profit firms and Organizations, edited by Edward L. hospitals, and Renaissance chapels. It underline the key distinctions between Glaeser, is available from the also discusses the theoretical under- for-profits and not-for-profits. University of Chicago Press for pinnings of governance issues and for- Glaeser is a research associate of $55.00. This NBER conference vol- profit behavior. Taken together, these the National Bureau of Economic ume describes governance issues in papers advance the empirical knowl- Research and a professor of econom- several different types of not-for-prof- edge of the dynamics involved in ics at Harvard University.

Means-Tested Transfer Programs in the United States

Means-Tested Transfer Programs in the Assistance to Needy Families (TANF). history, current rules and caseloads, United States, edited by Robert A. At the same time, the Earned Income expenditures, and recipient characteris- Moffitt, is available from the University Tax Credit (EITC) has grown from a tics; it also summarizes recent research. of Chicago Press for $95.00. In just the minor program to one of the most The result is a “one-stop source” for last ten years, we have seen means-test- important for low-income families, and information and analyses that have not ed transfer programs — programs for Medicaid has greatly expanded its eligi- been put together in a single accessible which some test of resources is bility to new groups. This NBER volume before. required for recipients — such as Aid Conference Report covers all nine Moffitt is an NBER Research to Families with Dependent Children major U.S. means-tested programs, dis- Associate and a professor of econom- (AFDC), replaced by Temporary cussing each program’s institutional ics at Johns Hopkins University.

Labor Markets and Firm Benefit Policies in Japan and the United States

Labor Markets and Firm Benefit assessment of the recent rise in the data that will allow such comparisons Policies in Japan and the United States, edit- U.S. economy and the prospects for in the future. ed by Seiritsu Ogura, Toshiaki remaining at the “peak.” Following an Ogura is associated with the Japan Tachibanaki, and David A. Wise, is extensive exploration of labor markets Center for Economic Research. available from the University of in Japan, chapters six through eight Tachibanaki is a professor at the Chicago Press for $99.00. This is the deal with health and pension issues in Institute of Economic Research of fourth volume resulting from a collab- the United States. The final four chap- Kyoto University. Wise directs the oration between the NBER and the ters examine benefit policies in Japan. NBER’s Program on Aging and is a Japan Center for Economic Research As with all volumes in this series, the professor of economics at Harvard’s (JCER); its thirteen papers feature papers collected here are primarily John F. Kennedy School of analyses of important features of investigations of discrete topics rather Government. Together the three also employment practices and fringe ben- than comparisons between the two coedited the most recent JCER-NBER efit systems in Japan and the United countries —studies undertaken in part collaboration, Aging Issues in the United States. The book opens with a general to produce a critical mass of empirical States and Japan, published in 2001.

NBER Reporter Summer 2003 49. Current Working Papers

NBER Working Papers On-Line A complete list of all NBER Working Papers with searchable abstracts, and the full texts of Working Papers (issued since November 1994) are available at http://www.nber.org/wwp.html to anyone located at a university or other organization that sub- scribes to the (hard copy) Working Paper series. If you believe that your organization subscribes, but you cannot access the online Working Paper service, please e-mail the NBER at [email protected] for more information and assistance. * Individual copies of NBER Working Papers, Historical Factors in Long-Run Growth Papers, and Technical Papers are avail- able free of charge to Corporate Associates. For all others, there is a charge of $10.00 per hardcopy or $5.00 per downloaded paper. (Outside the United States, add $10.00 per order for postage and handling.) Advance payment is required on all orders. To order, call the Publications Department at (617)868-3900 or visit www.nber.org/papers. Please have ready the num- ber(s) of any Working Paper(s) you wish to order. Subscriptions to the full NBER Working Paper series include all 500 or more papers published each year. Subscriptions are free to Corporate Associates. For others within the United States, the standard rate for a full subscription is $2200; for academic libraries and faculty members, $1275. Higher rates apply for foreign orders. The on-line standard rate for a full subscription is $1560 and the on-line academic rate is $630. Partial Working Paper subscriptions, delineated by program, are also available. For further information, see our Web site, or please write: National Bureau of Economic Research, 1050 Massachusetts Avenue, Cambridge, MA 02138-5398. * Titles of all papers issued since April 2003 are presented below. For previous papers, see past issues of the NBER Reporter. Working Papers are intended to make results of NBER research available to other economists in preliminary form to encourage dis- cussion and suggestions for revision before final publication. They are not reviewed by the Board of Directors of the NBER.

