NBER Reporter NATIONAL BUREAU OF ECONOMIC RESEARCH

A quarterly summary of NBER research No. 3, September 2017

The 2017 Lecture

Uncertainty and Large Swings in Activity

Mervyn King*

It is a great honor, as well as a personal pleasure and privilege, to be invited to deliver the Feldstein Lecture. I have known Marty and Kate for almost 50 years. I met Marty in the summer of 1970, when I presented my first ever paper at the Second World Congress of the Econometric Society in Cambridge, England. The subject was investment, and Marty presented a paper, jointly with the late John Flemming, on the same topic. Those were the early days of computer anal- ysis of data, and paper tape had not yet given way to the new technology of punched cards. But the application of rigorous theory to quantitative empirical analysis was a heady and seductive combination. Mervyn King A year later I was a graduate student at Harvard with Marty as my men- tor. A few years after that, Marty took over the National Bureau, and the first Summer Institute was held. “Oh, to be in Cambridge, England now that spring ALSO IN THIS ISSUE is here” became “Oh, to be in Cambridge, Massachusetts now that summer’s here”.1 Old Idea, New Insights: The Ricardian Revival And here we are at the 40th NBER Summer Institute. In the audience, I in International Trade 11 see economists who had not yet been born at the time of that first workshop in 1978. So in my lecture I want to trace the path that both Marty and I took from Integrated Assessment Models microeconomics to macroeconomics. In particular, I shall ask how far the so- of Climate Change 16 called workhorse or canonical models of modern macroeconomics can help us Resource Barriers to Postsecondary understand what has been going on in the world economy for the past quarter Educational Attainment 21 of a century. My focus will be on uncertainty and large swings in economic activ- Economic Development and ity — of the kind we saw in the Great Depression and more recently in the Great Gender Equality: Exceptions to the Rule 26 Recession of 2008–09 — and the unexpectedly slow and protracted recovery NBER News 29 since the financial crisis. Conferences 33 In so doing, I want to draw inspiration from what, in my view, is one Program Meeting 38 *This is an annotated, lightly edited version of the Martin Feldstein NBER Books 39 Lecture delivered July 19, 2017, at the NBER Summer Institute. Mervyn King is a former governor of the Bank of England, a member of the British House of Lords, an NBER research associate, and the Alan Greenspan Professor of Economics at New York University. The annual lecture is named in honor of the NBER’s president emeritus.

Reporter Online at: www.nber.org/reporter of Marty’s greatest strengths: his ability to com- bine a conviction that economics has a great deal NBER Reporter to offer in thinking about almost every aspect of our lives — Marty’s freshwater characteristic — and an imagination to develop models and new data sources to examine previously unexplored terri- The National Bureau of Economic Research is a private, nonprofit research orga- nization founded in 1920 and devoted to objective quantitative analysis of the tory — his saltwater dimension. American economy. Its officers and board of directors are: President and Chief Executive Officer — James M. Poterba Introduction Controller — Kelly Horak Corporate Secretary — Alterra Milone The fundamental question that has divided BOARD OF DIRECTORS economists since publication of The General Theory Chairman — Karen N. Horn in 1936 is whether a market economy with flex- Vice Chairman — John Lipsky ible wages and prices is self-stabilizing. The recent Treasurer — Robert Mednick financial crisis should have generated a more serious DIRECTORS AT LARGE debate about that question. But it takes a great deal Peter Aldrich Jacob A. Frenkel Michael H. Moskow to derail a conventional theory. As John Maynard Elizabeth E. Bailey Judith M. Gueron Alicia H. Munnell Keynes wrote in the preface to his great work, “The John H. Biggs Robert S. Hamada Robert T. Parry John S. Clarkeson Peter Blair Henry James M. Poterba ideas which are here expressed so laboriously are Kathleen B. Cooper Karen N. Horn John S. Reed extremely simple and should be obvious. The diffi- Charles H. Dallara Lisa Jordan Marina v. N. Whitman culty lies, not in the new ideas, but in escaping from George C. Eads John Lipsky Martin B. Zimmerman 2 Jessica P. Einhorn Laurence H. Meyer the old ones.” Mohamed El-Erian Karen Gordon Mills The crisis did not lead to an intellectual revo- lution. Instead, debate focused on the appropriate DIRECTORS BY UNIVERSITY APPOINTMENT policy response rather than the theoretical basis of Timothy Bresnahan, Stanford Benjamin Hermalin, California, Berkeley current macroeconomics.3 Indeed, the workhorse Pierre-André Chiappori, Columbia George Mailath, Pennsylvania Alan V. Deardorff, Michigan Marjorie B. McElroy, Duke model taught in courses on macroeconomics and Ray C. Fair, Yale Joel Mokyr, Northwestern used by policymakers survived the crisis better than Edward Foster, Minnesota Cecilia Elena Rouse, Princeton did our economies. Even adding banks and finan- John P. Gould, Chicago Richard L. Schmalensee, MIT Mark Grinblatt, California, Los Angeles Ingo Walter, New York University cial rigidities, with new first-order conditions, did Bruce Hansen, Wisconsin-Madison David B. Yoffie, Harvard not change its basic properties. The central idea is DIRECTORS BY APPOINTMENT OF OTHER ORGANIZATIONS that the economy moves in response to stochastic Jean-Paul Chavas, Agricultural and Applied Economics Association shocks around a steady-state or stationary long-run Martin Gruber, American Finance Association equilibrium. Philip Hoffman, Economic History Association It is interesting to ask how the stochastic, one- Arthur Kennickell, American Statistical Association Jack Kleinhenz, National Association for Business Economics sector models so much in favor today came to dom- Robert Mednick, American Institute of Certified Public Accountants inate macroeconomic thinking. Fifty years or so Peter L. Rousseau, American Economic Association ago, models of economic dynamics and models of Gregor W. Smith, Canadian Economics Association William Spriggs, American Federation of Labor and economic growth were quite separate. The former Congress of Industrial Organizations stimulated the construction of econometric mod- Bart van Ark, The Conference Board els with empirically estimated dynamic responses. The NBER depends on funding from individuals, corporations, and private The latter were concerned with long-run steady foundations to maintain its independence and its flexibility in choosing its growth and later expanded into multisector models research activities. Inquiries concerning contributions may be addressed to James of economic development.4 The first advance was M. Poterba, President & CEO, NBER, 1050 Massachusetts Avenue, Cambridge, MA 02138-5398. All contributions to the NBER are tax-deductible. to incorporate the ideas of Frank Ramsey into the formulation of optimal growth paths based on the The Reporter is issued for informational purposes and has not been reviewed by maximization of expected utility.5 The second was the Board of Directors of the NBER. It is not copyrighted and can be freely repro- the explicit modelling of expectations in a stochas- duced with appropriate attribution of source. Please provide the NBER’s Public Information Department with copies of anything reproduced. tic environment. It was natural to relate expecta- tions to the underlying long-run relationships driv- Requests for subscriptions, changes of address, and cancellations should be sent ing the economy, and so rational expectations came to Reporter, National Bureau of Economic Research, Inc., 1050 Massachusetts Avenue, Cambridge, MA 02138-5398 (please include the current mailing label), to the fore. Multisector models seemed to add little or by email to [email protected]. Print copies of the Reporter are only mailed to to the insights into behavior afforded by the ratio- subscribers in the U.S. and Canada; those in other nations may request electronic nal expectations revolution. Attention switched subscriptions at www.nber.org/drsubscribe/. 2 NBER Reporter • No. 3, September 2017 back, therefore, to one-sector models and the elaboration of stochas- GDP per Capita, U.S. and U.K., 1900–2016 tic shocks. And so we arrived at today’s consensus on the centrality of one-sector DSGE (dynamic stochastic general equilibrium) models. In 1990 dollars 40,000 But these models have their limitations, and two seem to me par- 1960–2016 trend ticularly serious. First, expected utility theory has come to dominate macroeconomic modelling even though its foundations are fragile when analyzing behavior in the presence of large, one-off macroeco- U.S. nomic shocks. Second, the one-sector framework leads policymakers 20,000 to focus exclusively on the level of aggregate demand rather than on its composition. Both features are, in my view, problematic in under- U.K. standing the world economy today, as I shall now try to illustrate with a rapid tour of some of the relevant data. 0 1900 1960 2016 Selected Data Source: Data from the Maddison Project and the IMF World Economic Outlook Database

The proposition that the U.S. economy follows a path described Figure 1 by random shocks around a steady-state growth rate is given some sup- port in Figure 1, which plots GDP per head at constant prices from 1900 to 2016.6 A trend line with a constant annual growth rate of 1.95 percent captures the upward path of GDP per head rather well. Distribution of Deviations from U.S. GDP Trend, 1900–2016 By far the largest deviations from this path were, of course, the Great Count of occurrences Depression and the boom experienced in the Second World War. It 20 is noticeable that, despite these large swings in activity, from 1950 onwards GDP per head resumed the path that would have been pro- 15 jected from an estimated trend over the period 1900 through 1930.

Figure 1 also shows data for the U.K. The underlying growth rate 10 is remarkably similar, although, unlike the U.S., the U.K. did not expe- rience the wild swings of the 1930s and 1940s. But at the end of the First World War the U.K. suffered a step down in the level of GDP per 5 head and did not return to the previous trend path. This was when the U.S. took over the mantle of the world’s financial leader. 0 ≤ -15 (-13,-11] (-9,-7] (-5,-3] (-1,1] (3,5] (7,9] (11,13] ≥ 15 Figure 2 plots the distribution of percentage deviations from Percent deviation trend GDP in the U.S. over the full 1900–2016 period. Whatever else Source: Calculations using data from the Maddison Project can be said, the chart does not look like a normal distribution. If the and the IMF World Economic Outlook Database underlying distribution of shocks is normal, then it must be shifting over time, suggesting non-stationarity of the shocks. Figure 2 For the period since 1960, Figure 1 shows the trend growth path for real GDP per capita for the U.S. and the U.K. over the 1960–2007 period — the period up to the beginning of the recent financial cri- Global Real Interest Rate, 1985–2017 sis. The growth rate is almost exactly the same, just over 2 percent a 7 Inflation-protected global government bonds year, in both countries. Again, a constant trend growth path seems 5% to fit reasonably well until the period beginning with the financial cri- sis. Since then the pattern of growth has been very different from its 4 earlier path. A persistent shortfall from the previous trend is evident. 3 Something significant has changed — and it is a matter of dispute as to whether the underlying productivity growth trend has fallen or 2 whether there is another reason for the pattern of persistently slow growth. 1 The most striking evidence of non-stationarity is shown in Figure 0 3. It plots the world real interest rate at a 10-year maturity, as calcu- -1 lated by David Low and myself from interest rates on government 1985 1995 2005 2015 bonds issued with inflation protection, from 1985 to the middle of 2017.8 From around the time when China and the members of the Source: NBER Working Paper No. 19887, data updated through Q2 2017 former Soviet Union entered the world trading system, long-term Figure 3 NBER Reporter • No. 3, September 2017 3 real interest rates have steadily declined to ability distributions over events that are summoned a plumber. You would hope reach their present level of around zero. Such drawn from a non-stationary economic that he might arrive with a large box of a fall over a long period is unprecedented. environment. Second, important move- tools, examine carefully the nature of And it poses a serious challenge to the one- ments in the world economy over the past the problem, and select the appropriate sector growth model. In order to salvage the quarter of a century cannot be explained tool to deal with it. Now imagine that model, much effort has been invested in the easily in terms of a one-sector model. The when the plumber arrived, he said that attempt to explain why the “natural” real minimum that is required is a two-sec- he was a professional economist but did rate of interest has fallen to zero or negative tor view of the world with both tradable plumbing in his spare time. He arrived levels. But there is nothing natural and non-tradable goods. How else are we with just a single tool. And he about a negative real rate of inter- looked around the kitchen for est. It is simpler to see Figure 3 as a Gross National Saving Rates, a problem to which he could disequilibrium phenomenon that China and Germany, 1980–2016 apply that one tool. You might cannot persist indefinitely. Percent of GDP think he should stick to eco- Part of the explanation lies 55 nomics. But when dealing with in saving behavior. Figure 4 China economic problems, you should shows the gross national saving 45 also hope that he had a box of rates for China and Germany tools from which it was possible 9 from 1980 through 2016. Their 35 to choose the relevant one. saving rates have risen markedly, And there are times when Germany especially in the case of China 25 there is no good model to and more recently in Germany. explain what we see. The prop- Figure 5 shows the saving rates osition that “it takes a model 15 in the U.S. and U.K. over the 1980 1986 1992 1998 2004 2010 2016 to beat a model” is rather pecu- same period. From the mid- Source: The IMF World Economic Outlook Database liar. Why does it not take a fact 1990s onwards, there has been to beat a model? And although a decline, again symptomatic of Figure 4 models can be helpful, why do non-stationarity. we always have to have one?10 What these charts show is that the to make sense of the changes in saving After the financial crisis, a degree of doubt experience of the past 25 years cannot and investment rates in the major econo- and skepticism about many models would easily be described as the outcome of sto- mies and the continuing current account be appropriate. chastic deviations from a stationary pro- imbalances? Let me now turn to the first of the cess. The data I have presented two failings I mentioned — the provide a prima facie case for Gross National Saving Rates, U.S. and U.K., 1980–2016 limitations of expected utility considering explanations based theory. Percent of GDP on a divergence from a sustain- 24 able growth path along which Uncertainty: The U.S. the composition and not just Fallacy of Bayesian the level of aggregate demand 20 Reasoning Outside a is a key driver of growth. To Frequentist Framework understand this requires going 16 beyond the one-sector work- U.K. I believe we need to face horse model that has come up to the challenge posed by 12 to dominate macroeconomic radical uncertainty — a state of teaching and policymaking. affairs in which we cannot enu- 8 That model, even with 1980 1986 1992 1998 2004 2010 2016 merate all the possible states modifications to first-order Source: The IMF World Economic Outlook Database of the world and hence can- conditions to allow for various not attach subjective probabili- new “frictions,” has two fail- Figure 5 ties to them. The only sensible ings. First, it leans heavily on the answers to the questions “Will assumption of forward-looking agents Now, I do not want to suggest that President Trump still be in the White who optimise over known probability we should abandon the workhorse model, House in 2021?” and “Will the U.S. econ- distributions of the shocks hitting the simply that we should be prepared to omy regain its pre-crisis trend growth economy. But there is little empirical consider other approaches. Imagine that path?” are “I don’t know.” None of the basis for computing the relevant prob- you had a problem in your kitchen, and possible outcomes represent a series of

4 NBER Reporter • No. 3, September 2017 repeated events in a stationary environ- estimates of probabilities. And it is not complete Arrow-Debreu markets where ment in which it would be possible to only history that casts doubt on the plau- people buy and sell in a single Walrasian construct probabilities based on observed sibility of the assumption of stationarity. “grand auction.”18 The two worlds are the frequencies. And the most important Learning from experience, including that same, and Savage was clear that the prop- future events that will shape the economy of others, means that expectations evolve osition that they describe a wide range of cannot even be imagined — new prod- over time and induce a non-stationarity in decisions was, in his own words, “utterly ucts emerged that were unimaginable 20 economic relationships. Large swings in ridiculous.” years ago. activity do not occur with sufficient fre- This isomorphism between complete We have all grown up with the simple- quency to permit a frequentist approach markets and the axiomatic basis for prob- minded methodological view that agents to estimating probabilities. They do not abilistic reasoning is no academic foot- can be modelled “as if ” they optimized occur in an environment that is stationary note. The world divides into two states. expected utility computed according to over the relevant time scales. And there In the first, we can construct probabilities Bayesian updated “personal probabilities.” is no basis on which to construct subjec- and markets are complete. In the second, Although useful in some — especially sta- tive probabilities other than to succumb radical uncertainty precludes the con- tionary — contexts, the expected utility to the temptation, described so clearly by struction of probabilities and markets are framework, fundamental to modern mac- Paul Romer, to impose priors to resolve incomplete. In the former, explanations roeconomics, has serious weaknesses that the identification problem.13 for macroeconomic fluctuations reflect make it unsuited to the analysis of major frictions in markets. In the latter, swings swings in economic activity. Equivalence of in activity are a natural consequence of Let me give a simple example. It Probabilistic Reasoning incomplete markets. In our toolbox there relates to my own experience when, as and Complete Markets is room for both approaches. But in try- deputy governor of the Bank of England, ing to understand large swings in activity, I was asked to give evidence before the What does radical uncertainty mean I favor the second. House of Commons Select Committee for macroeconomics? Much of macro- It is striking that the two major econ- on Education and Employment on economics and finance leans heavily on omists of the 20th century who took whether Britain should join the European models that assume, either explicitly or radical uncertainty seriously, Keynes and Monetary Union. I was asked how we implicitly, complete (Arrow-Debreu) Frank Knight, devoted their attention to might know when the business cycle in markets. As a result, those models are the two features of a capitalist economy the U.K. had converged with that on essentially static.14 What is less well that distinguish it from a Walrasian equi- the Continent. I responded that given understood is that a world of complete librium.19 Knight explored the nature the typical length of the business cycle, markets is isomorphic to a world in which of entrepreneurship, something that is and the need to have a minimum of subjective probabilities can be assigned to impossible to analyze outside radical 20 or 30 observations before one could all states of the world.15 uncertainty and incomplete markets. And draw statistically significant conclusions, In the mid-19th century, mathemati- Keynes wanted to understand why a capi- it would be 200 years or more before we cians started to develop an axiomatic basis talist economy was subject to large fluc- would know.11 And of course it would be for probability theory independent of tuations in output and employment. As absurd to claim that the stochastic pro- observed frequencies.16 Economists have Keynes was only too well aware, an idea cess generating the relevant shocks had been happy to adopt this approach to which is simple and obvious, but which is been stationary since the beginning of the uncertainty, even though its originators difficult to formalize mathematically, can Industrial Revolution. There was no basis were conscious of its limits. In his 1954 be resisted almost indefinitely. for pretending that we could construct a treatise on the foundations of statistics, L. The models used today assume fric- probability distribution. As I concluded, J. Savage was careful to assess the realism tions of various sorts to explain why “You will never be at a point where you of the axioms that underlay those founda- unemployment can persist. Yet it was this can be confident that the cycles have gen- tions. They rested on a theory of decisions view against which Keynes fought in the uinely converged; it is always going to be in which people looked ahead and antici- 1930s. He was adamant that even if wages a matter of judgment.”12 pated all possible branches of the decision were perfectly flexible, unemployment The fact that the economic processes tree. Savage described the world in which could persist. In distilling the essence generating growth and fluctuations do probabilistic reasoning applied in these of The General Theory, the most pene- not exhibit “stationarity” is of fundamen- words: “… acts and decisions, like events, trating analysis remains, in my view, the tal importance. It is why so many empiri- are timeless. The person decides ‘now’ 1975 review article by Don Patinkin. He cally estimated models break down. The once for all; there is nothing for him to highlights chapter 19, on money wages, world does not stand still long enough wait for, because his one decision pro- in which Keynes describes why a reduc- for an observer to measure the frequen- vides for all contingencies.”17 It is a “grand tion of money wages is not an effective cies that would enable her to construct decision.” But this is exactly the world of way to reduce unemployment: “The eco-

NBER Reporter • No. 3, September 2017 5 nomic system cannot be made self-adjust- uct does not calculate subjective prob- challenge of radical uncertainty. It is a ing along these lines.”20 Reductions in abilities and then maximize expected story that integrates the most impor- money wages increase desired employ- utility. There is no current price sig- tant pieces of information in order to ment, but if they also reduce expecta- nal to guide her decisions. Instead, she make a decision, and I provide examples tions of future incomes, then aggregate develops a narrative within which it is in my book. spending may fall and unemployment possible to understand the key param- persist. As Patinkin put it, “thus the eters determining the likely success of A Two-Sector Model with Slow General Theory is not a static theory the product, and makes a judgment. As Speed on the Turnpike Approach of unemployment equilibrium, but a Danny Kahneman put it: “No one ever dynamic theory of unemployment dis- made a decision because of a number. I turn now to the limitations of the equilibrium.”21 That is old-fashioned They need a story.”24 When the finan- one-sector nature of the canonical model language from the 1970s, but it points cial crisis hit in 2007, and took a major in macroeconomics. When Marty Feldstein to the centrality of the incompleteness turn for the worse in 2008, the reaction was a young man, multisector growth mod- of markets, which in turn rests on radi- of policymakers was not to update their els were all the rage. The optimal path from cal uncertainty.22 prior probabilities with each new obser- an initial starting point was to move toward What has been overlooked in the vation. It was to ask: “What is going on and then remain close to a balanced growth discussion of monetary policy in the here?” Or to quote Chuck Manski in a path along which all sectors grew at the same industrialised world today is that a sim- recent NBER Working Paper, rate. The early literature was concerned with ilar argument holds for interest rates. “Introspecting about how I finding conditions under which the opti- Central banks have flirted with nega- revise my own macroeconomic mal path would be close to the balanced tive interest rates. But for many econo- expectations after receipt growth path for most of the time — just as mists it has been a full-blown affair. The of new information, I often in a long car journey the optimal route is to prevailing view that the main obstacle get onto the highway and stay with it until find it difficult to conjecture 27 to our achieving macroeconomic stabil- an explicit sampling process. close to the final destination. Hence such ity is the zero lower bound on nominal Hence, I am unable to con- results were known as turnpike theorems interest rates is, I believe, more than a lit- and they were proven under rather general 23 sciously update in the Bayesian tle misleading. Negative interest rates manner.”25 have a substitution effect which raises current spending, but such a change Time does not Rebalancing toward Steady Growth in a Two-Sector Economy in policy may create expectations of permit a discussion Tradable goods and services Tradable goods and services future policy actions that would reduce of narratives as a way incomes. Aggregate spending could fall of describing macro- P* rather than rise. Such a possibility is economic events — I precluded by assumption in the work- provided one for the von Neumann ray C China, horse model. response of spending Germany Rational, or more accurately, to the financial crisis model-consistent, expectations proved in my book The End B U.S., U.K. invaluable in avoiding false inferences of Alchemy. But I do A about the impact of government inter- want to emphasize ventions. But if markets are incomplete, that I’m using the O Non-tradable goods and services Non-tradable goods and services it is easy to forget that expectations over word “narrative” in a Source: Author’s illustrative example future prices of goods for which there very different sense are no current futures markets will also from that deployed Figure 6 respond to changes in government poli- by Robert Shiller in cies. The Lucas critique applies equally his AEA Presidential Lecture earlier conditions. to incomplete and complete markets. this year. For him, a narrative is “a sim- Multisector models fell out of favor Feedback from negative interest rates to ple story or easily expressed explana- largely because of the focus on the steady- beliefs about future policies, and hence tion of events that many people want state of those models. If all sectors were incomes, cannot be ruled out. to bring up in conversation or on news growing at the same rate, then the models When confronted with radical or social media because it can be used added little to the insights provided by one- uncertainty, agents develop and evolve to stimulate the concerns or emotions sector models. But their real interest lies in narratives to cope with the challenge of of others.”26 It contrasts with a rational the adjustment path off the steady-state. making one-off decisions. An entrepre- view of the world. For me, a narrative is Without burdening you with formalities, neur thinking of launching a new prod- an entirely rational way to approach the Figure 6 shows an illustrative optimal path

6 NBER Reporter • No. 3, September 2017 for an economy with two sectors, tradable tor.28 And along the optimal path, measured sectors expand too rapidly, and for those and non-tradable goods and services. In the growth of total output will initially be weak that have experienced a relative decline in left panel, the solid line OP shows the bal- relative to the growth rate along the balanced their tradable goods sectors.30 One could anced growth path — often known as the path. In the one-sector model, the prob- easily imagine that the former illustrates lem does not really the challenge facing China and Germany Tradables to Non-Tradables Price Ratio, 1996–2014 arise. Deviations from today, whereas the latter represents the the steady-state path experience of the U.S. and the U.K. For 1.6 reflect random shocks, economies of both types, the task of real- which die away of their locating resources, including fixed capital,

1.4 own accord. may require a period of low growth and My two-sector falls in output in some sectors. The real U.K. division — between interest rate is important but it is not the 1.2 tradable and non- only relative price that matters in under- Germany tradable goods and standing slow growth today. 1 services — is styl- The key insight from such models is ized but captures, in simple but important: the composition my view, an impor- of demand matters. Trying to understand 0.8 1996 2014 tant division reflect- weak growth in the context of a single com- ing the imbalances in modity forces the debate into the arena of Source: Philip Lane, based on Berka et al. (2014), detailed in endnote 29 the world economy either weakness of aggregate demand or Figure 7 prior to the crisis, and slower productivity growth. But the turn- the need to rebalance pike theorem suggests that weak growth von Neumann ray — and the prices sup- now. Figure 7 shows the relative price of can be the property of an optimal response porting the path are given by the slope of the tradables versus non-tradables in both the to the need to rebalance the composition line orthogonal to that ray. The dotted line U.K. and Germany over the period 1996 of demand and output. I believe that that AB shows the path of an economy steadily through 2014, using data supplied by is exactly where we are today. moving away from balanced growth, as I Philip Lane, now Governor of the Central Figure 8 shows the imbalances last believe happened in the run-up to the cri- Bank of Ireland.29 It is clear that one of year among the four major parts of the sis, along which the relative price of the two the problems faced by the U.K. in trying goods differs from its steady-state value. to avoid unbalanced growth is the steady Having arrived at point B, the economy fall in the price of tradable goods and ser- Saving Rate and Current now needs to rebalance. Starting from point vices relative to non-tradables. Only fol- Account Surplus, 2016 B, the optimal path BC takes the economy lowing the sharp depreciation of sterling U.S. U.K. China Germany along the solid line which shows the optimal during the financial crisis was that relative 50 trajectory towards and then converging on price stable. For much of the period, there the balanced growth path. seems to be evidence that an unsustainably 0 For any initial composition of output, high real exchange rate led inexorably to Saving Rate (percent of GDP) such as B, the optimal path will stay within a a current account deficit and the need to 10 certain neighborhood of the von Neumann rebalance the economy. With the further ray for most of the time. But starting from appreciation of sterling in 2014 and 2015, 0 an unbalanced combination of tradable and the fall in the relative price resumed. All non-tradable sectors, the interesting turn- this puts the depreciation of sterling since -10 pike result is that in order to get back to last summer into perspective. In contrast, Current Account Surplus / Deficit a balanced economy, it pays to reallocate Germany has experienced, if anything, a (percent of GDP) 500 resources between the two sectors sooner rising price of tradables, and it is hence no rather than later. In the sector that has over- surprise that its current account surplus expanded, that may require a contraction of has risen to unsustainable levels, around 0 output and writing off of capital. Focusing on 8½ percent of GDP last year. -500 the adjustment to the equilibrium path — or The right panel in Figure 6 shows the Current Account Surplus / Deficit the “traverse” in [John] Hicks’ terminol- route back to a balanced growth path for (in billions of U.S. dollars) ogy — brings an Austrian flavor to the analy- economies with the same technology but Source: The IMF World Economic sis of growth in the two-sector model, espe- with different histories of their tradable Outlook database cially with the possibility that it is optimal to goods sectors. It shows the path for econ- discard capital invested in the “wrong” sec- omies that have seen their tradable goods Figure 8