NBER Working Papers

Paper Author(s) Title 9623 Willem H. Buiter Deflation: Prevention and Cure 9624 Charles W. Calomiris How to Restructure Failed Banking Systems: Lessons from the U.S. Joseph R. Mason in the 1930s and Japan in the 1990s 9625 Bruce A. Blonigen Evolving Discretionary Practices of U.S. Antidumping Activity 9626 Mark G. Duggan Does Medicaid Pay Too Much for Prescription Drugs? A Case Study of Atypical Anti-Psychotics 9627 Ronald G. Ehrenberg Who Bears the Growing Cost of Science at Universities? Michael J. Rizzo George H. Jakubson 9628 Robert A. Pollak Does Family Structure Affect Children’s Educational Outcomes? Donna K. Ginther 9629 Christian Gollier Collective Investment Decision Making with Heterogeneous Time Richard Zeckhauser Preferences 9630 Pinka Chatterji Does Returning to Work After Childbirth Affect Breastfeeding Kevin Frick Practices? 9631 Pravin Krishna Reciprocated Unilateralism: An Interest-Group Approach Devashish Mitra 9632 Jennifer Hunt Teen Births Keep American Crime High

50. NBER Reporter Summer 2003 Paper Author(s) Title 9633 Raghuram Rajan The Flattening Firm: Evidence from Panel Data on the Changing Nature Julie Wulf of Corporate Hierarchies

9634 Edward J. Kane Continuing Dangers of Disinformation in Corporate Accounting Reports

9635 Herschel I. Grossman “Fifty-four Forty or Fight!”

9636 Severin Borenstein Do Airline Bankruptcies Reduce Air Service? Nancy L. Rose

9637 Stephen Redding Geography and Export Performance: External Market Access Anthony J. Venables and Internal Supply Capacity

9638 Pinka Chatterji Alcohol Abuse and Suicide Attempts Among Youth — Dhaval Dave Correlation or Causation? Robert Kaestner Sara Markowitz

9639 Andrew B. Bernard Falling Trade Costs, Heterogeneous Firms, and Industry Dynamics J. Bradford Jensen Peter K. Schott

9640 Roberto Rigobon Spillovers Across U.S. Financial Markets Brian Sack

9641 Enrico Moretti Human Capital Externalities in Cities

9642 Yasushi Hamao Idiosyncratic Risk and the Creative Destruction in Japan Jianping Mei Yexiao Xu

9643 Joe Peek Unnatural Selection: Perverse Incentives and the Misallocation of Eric S. Rosengren Credit in Japan

9644 Patrick M. McGuire Bank Ties and Bond Market Access: Evidence on Investment-Cash Flow Sensitivity in Japan

9645 Colin Mayer Sources of Funds and Investment Activities of Venture Capital Funds: Koen Schoors Evidence from Germany, Israel, Japan, and the UK Yishay Yafeh

9646 Kazuo Ogawa Financial Distress and Employment: The Japanese Case in the 1990s

9647 Tokuo Iwaisako Household Portfolios in Japan

9648 Rasmus Fatum Effectiveness of Official Daily Foreign Exchange Market Intervention Michael Hutchison Operations in Japan

9649 Sean Nicholson Barriers to Entering Medical Specialties

9650 Reuben Gronau Time Vs. Goods: The Value of Measuring Household Production Technologies Daniel S. Hamermesh

NBER Reporter Summer 2003 51. Paper Author(s) Title 9651 Randall Morck Why Some Double Taxation Might Make Sense: The Special Case of Inter-corporate Dividends 9652 Brian A. Jacob Public Housing, Housing Vouchers,and Student Achievement: Evidence from Public Housing Demolitions in Chicago 9653 Brian A. Jacob Are Idle Hands the Devil’s Workshop? Lars Lefgren Incapacitation, Concentration, and Juvenile Crime 9654 Michael D. Bordo Exchange Rate Regime Choice in Historical Perspective 9655 Simon J. Evenett The Cross Border Mergers and Acquisitions Wave of the Late 1990s 9656 Jeffrey G. Williamson Was It Stolper-Samuelson, Infant Industry, or Something Else? World Trade Tariffs, 1789-1938 9657 Alon Brav Payout Policy in the 21st Century John R. Graham Campbell R. Harvey Roni Michaely 9658 V.V. Chari Financial Crises as Herds: Overturning the Critiques Patrick J. Kehoe 9659 Raymond Fisman Does Competition Encourage Credit Provision? Mayank Raturi Evidence from African Trade Credit Relationships 9660 Jon Faust Identifying the Effects of Monetary Policy Shocks on Exchange Rates Using John H. Rogers High Frequency Data Eric Swanson Jonathan H. Wright 9661 Jon Bakija Charitable Bequests and Taxes on Inheritance and Estates: Aggregate William Gale Evidence from Across States and Time Joel Slemrod 9662 Ashoka Mody The Role of Information in Driving FDI Flows: Host-Country Transparency Assaf Razin and Source Country Specialization Efraim Sadka 9663 Simon J. Evenett Do All Networks Facilitate International Commerce? U.S. Law Firms and the International Market for Corporate Control 9664 Michael W. Brandt A No-Arbitrage Approach to Range-Based Estimation of Return Francis X. Diebold Covariance and Correlations 9665 David Blau The Determinants and Consequences of Child Care Subsidies for Erdal Tekin Single Mothers 9666 Barry Eichengreen The Monetary Consequences of a Free Trade Area of the Americas Alan M. Taylor 9667 Jiri Jones Inflation Targeting in Transition Countries: Experience and Prospects Frederic S. Mishkin 9668 Christopher L. Gilbert Globalization and International Commodity Trade with Specific Reference Panos Varangis to the West African Cocoa Producers 9669 Drusilla K. Brown The Effects of Multinational Production on Wages and Working Conditions Alan V. Deardorff in Developing Countries Robert M. Stern 9670 Carmen M. Reinhart Twin Fallacies About Exchange Rate Policy in Emerging Markets Vincent R. Reinhart 9671 Paul Glewwe Teacher Incentives Nauman Ilias Michael Kremer 52. NBER Reporter Summer 2003 Paper Author(s) Title 9672 Stephen G. Cecchetti Inflation Targeting, Price-Path Targeting, and Output Variability Junhan Kim 9673 David S. Bates Maximum Likelihood Estimation of Latent Affine Processes 9674 Richard J. Rendleman, Jr. Diversification and the Taxation of Capital Gains and Losses Douglas A. Shackelford 9675 Steven Berry Product Quality and Market Size Joel Waldfogel 9676 Henry Saffer Alcohol Advertising and Alcohol Consumption by Adolescents Dhaval Dave 9677 Andrew Ang Do Demographic Changes Affect Risk Premiums? Angela Maddaloni Evidence from International Data 9678 Joshua Graff Zavin Risk Aversion, Liability Rules, and Safety Richard Just David Zilberman 9679 Edward P. Lazear Firm-Specific Human Capital: A Skill-Weights Approach 9680 Josh Lerner Cooperative Marketing Agreements Between Competitors: Jean Tirole Evidence from Patent Pools Marcin Strojwas 9681 Angela K. Dills Alcohol Prohibition and Cirrhosis Jeffrey A. Miron 9682 Robert J. Barro Religion and Economic Growth Rachel M. McCleary 9683 George M. Holmes Does School Choice Increase School Quality? Jeffrey S. DeSimone Nicholas G. Rupp 9684 Pierre Olivier Gourinchas The Elusive Gains from International Financial Integration Olivier Jeanne 9685 Scott Weisbenner Local Does as Local Is: Information Content of the Geography of Individual Zoran Ivkovich Investors’ Common Stock Investments 9686 Andrew F. Haughwout Local Revenue Hills: Evidence from Four U.S. Cities Robert P. Inman Steven Craig Thomas Luce 9687 Luis Bertola Globalization in Latin America Before 1940 Jeffrey Williamson 9688 Kosali Ilayperuma Simon Do Minimum Wages Affect Non-wage Job Attributes? Evidence on Robert Kaestner Fringe Benefits and Working Conditions 9689 Jeffrey A. Miron The Effect of Drug Prohibition on Drug Prices: Evidence from the Markets for Cocaine and Heroin 9690 Rosalie Liccardo Pacula Marijuana Decriminalization: What Does it Mean for the United States? Jamie F. Chriqui Joanna King 9691 Casey B. Mulligan Social Security, Retirement, and the Single-Mindedness of the Electorate Xavier Sala-i-Martin 9692 Alejandro Ibarra-Yunez Spaghetti Regionalism or Strategic Foreign Trade: Some Evidence for Mexico