NBER Reporter • No. 3, September 2017 7 world economy in which current account able to the tradable sector. No doubt five-year spot rate staying remarkably deficits and surpluses are significant. Both other explanations will be forthcoming. close to zero. the U.S. and U.K. had substantial current And in truth it is too soon to tell. account deficits, amounting in aggregate But do not be misled into think- Conclusions to around $600 billion, and China and ing that, because unemployment is low, Germany had correspondingly large cur- an unfortunate sequence of negative The moral of my story is that it is rent account surpluses. All four econ- shocks has come to an end, and nor- important not to be constrained by existing omies need to move back to models, nor to think that simply a balanced growth path. But Implied Forward Rates, U.S., 2004–17 tinkering with those models pro- far too little attention has been vides an answer to the challenges 4% paid to the problems involved Implied 10-year real rate 10 years forward posed by the crisis and by unex- in doing that. With unemploy- pectedly slow growth over the ment at low levels, the key prob- past decade. lem with slower-than-expected 2 I am not suggesting that growth is not insufficient aggre- we should abandon our existing gate demand but a long period tools. It is a question of horses away from the balanced path, 0 for courses. But the workhorse reflecting the fact that rela- model does not constitute a com- tive prices are away from their Implied 10-year real rate 10 years forward, minus 5-year real spot rate prehensive toolkit. Remember steady-state levels. The result is -2 the lesson of the good economics that the shortfall of GDP per 2005 2010 2015 plumber — carry many tools with head relative to the pre-crisis Source: Calculations using data from the U.S. Treasury you, and always pose the question: trend path was over 15 percent What is going on here? Designing in both the U.S. and U.K. at the Figure 9 practical policies to improve pub- end of last year. Policies which focus only mality is about to be restored. Figure 9 lic interventions is a continuing challenge, on reducing the real interest rate miss the shows the real interest rate that markets and one that Marty has explored through- point; all the relevant relative prices need expect to hold 10 years from now in the out his career in a wide variety of fields: to change, too. U.S.: the 10-year 10 year forward rate health, taxation, saving, social security, implied by the yield structure of real monetary and macroeconomic policies, and Another Story rates. The crisis dashed hopes that real even defense economics. Taken together, those contributions certainly add There are many stories 5-Year Real Spot Rate, U.S., 2004–17 up to a life well-lived. which purport to explain recent Marty is still a role model growth experience. There is 5% for younger economists who the decline in growth poten- want to be the kind of econom- tial emphasized by Robert ics plumber that every family Gordon,31 secular stagna- would trust with their kitchen. tion advanced by Lawrence And even after almost 50 years Summers,32 and others. For I look forward to a few more example, a recent paper by John 0 decades of learning from my Fernald, Robert Hall, James mentor. Stock, and Mark Watson attri- 1 butes slow growth to a declin- -2 To be precise, the first 2005 2010 2015 ing trend in total factor pro- lines of Robert Browning’s ductivity and a decline in labor Source: Calculations using data from the U.S. Treasury poem Home-Thoughts, from force participation.33 Perhaps, Abroad, written in 1845, are: perhaps not. Recent growth has Figure 10 “Oh, to be in England been very similar in the U.S. and U.K. rates might go back to something more Now that April’s there” But in the U.K., labor force participation normal and the current expectation is Return to text. has risen, not fallen. And it is possible to close to only 1 percent a year. It also 2 J. M. Keynes, The General Theory reconcile low unemployment with weak shows the same implied rate minus the of Employment, Interest and Money, growth as the property of a transition to five-year spot real rate. Again, there is London: Macmillan and Co., 1936, pp. a two-sector turnpike path during which little sign of market expectation of nor- viii. resources must shift from the non-trad- malization. And Figure 10 shows the Return to text.

8 NBER Reporter • No. 3, September 2017 3 Exceptions include P. Romer, mist, see E. Duflo, “The Economist as 18 See the discussion of the “grand auction” “The Trouble with Macroeconomics,” Plumber,” American Economic Review, in M. King, The End of Alchemy, New Stern School of Business, New York 107(5), 2017, pp. 1–26. York: W.W. Norton, 2016. University, September 2016. Return to text. Return to text. Return to text. 10 A similar point is made in R. Reis, 19 J. M. Keynes, Treatise on 4 An interesting discussion of the evo- “Is Something Really Wrong with Probability, London: Macmillan lution of growth models may be found Macroeconomics?” London School of and Co., 1921, and F. Knight, Risk, in S. Spear and W. Young, “Two- Economics, June 2017, mimeo. Uncertainty and Profit, Boston: Hart, Sector Growth, Optimal Growth, and Return to text. Schaffner & Marx, Houghton Mifflin the Turnpike: Amalgamation and 11 “In the longer run what is likely to be Co., 1921. Metamorphosis,” Macroeconomics the potential cost of a one-size-fits-all Return to text. Dynamics, 19(2), 2015, pp. 394–424, monetary policy? That is extremely hard 20 J. M. Keynes, The General Theory and in the classic survey paper, F. to judge and to be confident. The reason of Employment, Interest and Money, Hahn and R.C.O. Matthews “The why I think one can say you will never London: Macmillan and Co., 1936, p. Theory of Economic Growth: A Survey,” really know is that to have enough expe- 267. Page references are to the Royal Economic Journal, 74(96), 1964, pp. rience, enough observations, on business Economic Society Collected Writings 779–902. cycles to find out whether they have of John Maynard Keynes. Return to text. converged — the IMF Study did not Return to text. 5 F. Ramsey, “A Mathematical Theory cover very long periods — you need 200 21 D. Patinkin, “The Collected Writings of Saving,” Economic Journal, 38(152), or 300 years of data.” Select Committee of John Maynard Keynes: From the 1928, pp. 543–59. on Education and Employment Minutes Tract to the General Theory,” Economic Return to text. of Evidence, Thursday 27 May 1999, Journal, 85, June 1975, p. 257. 6 Data for 1900 through 2010 from Q uestion 46. Return to text. the Maddison Project at www.ggdc.net/ Return to text. 22 Many of these arguments about the maddison/maddison-project/home. 12 See previous citation. formation of expectations were expound- htm, (2013 version), updated from the Return to text. ed at length in the writings of G.L.S. IMF WEO Database for April 2017. 13 P. Romer, “The Trouble with (George) Shackle. For the methods used to construct these Macroeconomics,” Stern School of Return to text. data, see J. Bolt and J.L. van Zenden, Business, New York University, 23 I discuss the issue in chapter eight of “The Maddison Project: Collaborative September 2016. M. King (endnote 18). Research on Historical National Return to text. Return to text. Accounts,” The Economic History 14 See the discussion in S. Spear and W. 24 Q uoted in M. Lewis, The Undoing Review, 67(3), 2014, pp. 627–51. Young (endnote 4). As Reis (endnote 10) Project, New York: W.W. Norton, 2016, Return to text. has argued, there are recent efforts to p. 250. 7 The computed trend growth rates introduce elements of non-stationarity into Return to text. over 1960–2007 are 2.2 percent a year small theoretical models. But they have yet 25 C. Manski, “Survey Measurement in the U.S. and 2.3 percent a year in the to alter the mainstream of thinking about of Probabilistic Macroeconomic U.K. policy and still rest on the assumption of Expectations: Progress and Return to text. expected utility maximization. Promise,” NBER Working Paper 8 The data in M. King and D. Low, Return to text. No. 23418, forthcoming in NBER “Measuring the ‘World’ Real Interest 15 This issue is explored more fully in Macroeconomics Annual 2017, volume Rate,” NBER Working Paper No. forthcoming work by John Kay and myself. 32, M. Eichenbaum and J. Parker, eds., 19887, February 2014, with updates to Return to text. Chicago: University of Chicago Press. late 2016 by the authors. The extension 16 An excellent discussion of the develop- Return to text. to July 2017 relies on the change in the ment of probability theory may be found 26 R. Shiller, “Narrative Economics,” U.S. TIPS real yield published on the in L. Daston, Classical Probability in American Economic Review Papers U.S. Treasury website, because the Bank the Enlightenment, Princeton: Princeton and Proceedings, 107, 2017, pp. 967– of England has temporarily (as of June University Press, 1988. 1004. 2017) suspended publication of real Return to text. Return to text. yields while the estimation methodology 17 L. Savage, The Foundations of 27 A useful survey of turnpike models is reviewed. Statistics, New York: John Wiley & Sons, is given in L. McKenzie, “Turnpikes,” Return to text. 1954, p. 17. The page reference is to the American Economic Review, 88(2), 9 For a view of the economist as plumb- 1972 second revised edition. 1998, pp. 1–14. er from the perspective of a microecono- Return to text. Return to text.

NBER Reporter • No. 3, September 2017 9 28 See John Hicks’ article on neo-Aus- tradables divided by the sum of all item as in the U.K. and Germany. trian growth theory: J. Hicks, “A Neo- weights. Return to text. Austrian Growth Theory,” Economic Return to text. 31 R. Gordon, The Rise and Fall of Journal, 80(318), 1970, pp. 257–81. 30 It is important to remember the dis- American Growth: The U.S. Standard Return to text. tinction between traded and tradable of Living since the Civil War, 29 Tradable and non-tradable infla- goods and services. Although the defini- Princeton: Princeton University Press, tion rates are compiled as a weighted tions used in any particular empirical 2016. average of price changes of individual application are somewhat arbitrary, Return to text. consumption goods. Price changes and hence the data need to be used with 32 L. Summers, “The Age of Secular and item weights are drawn from the some circumspection, for Figure 7 the Stagnation: What It Is and Eurostat HICP database (ec.europa. share of tradable goods and services in What to Do About It,” Foreign eu/eurostat/web/hicp/data/database). total output in 2014 was 58.5 percent Affairs, March/April 2016. All items are classified as either trad- in Germany and 58.4 percent in the Return to text. able or non-tradable, following M. U.K., well above their respective shares 33 J. Fernald, R. Hall, J. Stock, Berka, M. Devereux, and C. Engel, in actual trade. An attempt to calculate and M. Watson, 2017, “The “Real Exchange Rates and Sectoral the relative price of tradables versus Disappointing Recovery of Output Productivity in the Eurozone,” NBER non-tradables in the U.S. has been after 2009,” NBER Working Paper No. Working Paper No. 20510, September made by Rui Mano of the IMF, and the 23543, June 2017, forthcoming in 2014, Table 1. The non(tradable) U.S. too has experienced a fall in the Brookings Papers on Economic consumption weight is the sum of all relative price of tradables with a simi- Activity. item weights that are classified as (non) lar proportion between the two sectors, Return to text.

10 NBER Reporter • No. 3, September 2017 Research Summaries

Old Idea, New Insights: The Ricardian Revival in International Trade

Arnaud Costinot and Dave Donaldson

Two centuries ago, David Ricardo dictions of a Ricardian model line up with wrote down a simple thought experi- data on trade patterns? In a famous chal- ment that changed the way economists lenge, a mathematician, Stan Ulam, once think about international trade. Suppose asked Paul Samuelson to name one prop- the residents of two nations, England and osition in the social sciences that is both Portugal, differ in their ability to pro- true and nontrivial. After much reflection, duce two goods, cloth and wine: Portugal Samuelson’s reply was: “Ricardo’s theory of Arnaud Costinot is a is more efficient at both, but its relative comparative advantage.” research associate in the advantage over England is weaker in cloth. The practical content of Ricardo’s the- NBER International Trade and If these countries are able to trade, what ory has received surprisingly little atten- Investment Program. A profes- will happen? Who will trade what with tion due to the challenges of connect- sor of economics at MIT, he whom? Who will gain from the trades? ing Ricardo’s ideas to data. Together with received his B.S. from École How large will the gains be? Ivana Komunjer, we have extended Eaton Polytechnique in 2000, his M.A. Ricardo’s famous example has been and Kortum’s quantitative model to study from École des Hautes Études en used to answer these fundamental ques- inter-industry Ricardian comparative Sciences Sociales in 2001, and his tions of international trade in countless advantage in a way that is amenable to Ph.D. from Princeton University textbooks: England imports Portuguese empirical scrutiny.2 in 2005. He is a co-editor of wine and everyone’s a winner, all the more The basic prediction of the Ricardian the Journal of International so the worse the English are at making model is that countries should export rel- Economics, a foreign editor of wine. atively more in sectors in which they are the Review of Economic Studies, Until recently, however, Ricardo’s relatively more productive. Our model cap- and an associate editor of the logic has had surprisingly little impact on tures this simple idea by providing closed- American Economic Review the way that economists use data from form solutions for relative bilateral trade and the Journal of Economic the world around them to answer ques- flows as a function of relative observed pro- Literature. tions about trade policy. Extending the ductivity. Crucially, the model takes into His research has focused on logic to a realistic economy with many account the fact that countries will not a variety of positive and norma- regions and products had seemed some- produce all varieties of every good. Rather, tive issues in international trade, where between impractical and impossible. a country will only produce those variet- including the foundations of the But thanks to a number of recent inno- ies in which it is relatively more efficient. theory of comparative advantage, vations — most importantly in the semi- This implies that differences in observed the measurement of the wel- nal work of Jonathan Eaton and Samuel productivity tend to be smaller than true fare gains from trade, the role Kortum — this is rapidly changing.1 In differences in productivity as a result of a of intermediaries in international this research report, we describe some of selection effect. exchanges, the impact of global our recent attempts to connect Ricardo’s Combining standard data on industry- supply chains, and the design of 200-year-old idea to the real world. level trade flows and productivity, we find optimal trade policy and capital that countries do indeed tend to export controls. Ricardian Comparative goods where their relative productivity is Costinot grew up in Advantage and Empirical higher, as this Ricardian model would pre- Dunkirk, in the north of France. Patterns of Trade dict. More precisely, a 1 percent change He currently lives in Brookline, in relative productivity is associated with Massachusetts, with his wife, We begin by asking the most basic of a 6.5 percent change in relative exports. Nadège, and their two children, empirical questions: How well do the pre- There is also support for the notion that Paul and Alice.