NBER Reporter Summer 2003 53. Paper Author(s) Title 9693 Ann Dryden Witte Impacts of the Eligibility Expansions and Provider Reimbursement Rate Magaly Queralt Increases on Child Care Subsidy Take-Up Rates,Welfare Use, and Work 9694 Steven Shavell Economic Analysis of Accident Law 9695 Steven Shavell Economic Analysis of Property Law 9696 Steven Shavell Economic Analysis of Contract Law 9697 Steven Shavell Economic Analysis of Litigation and the Legal Process 9698 Steven Shavell Economic Analysis of Public Law Enforcement and Criminal Law 9699 Steven Shavell Economic Analysis of the General Structure of the Law 9700 Steven Shavell Economic Analysis of Welfare Economics, Morality, and the Law 9701 Chang-Tai Hsieh Relative Prices and Relative Prosperity Peter J. Klenow 9702 Scott Adams Living Wage Effects: New and Improved Evidence David Neumark 9703 Thomas J. Kane A Quasi-Experimental Estimate of the Impact of Financial Aid on College-Going 9704 Donald Morgan Bank Integration and State Business Cycles Bertrand Rime Philip Strahan 9705 Henry Farber Nonunion Wage Rates and the Threat of Unionization 9706 Henry Farber Is Tomorrow Another Day? The Labor Supply of New York Cab Drivers 9707 Henry Farber Job Loss in the United States, 1981-2001 9708 Joshua D. Angrist Treatment Effect Heterogeneity in Theory and Practice 9709 Steve Dowrick Ideas and Education: Level or Growth Effects? 9710 Donald Morgan Foreign Bank Entry and Business Volatility: Philip Strahan Evidence from U.S. States and Other Countries 9711 Harrison Hong Thy Neighbor’s Portfolio: Word-of-Mouth Effects in the Holdings and Trades Jeffrey D. Kubik of Money Managers Jeremy C. Stein 9712 Jaap H. Abbring A Structural Empirical Model of Firm Growth, Learning, and Survival Jeffrey R. Campbell 9713 Michael D. Bordo IS-LM and Monetarism Anna J. Schwartz 9714 Thomas Davidoff Annuities and Individual Welfare Jeffrey R. Brown Peter A. Diamond 9715 Mihir Desai A Multinational Perspective on Capital Structure Choice and Internal C. Fritz Foley Capital Markets James R. Hines, Jr. 9716 Vanessa Strauss-Kahn The Role of Globalization in the Within-Industry Shift Away from Unskilled Workers in France 9717 Bronwyn H. Hall Business Method Patents, Innovation, and Policy 9718 Josh Ederington Footloose and Pollution-Free Arik Levinson Jenny Minier 9719 Luis Garicano Specialization, Firms, and Markets: Thomas N. Hubbard The Division of Labor Within and Between Law Firms

54. NBER Reporter Summer 2003 Paper Author(s) Title 9720 Bridget Terry Long Does the Format of a Financial Aid Program Matter? The Effect of State In-Kind Tuition Subsidies 9721 Glenn Ellison A Model of Add-on Pricing 9722 Patrick Bolton Executive Compensation and Short-termist Behavior in Speculative Jose Scheinkman Markets Wei Xiong 9723 Gordon H. Hanson Vertical Production Networks in Multinational Firms Raymond J. Mataloni Matthew J. Slaughter 9724 Jiann-Chyuan Wang Productivity Growth and R&D Expenditure in Taiwan’s Manufacturing Kuen-Hung Tsai Firms 9725 Manuel Trajtenberg Defense R&D Policy in the Anti-terrorist Era 9726 Martin Pesendorfer Identification and Estimation of Dynamic Games Philipp Schmidt-Dengler 9727 Saul Lach Incentives and Invention in Universities Mark Schankerman 9728 Ali Hortacsu Product Differentiation, Search Costs, and Competition in the Mutual Fund Chad Syverson Industry: A Case Study of S&P 500 Index Funds 9729 Carolyn L. Evans Distance, Time, and Specialization James Harrigan 9730 Bronwyn Hall Adoption of New Technology Beethika Khan 9731 Bronwyn Hall Prospects for Improving U.S. Patent Qualityvia Post-grant Opposition Stuart J. H. Graham Dietmar Harhoff David C. Mowery 9732 James J. Heckman Fifty Years of Mincer Earnings Regressions Lance J. Lochner Petra E. Todd 9733 Edward L. Glaeser Sprawl and Urban Growth Mathhew E. Kahn 9734 Richard A. Jensen The Disclosure and Licensing of University Inventions Jerry G. Thursby Marie C. Thursby 9735 Emmanuel Dechenaux Appropriability and the Timing of Innovation: Brent Goldfarb Evidence from MIT Inventions Scott A. Shane Marie C. Thursby 9736 John A. List Neoclassical Theory Versus Prospect Theory: Evidence From the Marketplace 9737 Aaron Tornell Credit Market Imperfections in Middle Income Countries Frank Westermann 9738 Sherry Glied Health Inequality, Education, and Medical Innovation Adriana Lleras-Muney 9739 Paul R. Bergin Endogenous Nontradability and Macroeconomic Implications Reuven Glick 9740 Pol Antràs Firms, Contracts, and Trade Structure