NBER Reporter • No. 3, September 2017 11 observed productivity differences are crop, that any location could achieve as a biased by Ricardian selection. We use our function of its environmental conditions. estimates to quantify the welfare impact For example, the Global Agro-Ecological of this Ricardian channel across sectors. Zones (GAEZ) project from the Food Cross-industry differences in technolo- and Agricultural Organization aggregates gies generate only a small part of the gains such predictions for all major crops and at from trade; comparative advantage oper- a detailed geographical level. The project ates mostly at the within-industry prod- includes information on 2.2 million par- uct level. cels of land around the globe. This permits one to “observe” not only the productivity Ricardo’s Difficult of a land parcel in its current use, but also Identification Problem in all potential uses. In a first paper, we focused on a direct “Ricardo’s difficult idea,” as Paul test of the Ricardian model.3 On the basis Dave Donaldson is a Krugman once referred to the theory of of GAEZ potential yield observations, re­search associate in the NBER’s comparative advantage, contains at its core we calculated the Ricardian model’s pre- Development Economics, a fundamental barrier to empirical anal- dictions about the pattern of produc- Development of the American ysis. Ricardo’s simple example involved tion — which crops growers at different Economy, and International a prediction about trade patterns as a locations would choose to specialize in, Trade and Investment Programs, function of four productivity numbers — at prevailing producer prices. These pre- and a professor of economics England’s and Portugal’s productivity lev- dicted patterns of production have signifi- at MIT. He is a co-editor of els in cloth and wine. But how is an ana- cant power to predict actual patterns of the American Economic Journal: lyst to measure England’s productivity in crop production around the world — per- Applied Economics, an editorial an activity, such as wine making, which it haps surprisingly so, given the many rea- board member of the Journal of does not engage in because it can import sons for actual productivities to differ Economic Literature, the Journal wine from Portugal? Without knowledge from those predicted by agronomists. of International Economics, of this missing productivity number, it Moving beyond testing, a core inter- the Review of Economic Studies, is impossible to test the model’s predic- est is in estimating the gains from trade and the Q uarterly Journal of tions about the patterns of trade. Yet the that exist in the world around us. As in Economics, and a program direc- very essence of the model implies that this all standard trade models, the Ricardian tor (for trade) at the International fourth number should not be observable model postulates that regions differ and Growth Centre. to an analyst. those differences give rise to potential Donaldson has investigated This empiricist’s Gordian knot — for- gains from specialization afforded by the the welfare and other effects of mally, an “identification” problem — is ability to trade. But how large are the market integration, the impact familiar in selection models throughout differences, and hence how large are the of improvements in transporta- economics, but standard solutions have gains? Unfortunately Ricardo’s identifica- tion infrastructure, how trade been difficult to apply to the study of tion problem again stands in the way. All can mediate the effects of cli- international trade. Previous attempts to four of Ricardo’s numbers are needed to mate change, and how trade test the Ricardian theory, including our evaluate the gains from trade, for either affects food security and famine. aforementioned paper, are therefore based England or Portugal. The key is to estimate He has been awarded the 2017 on strong functional-form assumptions just how much more efficient the world , as well that impose a particular structure on the is when England doesn’t have to produce as an Alfred P. Sloan Research distribution of productivity across goods any wine. That efficiency boost depends on Fellowship, and has benefited (and varieties of the same good) to allow how bad England is at producing the wine from support from the National an analyst to infer underlying productiv- that it doesn’t produce. Science Foundation. ity differences from observed differences. In a second recent paper, we again draw Donaldson is a native of In recent research, we have drawn on the GAEZ data to measure some of the Toronto. He received an under- on some unique features of the agricul- gains from agricultural trade.4 In particu- graduate degree in physics from tural sector in order to make progress lar, we ask how much of the growth in U.S. the University of Oxford and a on Ricardian empirical work despite the agricultural productivity over the period Ph.D. from the London School identification challenges. The key obser- 1880–1997 has come about because of of Economics. He lives in Boston vation is that this is a setting in which a the spatial integration of agricultural mar- with his wife, daughter, and three major scientific focus among agronomists kets across the United States. A key chal- sons. is to predict the productivity, for any lenge to incorporating information from

12 NBER Reporter • No. 3, September 2017 the GAEZ data in this historical case ing climate will portend lower yields for tially. The value of crop output would is that potential yields, for each crop in many crops in many locations. But the be predicted to fall by about 40 percent. each location, have changed, in unknown aggregate consequences of those mil- However, there is enough heterogeneity ways, due to technological progress. But lions of micro-shocks will depend on in these shocks over space that after real- under the assumption that those changes the extent to which farmers can shift locating production according to compar- do not reverse comparative advantage their growing patterns from one crop to ative advantage across crops within each within any county, we show how to use another, and the extent to which con- parcel, welfare losses become smaller by data from the Agricultural Census since sumers can change the trade linkages a factor of three. Furthermore, there is so 1880 to infer the unique set of prices and that connect them to particular farmers. much productivity heterogeneity across productivity shocks that To examine these pos- parcels within countries that there is lit- is consistent with profit U.S. Agricultural Gains sibilities we build a gen- tle to be gained from controlling coun- maximization and factor- from Increased Economic eral equilibrium model of tries’ capacity to adjust their trade flows market clearing in any Integration trade, in 10 leading crops, internationally. given county and year. Total output growth due to among each of 1.7 mil- integration and productivity These estimated, model- 80% lion land parcels around Reducing Ricardian Complexity consistent prices cor- the world. Each parcel relate well with data on 60 has its own Ricardian The Ricardian world can be a com- actual state-level prices productivity capabili- plicated place. The full equilibrium and show a clear trend 40 ties in each crop, so we implications of Ricardo’s four numbers, toward lower intra-U.S. have 17 million crop-par- in his simple two-by-two setting, took spatial price dispersion 20 cel estimates, and trade almost three decades to work out, as over time. Commodity occurs subject to trad- John Stuart Mill did in 1844.6 So what 0 markets were more inte- 1880–1920 1954–1997 ing frictions designed to is the Ricardian analyst — let alone the grated in 1997 than in match world trade flows reader — to make of the complexity of Source: Researchers’ calculations 1880. using data from the U.S. Census of today. We then shock a model like that discussed above with Agriculture and the Agricultural Time But how large are the Series-Cross Section dataset each parcel’s productiv- 17 million productivity ingredients? In benefits from this height- ity level in each crop in recent work with Rodrigo Adao, we ened integration? To Figure 1 the manner that clima- have developed a new methodology for shed light on this question, tologists and agronomists simplifying the empirical use of general we calculate the change in the value of believe will arise due to climate change neoclassical models, a class that includes nationwide output that the 1880 econ- by 2071–2100. This is feasible because the Ricardian model as a special case.7 omy would have enjoyed if inter-spatial the GAEZ project provides their agro- We first establish the equivalence price gaps in 1880 were set to their later nomic predictions both for contempo- between such models and reduced (say 1920) level, rather than to the actual rary climate conditions and also under exchange models in which countries 1880 level. The results are surprising. For the expected climate conditions that directly exchange factor services, extend- example, 1880–1920 gains were 79 per- correspond to each of the scenarios used ing an original insight from Charles cent, approximately the same as the gains by the UN’s Intergovernmental Panel on Wilson in the Ricardian case.8 This that we calculate are due to pure within- Climate Change. location productivity gains. Similar state- What are the Adaptation, Trade, and the Eects of Climate Change ments are true about a later period, 1954 consequences of this to 1997. Overall, the increasing exploita- change in the 17 mil- Estimated change in national welfare (percent of GDP) Canada United States Mexico tion of gains from intra-national trade in lion place- and crop- 0.8 this context appear to have been a major, specific productivi- Production and trade and perhaps underappreciated, contribu- ties? We find that 0.4 adjustments tor to aggregate economic growth. climate change would generate a large neg- 0.0 Comparative Advantage and ative productiv- -0.4 the Costs of Climate Change ity shock for many No production or Only production trade adjustments adjustments countries around -0.8 Together with Cory Smith, we have the world and that if also applied Ricardo’s logic to study of there were no reallo­ Authors' calculations using various data sources and climate model projections Source: Researchers’ calculations using data from the GAEZ project, the IPCC program, the consequences of climate change in cations around the and the FAOSTAT program agricultural markets around the world.5 world, welfare would There is little doubt that a warm- decrease substan- Figure 2

NBER Reporter • No. 3, September 2017 13 equivalence implies that, for an arbi- theoretical considerations in the agri- Ricardo’s Rejuvenation trary change in trade costs, counterfac- cultural sector. The market structure tual changes in the factor content of in this sector is plausibly close to the New data sources, new modeling trade, factor prices, and welfare only neoclassical ideal and, again, agronomic strategies, and new empirical procedures depend on the shape of a reduced fac- data enable a unique view of compara- have breathed new life into Ricardo’s tor demand system. We then provide tive advantage. We find that about half 200-year-old insights about comparative sufficient conditions for estimates of of the welfare gains from optimal trade advantage and trade flows. This revital- this factor demand system to be recov- taxes arise from the use of non-uniform ized line of work has generated important ered non-parametrically. Together, these trade taxes that vary monotonically with insights on a range of applied questions, results offer a strict generalization of comparative advantage. including the design of border taxes, the the parametric approach used in so- origins of aggregate productivity gains, called gravity models — like the version Home Demand as a and the expected harm from climate of the Ricardian model developed by Source of Ricardian change. We have recently surveyed many Eaton and Kortum — in which the fac- Comparative Advantage of these new developments.13 The impact tor demand system is isoelastic. of Ricardo’s path-breaking work may be In more recent work we have even greater in its third century than in Implications turned to a more basic question: Where its first two. for Optimal Trade Policy do Ricardo’s cross-country differences in relative productivities come from? What does the Ricardian model In models that incorporate increasing 1 J. Eaton and S. Kortum, “Technology, imply for the design of a nation’s opti- returns, be they of the Marshallian sort Geography, and Trade,” Econometrica, mal trade policy? Should it protect more that are external to firms or of the 70(5), 2002, pp. 1741–79. import sectors with weaker comparative monopolistic competition sort that play Return to text advantage? Conversely, should it subsi- out through firm entry in differentiated 2 A. Costinot, D. Donaldson, and I. dize less in export sectors with stronger product markets, the productivity of a Komunjer, “What Goods Do Countries comparative advantage? Perhaps surpris- given industry in a given nation rises as Trade? A Q uantitative Exploration of ingly, in spite of the importance of the its output increases. This opens up the Ricardo’s Ideas,” NBER Working Paper theory of comparative advantage in the possibility, for sufficiently strong aggre- No. 16262, August 2010, and Review of field of international trade, these ques- gate economies of scale, of what Staffan Economic Studies, 79(2), 2012, pp. 581– tions have not previously been inves- Linder and Krugman call the “home- 608. tigated formally. The goal of our work market effect,” in which a region’s home Return to text with Jonathan Vogel and Iván Werning demand base will become a source of 3 A. Costinot and D. Donaldson, “Ricardo’s is to shed light on these questions.9 endogenous Ricardian comparative Theory of Comparative Advantage: Old The main theoretical result of our advantage.10 Idea, New Evidence,” NBER Working Paper paper is that, in the context of a canoni- In work with Margaret Kyle and No. 17969, April 2012, and American cal Ricardian model, optimal import , we estimate the Economic Review Papers and Proceedings, tariffs should be uniform, whereas opti- strength of this effect in the context of 102(3), 2012, pp. 453–8. mal export subsidies should be weakly the global pharmaceutical industry.11 Return to text decreasing with respect to compara- Building on previous work on the 4 A. Costinot and D. Donaldson, “How tive advantage. While the latter pat- effect of demographic changes on inno- Large Are the Gains from Economic tern accords well with the observation vation and product entry by Daron Integration? Theory and Evidence from that countries tend to protect their Acemoglu and Joshua Linn,12 our U.S. Agriculture, 1880–1997,” NBER least competitive sectors in practice, paper establishes that countries that for Working Paper No. 22946, December larger subsidies do not stem from a demographic reasons are expected to 2016. greater desire to expand production in have high demand for a certain type of Return to text less competitive sectors. Rather, they drug are actually more likely to be net 5 A. Costinot, D. Donaldson, and C. reflect tighter constraints on the ability exporters of that drug. We find that the Smith, “Evolving Comparative Advantage to exploit monopoly power by contract- correlation between predicted home and the Impact of Climate Change in ing exports. Put simply, countries have demand and sales abroad is positive and Agricultural Markets: Evidence from 1.7 more room to manipulate world prices greater than the correlation between Million Fields Around the World,” NBER in the sectors in which they have a com- predicted home demand and purchases Working Paper No. 20079, April 2014, parative advantage. from abroad, which is strong evidence and Journal of Political Economy, 124(1), The final part of the paper explores for the role of the home market in cre- 2016, pp. 205–48. the quantitative importance of these ating comparative advantage. Return to text

14 NBER Reporter • No. 3, September 2017 6 J. S. Mill, Essays on Some Unsettled 9 A. Costinot, D. Donaldson, J. Vogel, 22538, August 2016. Questions in Political Economy, London: and I. Werning, “Comparative Advantage Return to text Longmans, Green, Reader, 1844; and Dyer, and Optimal Trade Policy,” NBER 12 D. Acemoglu and J. Linn, “Market 2nd edition, 1874. Working Paper No. 19689, December Size in Innovation: Theory and Evidence Return to text 2013, and Quarterly Journal of from the Pharmaceutical Industry,” NBER 7 R. Adao, A. Costinot, and D. Donaldson, Economics, 130(2), 2015, pp. 659–702. Working Paper No. 10038, October 2003, “Nonparametric Counterfactual Predictions Return to text and Quarterly Journal of Economics, in Neoclassical Models of International 10 S. Linder, An Essay on Trade and Trans­ 2004, 119(3), pp. 1049–90. Trade,” NBER Working Paper No. 21401, formation, Uppsala, Sweden: Almqvist Return to text July 2015, and American Economic & Wiksells, 1961; P. Krugman, “Scale 13 A. Costinot and J. Vogel, “Beyond Review, 107(3), 2017, pp. 633–89. Economies, Product Differentiation, and Ricardo: Assignment Models in Return to text the Pattern of Trade,” American Economic International Trade,” NBER Working 8 C. Wilson, “On the General Structure Review, 1980, 70(5), pp. 950–9. Paper No. 20585, October 2014, and of Ricardian Models with a Continuum Return to text Annual Review of Economics, 7, 2015, of Goods: Applications to Growth, 11 A. Costinot, D. Donaldson, M. Kyle pp. 31–62; D. Donaldson, “The Gains Tariff Theory, and Technical Change,” and H. Williams, “The More We Die, the from Market Integration,” Annual Review Econometrica, 1980, 48(7), pp. 1675–702. More We Sell? A Simple Test of the Home- of Economics, 7, 2015, pp. 619–47. Return to text Market Effect,” NBER Working Paper No. Return to text

NBER Reporter • No. 3, September 2017 15 Integrated Assessment Models of Climate Change

William Nordhaus

Many areas of the natural and lenge to natural and social scien- social sciences involve complex tists, who must incorporate a wide systems that link multiple areas variety of geophysical, economic, and disciplines. This is particu- and political disciplines into their larly true for the science, econom- diagnoses and prescriptions. ics, and policy of climate change, Integrated assessment mod- William Nordhaus is Sterling Professor of which involve a wide variety of els of climate change grew organi- Economics at Yale University and a research fields from atmospheric chemistry cally from energy models. One of associate in four NBER Programs: Economic to game theory.1 As understand- the earliest careful comparisons of Fluctuations and Growth; Environment and ing progresses across the differ- energy models was the Modeling Energy Economics; Productivity, Innovation, ent fronts, it is increasingly neces- Resource Group (MRG) analysis of and Entrepreneurship; and Public Economics. sary to link the different areas to the 1970s.2 This project, chaired He is a pioneer in analyzing economic conse- develop models and policies that by Nobel Prize-winning economist quences of climate change and one of the lead- reflect the complex interactions. Tjalling Koopmans, analyzed sev- ing global authorities on climate policy. A full analysis would reflect that eral energy models that projected Nordhaus developed the DICE and RICE economic activity drives emissions, energy demands and technologies models of the economics of climate change, which affect atmospheric concen- over a long time horizon. The ear- which are widely used in research on climate- trations, thence climate and the lier work of Koopmans on the lin- change economics and policies. He also has hydrological cycle, which in turn ear programming approach to pro- studied a range of additional topics in both affect human and natural systems, duction, as well as the Samuelson micro- and macroeconomics, including wage which ultimately contribute to the principle of “markets as maximiza- and price behavior, health economics, aug- determination of climate policies. tion,” formed the intellectual core mented national accounting, the political busi- Integrated assessment analyses of much of the energy modeling ness cycle, and the dynamics of productivity. He and models play a key role in put- starting at that time and proceed- is the author of three treatises on climate pol- ting the pieces together. Integrated ing to the present.3 icy, The Climate Casino, Managing the Global assessment models (IAMs) can be The first IAMs in climate Commons, and Warming the World, and is a co- defined as approaches that inte- change were basically energy models author, with the late Paul Samuelson, of the clas- grate knowledge from two or more with an emissions model included, sic textbook, Economics. domains into a single framework. and later with other modules such Nordhaus served as president of the These are sometimes theoreti- as a carbon cycle model and a small American Economic Association (AEA) in 2015. cal but are increasingly comput- climate model. My early approaches He is a member of the National Academy of erized, empirical, dynamic, non- were partial equilibrium energy Sciences and a Fellow of the American Academy linear models of varying levels of models with exogenous output.4 of Arts and Sciences, as well as a Distinguished complexity. A. S. Manne’s model, the first to Fellow of the AEA. He also has been a mem- The challenge of coping with imbed an energy system in a full ber of the President’s Council of Economic global warming is particularly economic-growth model, was an Advisers (1977–9), provost of Yale University daunting, because it spans many important landmark.5 The earliest (1986–8), a member of the Brookings Panel on disciplines and parts of society. versions of my models adopted a Economic Activity since 1972, and a director Ecologists may see it as a threat to growth-theoretic framework simi- and chair (2013–14) of the board of directors of ecosystems, marine biologists as a lar to Manne’s and extended it to the Federal Reserve Bank of Boston. problem leading to ocean acidifi- geophysical variables.6 A native of Albuquerque, New Mexico, who cation, and coastal communities as IAMs are increasingly used in received his B.A. from Yale and his Ph.D. from a lottery with intense hurricanes, analyses by national governments MIT, Nordhaus lives in New Haven with his while ski resorts may view it as and in international assessments. wife, Barbara, who works at the Yale Child Study a mortal danger to their already Among the most important appli- Center. short seasons. It also poses a chal- cations are:

16 NBER Reporter • No. 3, September 2017 • Making projections that Climate and the Economy (RICE) specified using the Negishi approach, have consistent inputs and out- models have gone through several in which regions are aggregated using puts of the different components revisions since their initial develop- time- and region-specific weights sub- of the system so that, for example, ment around 1990. The latest pub- ject to budget constraints. the GDP projections are consistent lished versions are the RICE-2010 This sketch of a pair of IAMs in with the emissions projections. and DICE-2016R2 models. The latest the DICE and RICE models makes it • Calculating the impacts of DICE model is available in GAMS, clear that they are highly simplified alternative assumptions on impor- a fine mathematical software system, representations of complex economic tant variables such as output, emis- and a full description of the earlier and geophysical realities — what sions, temperature change, and version is available.7 might be called geo-macroeconomics. the effect of economic activity on DICE is a globally aggregated While small and comprehensive mod- climate. model. RICE is essentially the same, els have many advantages, they also • Tracing the effects of alter- except that output and abatement have major shortcomings because of native policies on all variables in a have structures for 12 regions. This their simplifications. consistent manner as well as esti- discussion will use the term “DICE A useful analogy here is to return mating the costs and benefits of model,” and for most modules the to the animal kingdom. Each model is alternative strategies. analysis applies equally to the RICE like an animal that has its fruitful niche • Estimating the uncertainties model. in the analytical ecosystem. Small mod- associated with alternative variables The DICE model views the eco- els can be fleet and can adapt easily to and strategies along with the value nomics of climate change from the a changing environment or new data, of research and new technologies. perspective of neoclassical economic while large models take many years to There are dozens of IAMs today, a growth theory. In this approach, mature but are able to handle much handful of which have a track record economies make investments in cap- larger and more complex tasks. of at least a decade. Models range ital, education, and technologies, from small ones like the DICE model thereby reducing consumption today Illustrative Results described below to enormous ones in order to increase consumption in that may have as many as a half-mil- the future. The DICE model extends Here are some representative results lion variables. Different IAMs are like this approach by including the “natu- using the DICE model. These results different animals in terms of com- ral capital” of the climate system as an are from the most recent version of the parative strengths and weaknesses in additional kind of capital stock. By model DICE-2016R2.8 One applica- tackling the various questions above. devoting output to emissions reduc- tion is to compare the economic and Small comprehensive models can yield tions, economies reduce consump- climate trajectories associated with dif- a full cost/benefit analysis, but are tion today but prevent economically ferent policy approaches. Here are four weak on regional or industrial detail. harmful climate change and thereby different policy options: Larger species provide great detail, but increase consumption possibilities in Baseline: No climate-change poli- may be unable to trace impacts and the future. cies are adopted. damages, are less transparent, and are The DICE model has 12 behav- Optimal: Climate-change policies unable to do full uncertainty analy- ioral equations, two variables to be maximize economic welfare, with full ses. Some models are able to trace the optimized, and several identities. participation by all nations starting in impacts of policies on land use. Others In the GAMS version, the simplest 2020. can investigate a wide range of tech- model has about 240 lines of opera- Temperature-limited: The optimal nologies. A few have full damage func- tional code. A run of 1,000 years takes policies are undertaken subject to a tions, while others include a limited five seconds, so it can be used for proj- further constraint that global tempera- number of technologies and engineer- ects with multiple states of the world ture does not exceed 2.5 °C above the ing variables. The great diversity of and Monte Carlo experiments. 1900 average. (The international goal the modeling ecosystem allows most The RICE model has the same of 2 °C is not feasible with current important questions to be addressed. basic economic and geophysical struc- DICE estimates without technologies ture, but contains a regional elabo- that allow negative emissions by mid- The DICE and RICE ration. The specification of prefer- 21st century.) Models as Examples ences in RICE is different because Stern discounting: These are results it must encompass multiple regions. associated with an extremely low dis- The Dynamic Integrated model The general preference function is count rate as advocated by The Stern of Climate and the Economy (DICE) a Bergson-Samuelson social welfare Review on the Economics of Climate and Regional Integrated model of function over regions. The model is Change.9

NBER Reporter • No. 3, September 2017 17 Figures 1–3 show outcomes associated with the four Projected CO Emissions in Dierent Scenarios policy options. A few key results emerge. First, emissions Global industrial emissions of CO (Gt / year) differ sharply, with major cuts in emissions in cases with 80 Baseline ambitious policies. Second, the temperature in the uncon- 70 trolled scenario continues to rise sharply over the current 60 century. 50 Figure 3 is the outcome that is generated uniquely by

40 IAMs: the carbon prices associated with each policy. This Optimal 30 is a concept that measures the marginal costs of reductions of emissions of greenhouse gases. In a market environment 20 Optimal with low Optimal with maximum temperature increase of 2.5 °C such as a cap-and-trade regime, the carbon prices would be 10 discount rate (Stern discounting) the trading price of carbon emission permits. In a carbon- 0 2010 2025 2040 2055 2070 2085 2100 tax regime, they would be the harmonized carbon tax among participating regions. Carbon prices in the baseline scenario Source: W.D. Nordhaus, NBER Working Paper No. 22933 are the current average prices in world markets, roughly $2 per ton of CO2. Prices under the optimal and temperature- Figure 1 limited scenarios at first rise to $35 and $229 per ton of CO2, respectively, by 2020. The carbon prices associated with the low-discounting scenario are close to those of the Temperature Change in Di erent Scenarios temperature-limited policy. The carbon price is closely linked to an important pol- Global mean temperature increase since 1900 (°C) icy instrument, the social cost of carbon, or SCC. This con- Baseline cept represents the economic cost of an additional ton of 4 carbon dioxide emissions (or, more succinctly, carbon) or its Optimal equivalent. The advantage of IAMs is that they can calculate 3 Optimal with maximum the shadow price of carbon emissions along a reference path temperature increase of 2.5 °C of output, emissions, and climate change. In an optimized 2 Optimal with low discount rate climate policy, abstracting away from distortions, the social (Stern discounting) cost of carbon will equal the carbon price or the carbon tax. 1 Estimates of the SCC are a critical ingredient in cli- mate-change policy. They provide policymakers a guidepost 0 2010 2025 2040 2055 2070 2085 2100 to aim for if they are seeking an economically efficient pol- icy for carbon pricing. Another application is for rulemak- Source: W.D. Nordhaus, NBER Working Paper No. 22933 ing where countries do not have comprehensive policies cov- Figure 2 ering all greenhouse gases. In this context, regulators might use the SCC in a calculation of social costs and benefits of policies involving energy or climate-affecting decisions. For example, the U.S. government has undertaken rulemaking Carbon Prices in Di erent Scenarios proceedings to determine the SCC for use in such areas as subsidies for the installation of low-carbon energy sources, Price per ton of CO (2010 U.S. dollars) 500 regulations requiring energy efficiency standards in build- ings and motor vehicles, and for power plants. Current regu- 400 lations using the SCC have more than $1 trillion in benefits, Optimal with low discount rate 10 (Stern discounting) according to the U.S. Environmental Protection Agency. 300 Estimates of the SCC vary by model and approach. Optimal with maximum temperature increase of 2.5 °C Table 1, on the following page, shows estimates from the 200 most recent DICE model. Two points emerge here. The Optimal 100 first is the critical importance of discounting. Looking at 2020, the SCC ranges from $22 to $133 per ton of CO2 as Baseline 0 the real discount rate ranges from 2.5 to 5 percent per year. 2010 2025 2040 2055 2070 2085 2100 Second, the SCC is extremely high — around $200 per ton Source: W.D. Nordhaus, NBER Working Paper No. 22933 of CO2 — for damage functions that would justify the tem- perature-limited objective (2 °C) that has been adopted at Figure 3 international meetings in Copenhagen and Paris.

18 NBER Reporter • No. 3, September 2017 economic impacts. tant feature of damage studies is that they Social Cost of Carbon (2010 U.S. Dollars / Ton of CO) Take the “simple” are generally limited to global temperature example of tropical increases of up to 3 °C, with the upper limits Scenario 2015 2020 2025 2030 2050 cyclones — hurricanes shown in Figure 2 not well-studied.