NBER Reporter Summer 2003 55. Paper Author(s) Title 9741 Michael C. Davis Why Are Prices Sticky? The Dynamics of Wholesale Gasoline Prices James D. Hamilton 9742 Hugh Rockoff Prodigals and Projectors: An Economic History of Usury Laws in the United States from Colonial Times to 1900 9743 Kent Daniel Market Reactions to Tangible and Intangible Information Sheridan Titman 9744 Nicholas G. Rupp Airline Schedule Recovery After Airport Closures: George M. Holmes Empirical Evidence Since September 11th Jeff DeSimone 9745 Audrey Light Why Parents Play Favorites: Explanations of Unequal Bequests Kathleen McGarry 9746 Enrique G. Mendoza A Quantitative Analysis of Tax Competition vs. Tax Coordination Linda L. Tesar under Perfect Capital Mobility 9747 Lars E. O. Svensson Implementing Optimal Policy through Inflation-Forecast Targeting Michael Woodford 9748 Brian Knight Parochial Interests and the Centralized Provision of Local Public Goods: Evidence from Congressional Voting on Transportation Projects 9749 Peter H. Lindert Voice and Growth: Was Churchill Right? 9750 Frank R. Lichtenberg Does Misery Love Company? Evidence from Pharmaceutical Markets Joel Waldfogel Before and After the Orphan Drug Act 9751 Robert A. Moffitt The Negative Income Tax and the Evolution of U.S.Welfare Policy 9752 Howard Bodenhorn Just and Reasonable Treatment: Racial Treatment in the Terms of Pauper Apprenticeship in Antebellum Maryland 9753 Willem H.Buiter James Tobin: An Appreciation of his Contribution to Economics 9754 Frank R. Lichtenberg The Impact of New Drug Launches on Longevity: Evidence from Longitudinal, Disease-Level Data from 52 Countries, 1982-2001 9755 George J. Borjas The Labor Demand Curve is Downward Sloping: Reexamining the Impact of Immigration on the Labor Market 9756 Simeon Djankov The Regulation of Labor Rafael La Porta Florencio Lopez-de-Silanes Andrei Shleifer Juan Botero 9757 Patrick Bajari Auctions Versus Negotiations in Procurement: Robert S. McMillan An Empirical Analysis Steve Tadelis 9758 James Dow Equilibrium Asset Prices Under Imperfect Corporate Control Gary Gorton Arvind Krishnamurthy 9759 John Y. Campbell Household Risk Management and Optimal Mortgage Choice Joao F. Cocco 9760 Bruce D. Meyer Measuring the Well-Being of the Poor Using Income and Consumption James X. Sullivan 9761 Shane Greenstein Differentiation Strategy and Market Deregulation: Michael Mazzeo Local Telecommunication Entry in the Late 1990s

56. NBER Reporter Summer 2003 Paper Author(s) Title 9762 David Popp ENTICE: Endogenous Technological Change in the DICE Model of Global Warming 9763 Walter Novaes Bureaucracy as a Mechanism to Generate Information Luigi Zingales 9764 George-Marios Angeletos Idiosyncratic Production Risk, Growth, and the Business Cycle 9765 Gary S. Becker The Quantity and Quality of Life and the Evolution of World Inequality Tomas J. Philipson Rodrigo R. Soares 9766 Alejandra Cox Edwards International Migration, Remittances, and Schooling: Evidence from Manuelita Ureta El Salvador 9767 George-Marios Angeletos Coordination and Policy Traps Christian Hellwig Alessandro Pavan 9768 Maury Gittleman The Vintage Effect in TFP Growth: Thijs ten Raa An Analysis of the Age Structure of Capital Edward N. Wolff 9769 Robert Kaestner The Effect of Welfare Reform on Prenatal Care and Birth Weight Won Chan Lee 9770 William J. Collins Historical Perspectives on Racial Differences in Schooling in the Robert A. Margo United States 9771 Robert W. Fogel Secular Trends in Physiological Capital: Implications for Equity in Health Care 9772 Craig Burnside Fiscal Shocks and their Consequences Martin Eichenbaum Jonas Fisher 9773 Jordi Gali Fiscal Policy and Monetary Integration in Europe Roberto Perotti 9774 Avi Dor The Effects of Private Insurance on Measures of Health: Joseph Sudano Evidence from the Health and Retirement Study David W. Baker 9775 Rafael LaPorta Judicial Checks and Balances Florencio Lopez-de-Silanes Christian Pop-Eleches Andrei Shleifer 9776 Jean-Marie Grether Globalization and Dirty Industries: Jaime de Melo Do Pollution Havens Matter? 9777 Kristin J. Forbes One Cost of the Chilean Capital Controls: Increased Financial Constraints for Smaller Traded Firms 9778 Ricardo J. Caballero Hedging Sudden Stops and Precautionary Recessions: A Quantitative Framework 9779 Steven G. Allen Phasing into Retirement Robert L. Clark Linda S. Ghent 9780 Lynne G. Zucker Measuring Success of Advanced Technology Program Participation Michael R. Darby Using Archival Data 9781 George J. Borjas Welfare Reform, Labor Supply, and Health Insurance in the Immigrant Population