Base parameters in the United States. A difficulty in assessing IAMs is the Baseline 30.0 35.7 42.3 49.5 98.3 Basic physics indicates inability to use standard statistical tests Optimal controls 29.5 35.3 41.8 49.2 99.6 that a warmer ocean because of the lack of a probabilistic struc- is likely to increase ture. One concern, alluded to above, is 2.5 degree maximum the intensity of hurri- the major uncertainties associated with Maximum 184.1 229.0 284.0 351.0 1,008.4 canes; more precisely, the results of IAMs. The uncertainties are Max. for 50 years 147.2 183.2 227.2 280.4 615.6 the probability distri- particularly pronounced because of the bution of damage from long time periods — literally hundreds Stern Review discounting Uncalibrated 256.5 299.6 340.7 381.7 615.5 wind speed will shift to of years — required for estimates of opti- the right. IAMs need mal policies and the SCC. A recent study

Alternative discount rates to integrate this find- using the DICE-2016R2 model examined 2.5% 111.1 133.4 148.7 162.3 242.6 ing with economic the uncertainties of major outcomes from 3% 71.6 85.3 94.4 104.0 161.7 impacts. A major sur- parametric uncertainty of five major vari- 4% 34.0 39.6 44.5 49.8 82.1 5% 18.9 21.7 24.8 28.1 48.4 prise of research here ables: equilibrium temperature sensitivity, was that the economic productivity growth, the damage function, impacts of hurricanes the carbon cycle, and the rate of decarbon- Source: W.D. Nordhaus, NBER Working Paper No. 22933 are a high-power func- ization. Table 2, on the following page, tion of maximum wind shows the results. The best guess is the stan- Table 1 speed; normalized by dard DICE model where parameters are economic variables set at their expected values, while the oth- Calibration in IAMs such as local capital, damage is estimated ers are the distribution of outcomes. For to be wind speed to the ninth power.11 the important SCC calculation, the mean IAMs belong to a class of models, both But that is only a small slice of damage. value with full uncertainty is about 15 per- in economics and more generally in applied To construct damage functions, research- cent above the best guess. The change in sciences, that rely on calibration rather than ers need to aggregate across sectors, regions, temperature for 2100 is only slightly higher econometric estimation. Calibration involves levels of development, and climate change than the best guess. On the other hand, determination of system param- output is much higher because of eters and behavior using external Damage and Temperature Estimates from Major Studies the large estimated uncertainty of evidence rather than statistical productivity growth. Estimated damage (percentage of global output) systems estimation. Calibration 5 Another approach to esti- gained widespread use with the mating uncertainty, illustrative introduction of computable gen- 0 rather than statistical, is to exam- eral equilibrium models, of which ine model revisions. For this pur- -5 DICE and RICE are examples, pose, I looked at revisions of the and more recently with real busi- -10 DICE model over its quarter- ness cycle models. Outside eco- century history. The study found nomics, important examples of -15 that the major revisions have calibrated models are earth-sys- come primarily from economic -20 tem models such as the familiar 0 2 4 6 8 10 aspects of the model, whereas the climate models, transportation Temperature increase (°C) revisions to environmental mod- Area of the circles represents weights assigned to each study models, and engineering stud- Source: W.D. Nordhaus and A. Moƒat, NBER Working Paper No. 23646 ules have been much smaller. ies. Generally, while estimation is Particularly sharp revisions have desirable, calibration is necessary Figure 4 occurred for global output, dam- when the model reaches a certain complexity scenarios. This has proven the most difficult ages, and the social cost of carbon. These or when there are no relevant data to use for part of IAMs. In aggregate models, damage results indicate that the economic projec- estimation. Both of these are true for IAMs is often represented as a quadratic function tions are the least precise parts of IAMs of future climate change. of global mean temperature. Andrew Moffat and deserve much greater study than has As an example, one of the most con- and I recently did a systematic review of been the case up to now, especially careful troversial aspects of IAMs is the damage damage estimates; Figure 4 summarizes the studies of economic growth prospects to function, which relates climate change to findings of different studies.12 One impor- 2100 and beyond.13

NBER Reporter • No. 3, September 2017 19 DICE_Manual_100413r1. Uncertainty Analysis in Baseline Scenario pdf

DICE 50th Standard Interquartile Coef. of Models available at Variable (All except SCC refer to the year 2100) Mean best guess percentile deviation range variation www.econ.yale. edu/~nordhaus/ Social cost of carbon, 2015 (2010 U.S. dollars) 34.5 30.0 25.3 32.5 39.3 0.94 homepage/ Temperature increase from 1990 (°C) 4.22 4.10 4.08 1.12 2.01 0.22 DICEmodels09302016.

Carbon concentrations (ppm) 969.6 826.6 829.8 413.3 488 0.12 htm. Return to text World output (trillions 2010 U.S. dollars) 1,433 759 766 1,660 1,056 0.67 8 W. D. Nordhaus, CO emissions (Gt / year) 109.4 70.9 71.1 105.2 114.5 0.63 “Projections and Damages (percent of output) 4.3% 3.8% 3.2% 3.7% 4.9% 0.67 Uncertainties about Climate Change in an Era of Real interest rate (percent / year) 3.6% 3.6% 3.6% 1.6% 2.2% 0.33 Minimal Climate Policies,” NBER Working Paper No. Source: W.D. Nordhaus, NBER Working Paper No. 22933 22933, December 2016, revised Table 2 September 2017. Conclusion 2 Report of the Modeling Resource Group, Return to text Synthesis Panel of the Committee on Nuclear 9 The Economics of Climate Change: IAMs are important tools for under- and Alternative Energy Systems, National The Stern Review, Cambridge, England: standing the implications and policy aspects Research Council, Energy Modeling for an Cambridge University Press, 2007; W. of climate change. They have fundamentally Uncertain Future, Washington, D.C.: The D. Nordhaus, “The ‘Stern Review’ on the transformed the way economists and envi- National Academies Press, 1978. Economics of Climate Change,” NBER ronmentalists approach climate policy, shift- Return to text Working Paper No. 12741, December ing from a pure engineering approach — “do 3 P. A. Samuelson, “Market Mechanisms 2006; and “A Review of the Stern Review this and don’t do that” — to approaches like and Maximization,” Santa Monica, CA: on the Economics of Climate Change,” cap-and-trade or carbon taxes that empha- Rand Corporation, P-69, March 28, 1949. Journal of Economic Literature, 45, size market mechanisms. Return to text September 2007, pp. 686–702. As hurricanes Harvey and Irma remind 4 W. D. Nordhaus, “Can We Control Return to text us, the impacts of weather events can be Carbon Dioxide?” International Institute 10 W. D. Nordhaus, “Revisiting the Social extremely large. So climate change is likely for Applied Systems Analysis Working Paper, Cost of Carbon,” Proceedings of the National to continue growing as an economic prob- WP 75-63, Laxenburg, Austria, June 1975; Academy of Sciences, 114(7), 2017, pp. lem. Improving integrated assessment mod- W. D. Nordhaus, “Economic Growth and 1518–23. els is therefore an important research area for Climate: The Case of Carbon Dioxide,” Return to text economists — full of puzzles, challenges, and American Economic Review, 67(1), 1977, 11 W. D. Nordhaus, “The Economics of policy applications. pp. 341–6. Hurricanes in the United States,” NBER Return to text Working Paper No. 12813, December 2006, 1 This research report draws upon W. 5 A. S. Manne, “ETA: A Model for Energy and Climate Change Economics, 2010, 1 D. Nordhaus, The Climate Casino: Technology Assessment,” Bell Journal of (1), pp. 1–20. Risk, Uncertainty, and Economics for a Economics, 7(2), 1976, pp. 379-406. Return to text Warming World, New Haven, CT: Yale Return to text 12 W. D. Nordhaus and A. Moffat, “A University Press, 2013; W. D. Nordhaus, 6 W. D. Nordhaus, “An Optimal Transition Survey of Global Impacts of Climate “Integrated Economic and Climate Path for Controlling Greenhouse Gases,” Change: Replication, Survey Methods, Modeling,” in Handbook of Computable Science, 258, 1992, pp. 1315–9; W. D. and a Statistical Analysis,” NBER General Equilibrium Modeling, P. Dixon Nordhaus, Managing the Global Working Paper No. 23646, August 2017. and D. Jorgenson, eds.: Elsevier, 2013, pp. Commons: The Economics of Climate Return to text 1069–131; W. D. Nordhaus, “Projections Change, Cambridge, MA: MIT Press, 1994. 13 W. D. Nordhaus, “Evolution of and Uncertainties About Climate Change in Return to text Assessments of the Economics of Global an Era of Minimal Climate Policies,” NBER 7 W. D. Nordhaus, P. Sztorc, “DICE Warming: Changes in the DICE Model, Working Paper No. 22933, December 2016, 2013R: Introduction and User’s Manual,” 1992–2017,” NBER Working Paper No. revised September 2017. October 2013. http://www.econ.yale. 23319, April 2017. Return to text edu/%7Enordhaus/homepage/documents/ Return to text

20 NBER Reporter • No. 3, September 2017 Resource Barriers to Postsecondary Educational Attainment

Michael Lovenheim

U.S. economic growth in recent enrollment. Supply-side resource barri- decades has favored high-skilled, service- ers are driven by declining state subsi- based occupations and industries. As a dies for public higher education, as well result, the demand for skilled relative as the higher propensity of lower-income to unskilled labor has grown markedly, students to attend universities with lower which has been the source of much atten- per-student resources. tion and concern among policymakers In a series of research papers, my co- Michael Lovenheim is a and researchers. Increasingly, the labor authors and I have examined how fam- research associate in the NBER market outcomes of working-age adults ily financial resources and postsecondary Economics of Education and are linked to their educational attain- institutional resources affect collegiate Public Economics Programs. He is ment. Earnings gains have flowed dispro- attainment. We estimate resource effects an associate professor in the depart- portionately to those with four-year col- on both the demand and supply sides of ment of policy analysis and man- lege degrees. One might expect that this the higher education market and provide agement at Cornell University growth in the demand for skilled labor insight into policies that could reduce and is co-editor of the Journal of would be met with a substantial increase barriers to college completion. Policy Analysis and Management. in the production of such labor, but this Policymakers and researchers have Lovenheim earned his B.A. in eco- has not been the case. focused a significant amount of atten- nomics from Amherst College in The anemic response of collegiate tion on college access, with the goal 2000 and received his Ph.D. in attainment to persistent increases in of increasing college enrollment either economics from the University labor market returns has occurred along- overall or for specific groups. Much of of Michigan in 2007. He was a side rising inequality in postsecond- my research is motivated by the widen- Searle Freedom Trust Postdoctoral ary outcomes.1 Although education is ing gap between enrollment and degree Fellow at the Stanford Institute for often discussed as a means to reduce attainment: A large component of both Economic Research for two years economic inequality and induce upward the increased inequality in postsecond- prior to joining the Cornell faculty. social mobility, large and growing attain- ary attainment and the sluggish increase Lovenheim’s research focuses ment gaps among students from differ- in postsecondary attainment overall is on empirical, policy-relevant issues ent socioeconomic backgrounds, coupled degree non-completion. Simply put, if in the economics of education and with high labor market returns to post- most of the students who enroll in col- public finance. In higher education, secondary education, have led to con- lege were to successfully obtain a degree, his research examines questions cerns that the higher education system is postsecondary attainment would rise dra- related to the effect of personal and exacerbating inequality. matically and inequality in attainment institutional resources on student The fact that the supply of college- would decline. Increasing the supply and outcomes and institutional produc- educated workers has not kept up with altering the composition of college-edu- tivity. Lovenheim also has an active demand along with growing inequality in cated workers thus requires understand- research agenda on teacher labor postsecondary outcomes suggests there ing barriers to completion among the markets, focusing specifically on are barriers precluding many students students who enroll, as well as under- teachers unions, performance pay, from obtaining a postsecondary degree. standing barriers to enrollment. and teacher pension policies. His A particularly important class of barri- My collaborators and I have exam- research on public economics stud- ers, especially for low-income students, ined how financial resource barriers ies questions related to interstate centers around financial resources. Such can affect dimensions of postsecondary cigarette and alcohol regulation as barriers can occur on the demand (i.e., investment behavior beyond enrollment, well as the nutritional effects of student) or supply (i.e., institutional) such as what types of colleges students consumption taxes. In addition to side of the higher education market. attend and whether students complete his academic research, he recently Demand-side resource constraints mostly college conditional on attending. We co-authored, with Sarah Turner, consist of difficulties in paying the often look beyond access to the various dimen- a textbook on the economics of high tuition price associated with college sions along which financial resources can education.

NBER Reporter • No. 3, September 2017 21 influence higher education attainment. ing and magnitude of the urban hous- with incomes between $70,000 and This summary describes our recent ing boom between the late 1990s and $125,000, and families with incomes research findings and discusses their mid-2000s, across cities, to generate above $125,000. I present the effect policy implications. such variation.2 This period saw an of enrollment from a $10,000 home unprecedented growth in the value of equity increase relative to the mean Demand-Side Resource Effects housing as well as in the liquidity of college enrollment rate for each group, housing wealth; it became much eas- as well as the effect of the mean home A popular view among parents, ier to extract equity from the home equity increase experienced by each policymakers, and the media is that the through home equity loans, lines of group between 2001 and 2005, the cost of college presents a substantial credit, and cash-out refinances. Home heart of the housing boom. For all fam- impediment to postsecondary invest- price increases varied considerably ilies, enrollment increases by a statis- ment for many students. While col- across cities, with some such as Las tically significant 1.4 percent for each lege tuition and fees have indeed risen Vegas and Miami experiencing enor- $10,000 of additional home equity. precipitously over time, so has finan- mous increases over a short period of During the housing boom of the early cial aid. The United States has one of time, while others experienced rela- 2000s, the average homeowner expe- the most generous financial aid systems tively modest growth. The idea under- rienced an almost $58,000 increase in in the world, especially for very low- lying my approach is to consider high home equity, which my estimates indi- income students. The goal of this sys- school seniors in the same year whose cate increased college enrollment by 7.9 tem is to decouple students’ financial parents own a home in cities that expe- percent relative to the baseline level. background from their ability to invest rienced different recent housing price Students in families with earn- in a postsecondary degree. Finding growth. Families in high-increase cities ings under $70,000 per year are par- that students’ college choices are caus- received a financial windfall just before ticularly responsive to home equity ally linked to their family’s financial their children made college choices, changes: $10,000 of additional home resources is evidence that the equity increases college enroll- current financial aid system is Home Equity, Family Income and College Enrollment ment by 13.7 percent relative not sufficient to achieve this to the mean level. When mul- goal. Increase in college enrollment rate Enrollment rate increase due to home tiplied by the average home due to a $10,000 increase in home equity equity gains during the early 2000s Prior research has strug- housing boom equity increase experienced by gled to obtain credible esti- 25% this group in the early 2000s, mates of the causal effect 20 the effect is 21.4 percent. of family financial resource 15 Among students from higher- variation on postsecond- income households, enroll- ary attainment. Estimating 10 ment responds more modestly such an effect is challenging 5 to housing wealth changes. For because income and wealth both higher income groups, 0 are not randomly assigned < $70 $70–125 > $125 Family Income < $70 $70–125 > $125 the effect of housing wealth is across students: Families with (thousands) much smaller and is not sta- lower resources at the time of tistically significantly differ- Source: Researcher’s calculations using data from the National Longitudinal Survey of their children’s college entry Youth 1997 and the Conventional Mortgage Housing Price Index ent from zero. However, the decision typically had fewer point estimates are positive resources throughout the chil- Figure 1 and are sizable in magnitude dren’s lives to invest in their education. while families in lower-increase cit- when multiplied by the large increases The result is that students from lower- ies experienced a much more mod- in home equity experienced by these resource households tend to be, on est change in resources. I therefore families during the housing boom. The average, less academically prepared for leverage the timing, magnitude, and fact that students from higher income college than their counterparts from geographic dispersion of the housing households are less affected by housing more affluent backgrounds. boom to generate variation in house- wealth changes is likely because these What is needed is a source of fam- hold resources that are unrelated to the students face fewer resource constraints ily resource variation unrelated to the underlying characteristics of students. in financing a college education than myriad attributes of students that are I found college enrollment was their less affluent counterparts. I also correlated with the costs and bene- responsive to housing wealth during show that the housing wealth-enroll- fits of attending college, such as moti- the housing boom. Figure 1 shows the ment relationship was not present prior vation and academic achievement. I results graphically for families with to the housing boom, which suggests an have exploited differences in the tim- incomes below $70,000, families important role for the increased liquid-

22 NBER Reporter • No. 3, September 2017 ity of home equity in the early 2000s. tion rate of bachelor of arts degrees, a have higher per-student expenditures, In a follow-up paper, C. Lockwood longer time required by college com- higher-achieving student bodies, and Reynolds and I use the same source of pleters to complete a B.A., and an higher-paid and more research-pro- variation to examine how housing increased likelihood of a subsequent ductive faculties. The result is that wealth impacts the type of schools criminal conviction. Excluding these resources are increasingly being con- students choose, and college comple- students from financial aid eligibility centrated in a small set of “elite” insti- tion.3 We find that when families negatively affects their life outcomes tutions that serve students with high experience more home price growth and produces substantial social costs. precollegiate achievement levels. A when their child is in high school, Sarah Cohodes, Daniel Grossman, growing body of research in econom- their child is more likely to attend a Samuel Kleiner, and I examine another ics seeks to estimate the labor mar- state flagship university and is less source of household resource varia- ket return to enrolling in one of these likely to attend a community col- tion: access to Medicaid. This occurs highly selective schools, which is dif- lege. Interestingly, the flagship effect earlier in life than the resources I ficult because students with higher is driven by increased applications, examined in my other research.5 earning potential select into these which suggests that changes in family Medicaid is the primary means higher quality institutions. resources impact the types of schools through which lower-income children Rodney Andrews, Jing Li, and students consider attending. Low- receive health insurance, which can I contribute to this literature using income students whose families expe- improve their health and their par- administrative data on all public rienced home price increases during ents’ financial standing. This resource school students in the state.6 We link the housing boom were more likely to variation is different from those pre- educational records for all public K-12 complete a four-year degree as well. viously discussed because it does not students in Texas to postsecondary Another way to test for household just impact the ability to pay for col- records for all public higher education resource effects is to study variation in lege. Instead, it can affect the level students in the state, and merge these the amount of financial aid available and productivity of early childhood data with quarterly earnings records. to students. This has proved difficult. investments in education. We examine Linked administrative data are becom- Because most financial aid is federal, the large Medicaid eligibility expan- ing more prevalent in education eco- there is little variation in aid eligibil- sions experienced by those born from nomics research; they provide both a ity across students that is not directly 1980 through 1990. Using the fact wealth of information about students tied to their family finances and back- that children born in different states over time as well as large sample sizes. ground. Emily Owens and I studied and years had very different eligibil- We use pre-collegiate demographic an unusual policy change enacted by ity for Medicaid over the course of and academic achievement informa- the federal government in 2001 that their childhood due to state and fed- tion to account for student selection. excluded anyone with a drug convic- eral Medicaid law changes, we esti- Our findings indicate that graduating tion from receiving federal financial mate how Medicaid eligibility trans- from the University of Texas at Austin aid.4 While a small group, students lates into educational attainment later or Texas A&M University, the flagship with drug convictions tend to be from in life. We find that a 10 percentage universities in Texas, increases earn- more disadvantaged backgrounds, point increase in average Medicaid ings by 12 and 21 percent, respectively, and there may be particularly large eligibility during childhood decreases relative to graduating from a non- social returns to increasing their edu- the high school dropout rate by 4 flagship public university. Graduating cational attainment. We compare the percent and increases the likelihood from a community college is associ- change in college enrollment among of B.A. completion by 2.5 percent. ated with lower earnings by 11 percent those with a drug conviction when the These results suggest that policies tar- relative to obtaining a degree from a rule was implemented to the change geting resources to low-income fam- non-flagship public university. among those with no conviction. Our ilies with young children can have We also examine how college qual- findings indicate that college enroll- sizable effects on their ultimate colle- ity affects the distribution of earnings. ment within one year of high school giate attainment. Going beyond mean earnings effects graduation dropped by 22 percent is important, because the average may among those with a drug conviction Supply-Side Resource Effects mask a large amount of variability in relative to those without, which sug- labor market returns across the earn- gests financial resources are a relevant One reason studying postsecond- ings distribution. We estimate quan- barrier to postsecondary investment ary institutional resources is impor- tile treatment effects of college sector for many families. We also present evi- tant is the high degree of resource on earnings; the results are presented dence that the reduction in financial stratification within the higher educa- in Figure 2, on the following page. aid leads to a reduction in the comple- tion sector. More selective institutions These curves show the differences in

NBER Reporter • No. 3, September 2017 23 resources on colle- all of the observed decline in college Flagship Universities, Community Colleges, and Earnings giate attainment is completion rates. Men’s earnings dierence relative to graduating that a large amount In a follow-up paper, we conduct from a non-flagship public university in Texas 50% of money is spent a similar decomposition analysis with 40 by federal and state respect to lengthening time to degree Texas A&M 8 30 governments to among B.A. recipients over time. 20 subsidize higher While the supply-side effects are not 10 UT-Austin education. For pub- as strong, we find reductions in per- Non-flagship baseline 0 lic institutions, student resources in the less selective -10 Community Colleges state appropria- public four-year sector to be a core con- -20 tions are a particu- tributor to the longer time it is taking -30 larly important part students to complete B.A. degrees. 0 25 50 75 100 of the budget, and Earnings distribution percentile they have declined Comprehensive Interventions Source: Researcher’s calculations using data from the Texas Education Agency, substantially over the Texas Higher Education Coordinating Board, and the Texas Workforce Commission time. John Bound, Students from low-income back- Figure 2 Sarah Turner, and grounds face several barriers to post- I examine whether secondary success, including difficulty earnings, adjusted for observed stu- changes in supply- in financing postsecondary enroll- dent characteristics, between gradu- side resources contribute to declining ment, lack of information about the ates in the given sector and those in completion rates over time.7 Between postsecondary system that leads to less the non-flagship four-year sector at the mid-1970s and mid-1990s, college enrollment and enrollment in lower- each percentile of the earnings distri- completion rates conditional on ever quality colleges, and lower pre-colle- bution. For UT-Austin graduates, the having attended college dropped from giate academic achievement. There has mean effect of 12 percent does a poor 52 to 43 percent. The largest declines been a policy trend toward attempt- job of characterizing the effect on the were experienced by students attend- ing to address these multiple dimen- entire distribution. At the bottom of ing non-top-50 ranked public four-year sions of disadvantage that low-income the distribution, earnings returns to schools and community colleges. students face using comprehensive UT-Austin are quite low, and then Supply-side forces can play two interventions. Examples of such pro- they grow to more than 30 percent at roles in explaining this decline. First, as grams are the Longhorn Opportunity the top of the distribution. The effects more students enter college over time, Scholarship (LOS) in Texas, the Susan are much more constant among Texas an increasing proportion sort into less Thompson Buffett Foundation (STBF) A&M graduates, however. We argue selective and less resourced schools scholarship in Nebraska, and the ASAP the differences across the two flagship because these are the institutions that program at the City University of New universities are likely due to differ- expand their enrollment due to higher York . ences in field of study, as Texas A&M student demand. Second, per-student Andrews, Scott Imberman, and students are much more likely to major resources at the less selective institu- I study the LOS program in Texas in high-earning, low-variance fields tions have declined due to reductions using the linked administrative data such as engineering. Finally, we exam- in state appropriations, as these schools discussed previously.9 The LOS pro- ine community colleges and show that are particularly reliant on state fund- gram is run by the UT-Austin and con- the earnings penalty to a community ing. We conduct a decomposition anal- sists of recruiting students at urban, college relative to a non-flagship pub- ysis that shows how college completion low-income, and heavily minority high lic university is driven by low earn- rates would have changed had institu- schools, offering grant aid if they enroll ers. At the top of the earnings distri- tional resources (proxied by student- at UT-Austin, and providing a series of bution, community college graduates faculty ratios) and the distribution of academic support services once they are earn the same as their non-flagship students across postsecondary sectors enrolled. This program thus combines four-year counterparts. This is despite not changed over time. We find that demand-side and supply-side resource the fact that the community college the increase in student-faculty ratios supports. We find that among high- degree requires two fewer years of can explain about a quarter of the com- achieving students who were the targets study; for a portion of students, the pletion rate decline, while the rest can of this program, the LOS intervention payoff to community college enroll- be explained by students increasingly substantially increased the likelihood ment is relatively high. attending lower-quality colleges and that students both enrolled at and A second reason economists are universities. Thus, we argue that sup- graduated from UT-Austin. Among interested in the effect of supply-side ply-side resource changes can explain students from targeted high schools

24 NBER Reporter • No. 3, September 2017 who attended UT-Austin, earnings Reynolds, “The Effect of Housing 18068, May 2012, and “Q uantile increased by 82 percent 12 or more Wealth on College Choice: Evidence Treatment Effects of College Q ual- years after high school relative to simi- from the Housing Boom”, NBER ity on Earnings,” Journal of Human lar students who were not exposed to Working Paper No. 18075, May 2012, Resources, 51(1), 2016, pp. 200–38. this program. These results show that and Journal of Human Resources, Return to text combining supply-side and demand- 48(1), 2013, pp. 1–35. 7 J. Bound, M. F. Lovenheim, and side resource increases for disadvan- Return to text S. Turner, “Why Have College taged students can be particularly effec- 4 M. F. Lovenheim and E. G. Completion Rates Declined? An tive in supporting their postsecondary Owens, “Does Federal Financial Analysis of Changing Student attainment and future earnings. Aid Affect College Enrollment? Preparation and Collegiate Resources,” Evidence from Drug Offenders and NBER Working Paper No. 15566, the Higher Education Act of 1998,” December 2009, and American 1 M. J. Bailey and S. Dynarski, “Gains NBER Working Paper No. 18749, Economic Journal: Applied and Gaps: Changing Inequality in February 2013, and Journal of Urban Economics, 2(3), 2010, pp. 129–57. U.S. College Entry and Completion,” Economics, 81, 2014, pp. 1–13. Return to text NBER Working Paper No. 17633, Return to text 8 J. Bound, M. F. Lovenheim, December 2011, and published as 5 S. Cohodes, D. Grossman, S. Kleiner, and S. Turner, “Increasing Time to “Inequality in Postsecondary Education” and M. F. Lovenheim, “The Effect Baccalaureate Degree in the United in G. J. Duncan and R. J. Murnane, of Child Health Insurance Access States,” NBER Working Paper No. eds., Whither Opportunity? Rising on Schooling: Evidence from Public 15892, April 2010, and Education Inequality, Schools, and Children’s Life Insurance Expansions,” NBER Working Finance and Policy, 7(4), 2012, pp. Chances, New York, NY: Russell Sage, Paper No. 20178, May 2014, and 375–424. 2011, pp. 117–32. Journal of Human Resources, 51(3), Return to text Return to text 2016, pp. 727–59. 9 R. J. Andrews, S. A. Imberman, 2 M. F. Lovenheim, “The Effect of Return to text and M. F. Lovenheim, “Recruiting Liquid Housing Wealth on College 6 R. J. Andrews, J. Li, and M .F. and Supporting Low-Income, High- Enrollment,” Journal of Labor Lovenheim, “Q uantile Treatment Achieving Students at Flagship Economics, 29(4), 2011, pp. 741–71. Effects of College Q uality on Earnings: Universities,” NBER Working Paper Return to text Evidence from Administrative Data No. 22260, May 2016. 3 M. F. Lovenheim and C. L. in Texas,” NBER Working Paper No. Return to text

NBER Reporter • No. 3, September 2017 25 Economic Development and Gender Equality: Exceptions to the Rule