NBER Reporter Summer 2003 57. Paper Author(s) Title 9782 Michael Baker The Married Widow: Marriage Penalties Matter! Emily Hanna Jasmin Kantarevic 9783 Patrick Bajari House Prices and Consumer Welfare C. Lanier Benkard John Krainer 9784 Brian J. Hall The Trouble with Stock Options Kevin J. Murphy 9785 Jan Boone The Optimal Taxation of Unskilled Labor with Job Search and Social Lans Bovenberg Assistance 9786 Craig Burnside Government Finance in the Wake of Currency Crises Martin Eichenbaum Sergio Rebelo 9787 Graciela L. Kaminsky Short-Run Pain, Long-Run Gain: Sergio Schmukler The Effects of Financial Liberalization 9788 Anne Case From Cradle to Grave? The Lasting Impact of Childhood Angela Fertig Health and Circumstance Christina Paxson 9789 Andrew B. Bernard Product Choice and Product Switching Peter K. Schott Stephen Redding 9790 John A. Tauras The Impact of Physician Intervention and Tobacco Control Policies Lan Liang 9791 Sergey Iskoz How to Tell if a Money Manager Knows More Jiang Wang 9792 Christopher M. Meissner Committee Structure and the Success of Connected Lending in Nineteenth Century New England Banks 9793 Herschel I. Grossman ". . . and six hundred thousand men were dead." 9794 Herschel I. Grossman Distributional Disputes and Civil Conflict 9795 Alexandre Ziegler The Dominance of Retail Stores Edward P. Lazear 9796 N. Gregory Mankiw Disagreement about Inflation Expectation Ricardo Reis Justin Wolfers 9797 Kenneth Leonard Outcome Versus Service Based Payments in Health Care: Joshua Graff Zivin Lessons from African Traditional Healers 9798 Jeff Dominitz “Will Social Security Be There for You?”: Jordan Heinz How Americans Perceive Their Benefits Charles F. Manski 9799 Dennis Epple The Practice and Proscription of Affirmative Action in Higher Education Richard Romano Holger Sieg 9800 William F. Maloney Measuring the Impact of Minimum Wages: Jairo Nunez Mendez Evidence from Latin America

58. NBER Reporter Summer 2003 Paper Author(s) Title 9801 Per G. Fredriksson Chasing the Smokestack: John A. List Strategic Policymaking With Multiple Instruments Daniel L. Millimet 9802 Daron Acemoglu Incentives in Markets, Firms, and Governments Michael Kremer Atif Mian 9803 John C. Ham Did Expanding Medicaid Affect Welfare Participation? Lara D. Shore-Sheppard 9804 Xavier Sala-i-Martin Addressing the Natural Resource Curse: Arvind Subramanian An Illustration from Nigeria 9805 Felix Oberholzer-Gee Social Learning and Coordination in High-Stakes Games: Joel Waldfogel Evidence from Friend or Foe Matthew White 9806 Mario Draghi Transparency, Risk Management, and International Francesco Giavazzi Financial Fragility Robert C. Merton 9807 Steven Kaplan Private Equity Performance: Antoinette Schoar Returns, Persistence, and Capital 9808 Guillermo Calvo The Mirage of Exchange Rate Regimes for Emerging Market Countries Frederic S. Mishkin 9809 Alberto Alesina Choosing (and Reneging on) Exchange Rate Regimes Alexander Wagner

NBER Reporter Summer 2003 59. Nonprofit Org. U.S. Postage NBERReporter PAID National Bureau of NATIONAL BUREAU OF ECONOMIC RESEARCH Economic Research

1050 Massachusetts Avenue Cambridge, Massachusetts 02138-5398 (617) 868-3900

Change Service Requested