Seema Jayachandran

Women lag further behind men in couples’ fertility decisions and their invest- poor countries than in rich countries in ments in their children, with important terms of educational outcomes, health sta- implications for child health. It creates not tus, decision-making power in the family, only inequality between boys and girls, but and other aspects of well-being. As a rule, also among boys and among girls, and it the process of economic development lev- can reduce average child health outcomes. els the playing field for women in the labor Rohini Pande and I found that eldest Seema Jayachandran is an market. As an economy grows, its service son preference helps explain India’s unex- associate professor of economics sector expands, while agriculture and other pectedly high rate of stunting among chil- at Northwestern University and primary industries shrink in importance; dren under age five.3 Height is a com- a research associate in the NBER women have a comparative advantage in monly used measure of child well-being, Development Economics and jobs that require brains rather than brawn. not because we care about height per se, Health Care Programs. Similarly, economic development brings but because height reflects the nutrition Jayachandran’s recent work about — and is often enabled by — lower and illness a child has experienced during focuses on gender equality in devel- lifetime fertility for women, which frees up the critical early years of life. Childhood oping countries. She also is currently their time to invest in their careers. Better stunting has been linked to low cognitive working on projects related to envi- career prospects for women, in turn, lead ability, poor health, and low earnings later ronmental conservation, health, parents to invest more in their daughters’ in life. India is a poor country, so it is unsur- labor markets, and education. educations, and economically empowered prising that stunting is more prevalent than Jayachandran co-chairs the women have more say in their households. in a rich country such as the United States, health sector of the Abdul Latif I have recently reviewed the large literature but researchers have been puzzled by the Jameel Poverty Action Lab and on how economic progress brings about high level of stunting in India compared serves on the board of the Bureau for progress for women.1 to other countries at a similar level of eco- Research and Analysis of Economic However, there are also exceptions to nomic development. Pande and I compare Development. She is a reviewing the rule — ways in which cultural norms India to a set of sub-Saharan African coun- editor at Science and an associate about gender stubbornly persist in the face tries which, though they are poorer than editor for the American Economic of economic progress.2 My recent research, India, have lower levels of stunting. Review, the Q uarterly Journal focusing on India, has examined situations We find that child height varies con- of Economics, and the Journal of in which movement toward gender equity siderably among siblings, and specifically Economic Perspectives. has leveled off or regressed. Specifically, my that there is a strong drop-off with birth Jayachandran received her colleagues and I have examined negative order. This birth order gradient is observed Ph.D. in economics from Harvard ramifications of parents’ favoritism toward in almost all societies, but it is especially University in 2004. She also holds their eldest son and obstacles to women’s strong in India. Firstborn Indian chil- an M.A. in physics from Harvard, success in the labor market. This summary dren are no shorter than their sub-Saha- an M.A. in physics and philoso- describes some of our findings. ran African counterparts. The India height phy from the University of Oxford, puzzle is mostly concentrated among later- and an S.B. in electrical engineer- Implications of Parents’ born children, as shown in Figure 1, on the ing from MIT. Prior to joining Strong Desire to Have a Son following page. Northwestern, she was on the fac- Firstborn children’s genes do not differ ulty at from In India, older couples live with their systematically from the genes of younger 2006–11 and a Robert Wood eldest son, and the eldest son gets priority siblings. Infrastructure for clean water and Johnson Scholar in Health Policy for inheritance. He also plays important sanitation also do not vary much among Research at the University of roles in Hindu funeral rites. Having a son siblings; if anything, they improve over California, Berkeley from 2004– to fulfill these roles is therefore very impor- time, which would help later-born chil- 06. She is a recipient of an NSF tant to couples, and they tend to favor dren. Disparities within families suggest CAREER Award and a Sloan their eldest son over their other children. that parents’ choosing to provide more Research Fellowship. This eldest son preference strongly shapes inputs for some children than for others

26 NBER Reporter • No. 3, September 2017 explains a lot of the height puzzle. We and I studied another way that Indian increased access to ultrasound tests and find that health inputs such as vaccina- parents’ desire to have a son as quickly other technology that can determine fetal tions and prenatal visits also exhibit an as possible has negative spillovers, spe- sex. My research has highlighted another unusually steep drop-off with contributing factor: the move birth order in India. toward smaller families. Rapidly Height for Age and Birth Order, Ages 0–5 We also investigate the role changing norms about family of eldest son preference, and Normalized di­erence between mean child height and WHO growth standard size in India have collided with find that it is at the root of -1 the persistent cultural impor- India’s steep birth order gradi- tance of eldest sons.6 It is not ent. If parents dote on the first- -1.2 the case that parents in India born son, there will be a steep want all their children to be -1.4 birth-order gradient among Africa India boys; if they were to have two boys. How favoritism for eldest -1.6 children, most would ideally sons hurts later-born girls rela- want one son and one daugh- tive to their older sisters is more -1.8 ter. But the exalted role of the subtle. First, an earlier-born girl eldest son means that having at is less likely to have a brother yet, -2 least one son is very important. and so is less likely to be com- First born Second born Third born Couples who want four or five peting with a boy for resources children are very likely to end Source: Researcher’s calculations using data from the Demographic and when she is young. We find Health Surveys Program and the World Health Organization up with a son naturally (94 and much less inequality among sib- 97 percent, respectively), but at lings in regions of India where Figure 1 a family size of two — which son preference is weaker, such is increasingly the desired fam- as the matrilineal state of Kerala, and for cifically for girls.4 When parents with- ily size in India — a quarter of couples Muslims, who do not have strong eldest out a son give birth to a girl, their eager- would end up with two daughters if they son preferences, compared to Hindus. ness to “try again” for a son reduces the did not intervene, and thus many opt for A second key channel is parents decid- duration that their daughter is breastfed. sex-selective abortions. Using a novel way ing to have more children than originally Some mothers know that breastfeeding of eliciting fertility preferences, I show planned in order to have a son. Consider a suppresses fecundity and therefore wean that the growing desire for small families couple who desire two children; they will their newborn girl quickly. Others might explains a third to a half of India’s worsen- likely view the birth of a second daugh- simply become pregnant again — nurs- ing sex imbalance in recent decades. ter as an unpleasant surprise and decide ing does not make a woman completely These findings also highlight a para- to keep trying for a son. They will need to infecund — which triggers them to stop dox. Because women might value daugh- start economizing because now they will breastfeeding. In contexts where water ters more than men do, and because have more children to support than they and food are unsanitary, a shortened dura- female education strengthens women’s had planned, and those spending cutbacks tion of breastfeeding can be harmful to say in household decisions, we might hit the second-born daughter in her criti- child health. We find that this behavior is expect that educating women would cal early years. This behavior generates an important contributor to girls’ mortal- reduce the prevalence of gender selec- some quite specific predictions, which are ity in India. This harm to girls is an unin- tion. However, there is also another borne out in the data, about how parents’ tended consequence rather than a con- important force: Educated women investments depend on the composition scious choice resulting from a decision to typically want fewer children, which of the sibling group. invest more in sons than daughters. It nev- increases their “need” to resort to gen- Finally, we show that within-family ertheless disadvantages girls. der selection. Empowering women will inequality lowers average child health out- not necessarily eliminate the problem comes in India. With diminishing returns The Worsening Problem of “missing women.” This conclusion to investment, concentrating investments of ‘Missing Women’ raises the question of how, if at all, pol- on a favored child hurts aggregate out- icy might address the challenge of “miss- comes. Parents might be maximizing their Perhaps the most extreme conse- ing women.” In current work, Diva Dhar, own well-being by favoring one child, but quence of the desire for sons is selective Tarun Jain, and I are evaluating whether given potential positive externalities of abortion of female fetuses. Amartya Sen incorporating classroom discussions of a healthier and more skilled workforce, called attention to this problem of “miss- gender norms into the government sec- their favoritism could be inefficient, and a ing women” in 1990, and it has worsened ondary school curriculum makes adoles- drag on India’s economic progress. in India, as well as in China and elsewhere, cents less tolerant of gender discrimina- In an earlier paper, Ilyana Kuziemko in the three decades since.5 One reason is tion and begins to reshape those norms.7

NBER Reporter • No. 3, September 2017 27 Female Employment norm promoted entrepreneurship among 5 A. Sen, “More Than 100 Million and Friends’ Support women. Women Are Missing,” The New York Review of Books, 37(20), 1990, pp. Another way in which India’s cultural 61–6. norms might be constraining its growth 1 S. Jayachandran, “The Roots of Return to text relates to female employment. The female Gender Inequality in Developing 6 S. Jayachandran, “Fertility Decline labor force participation rate has been ris- Countries,” NBER Working Paper and Missing Women,” NBER Working ing in most developing countries recently, No. 20380, August 2014, and Annual Paper No. 20272, July 2014, and but in India it has fallen over the past Review of Economics, 7, 2015, pp. American Economic Journal: Applied decade.8 Social restrictions on women’s 63–88. Economics, 9(1), 2017, pp. 118–39. ability to interact with men outside their Return to text Return to text family or to travel unaccompanied within 2 There is a growing literature that dis- 7 D. Dhar, T. Jain, and S. their village or city are barriers to women’s cusses how cultural norms are shaped by Jayachandran, “Intergenerational employment. Many women turn to home- economic forces, then linger even after the Transmission of Gender Attitudes: based self-employment, such as tailor- economic environment has transformed. Evidence from India,” NBER Working ing clothes or making incense sticks, but See, for example, A. F. Alesina, P. Paper No. 21429, July 2015. the same social restrictions can hurt the Giuliano, N. Nunn, “On the Origins of Return to text success of these small businesses, in part Gender Roles: Women and the Plough,” 8 R. Heath and S. Jayachandran, “The because the women have limited oppor- NBER Working Paper No. 17098, Causes and Consequences of Increased tunity to network with and learn from May 2011, and Quarterly Journal of Female Education and Labor Force peers. Erica Field, Pande, Natalia Rigol, Economics, 128(2), 2013, pp. 469–530. Participation in Developing Countries,” and I evaluated a popular type of program Return to text NBER Working Paper No. 22766, to help small-scale entrepreneurs, namely 3 S. Jayachandran and R. Pande, “Why October 2016, and forthcoming in S. business skills training.9 In one variant of Are Indian Children So Short?,” NBER Averett, L. Argys, and S. Hoffman, eds., the program, we added the feature that Working Paper No. 21036, March Oxford Handbook on the Economics women could bring a friend with them to 2015, and American Economic Review, of Women, New York, NY: Oxford the classes. Program participants took out 107(9), 2017, pp. 2600–29. University Press, 2018. more business loans and earned higher Return to text Return to text incomes, but only in the variant where 4 S. Jayachandran and I. Kuziemko, 9 E. Field, S. Jayachandran, R. Pande, they could bring a friend. We cannot say “Why Do Mothers Breastfeed Girls Less and N. Rigol, “Friendship at Work: for sure what led to this result, but having Than Boys? Evidence and Implications Can Peer Effects Catalyze Female a friend’s support seemed to give women for Child Health in India,” NBER Entrepreneurship?,” NBER Working the wherewithal to set more ambitious Working Paper No. 15041, June 2009, Paper No. 21093, April 2015, and business goals and achieve them. In this and Quarterly Journal of Economics, American Economic Journal: Economic case, without reshaping the cultural norm, 126(3), 2011, pp. 1485–538. Policy, 8(2), 2016, pp. 125–53. redesigning a program in light of the Return to text Return to text

28 NBER Reporter • No. 3, September 2017 NBER News

Karen Horn Elected Chair of NBER Board of Directors, John Lipsky Elected Vice Chair

Karen N. Horn was elected chair of the NBER’s Ph.D. from Johns Hopkins University and was Board of Directors at the board’s September 25 elected to the NBER Board of Directors in 1993. meeting. She succeeds Martin Zimmerman, a pro- The board also elected director John Lipsky, Karen Horn fessor of business administration at the University a senior fellow at Johns Hopkins’ Paul H. Nitze of Michigan’s Ross School of Business and former School of Advanced International Studies, as vice vice president for corporate affairs at Ford Motor chair. Lipsky, who received his Ph.D. at Stanford Company, who had served since 2014. Horn, a and joined the NBER board in 1998, served as partner in the Brock Capital Group, is a former first deputy managing director of the International chair and CEO of Bank One and a former presi- Monetary Fund between 2006 and 2011. Prior to dent of the Federal Reserve Bank of Cleveland. his IMF service, he was vice chair of the JPMorgan She was also managing director and president of Investment Bank. Lipsky is co-chair of the Aspen Global Private Client Services at the Marsh sub- Institute’s Program on the World Economy and a sidiary of Marsh and McLennan. She received her life member of the Council on Foreign Relations. John Lipsky

Four New Directors Join NBER Board

Philip Hoffman, the Rea A. and Lela G. Axline of Pennsylvania. His research focuses on micro- Professor of Business Economics and History at the economic theory, particularly on pricing, repeated California Institute of Technology, is the new rep- games, non-cooperative and evolutionary game resentative of the Economic History Association theory, social norms, and the theory of repu- (EHA). His research combines economic the- tations. Mailath is a Fellow of the American ory and historical evidence to explain long-term Academy of Arts & Sciences, the Econometric changes in politics, society, and the economy, Society, the Game Theory Society, and the Society Philip Hoffman with particular attention to economic growth and for the Advancement of Economic Theory. He political development. He is currently studying has served on the Councils of the Econometric cross-country differences in long-term patterns of Society and the Game Theory Society. Mailath economic development, the evolution of financial received his undergraduate degree from ANU and institutions, and the way states develop the capac- his Ph.D. from Princeton. ity to levy taxes and provide public goods. He has Karen Gordon Mills, a senior fellow at served as president of the EHA and as co-editor Harvard Business School and president of MMP of the Journal of Economic History, and has been a Group, Inc., which invests in financial services, John Simon Guggenheim Fellow. He received his consumer products, and other technology-enabled A.B. from Harvard College and his Ph.D. from businesses, is a new at-large board member. Mills Yale University. He succeeds Alan Olmstead as the served as the 23rd administrator of the U.S. George Mailath EHA representative. Small Business Administration, 2009 to 2013, George Mailath, Walter H. Annenberg overseeing the agency’s initiatives to assist small Professor in the Social Sciences and professor businesses during the economic recovery follow- of economics at the University of Pennsylvania ing the 2008 financial collapse. She is an expert and Goldsmith Professor in the Research School on the economic health and well-being of the of Economics at Australian National University nation’s small businesses and on U.S. competitive- (ANU), is the new representative of the University ness. Mills is the author of seminal research on

NBER Reporter • No. 3, September 2017 29 capital markets for small busi- Corporate Governance, and of the professional journals in nesses. She chairs the advisory Ethics emeritus at New York these fields and is the author, committee of the Private Capital University’s Stern School of coauthor or editor of 27 books, Research Institute, co-chairs Business, is the inaugural repre- most recently Regulating Wall the Bipartisan Policy Center’s sentative of New York University Street. He received B.S. and M.S. Main Street Finance Task Force, (NYU) on the board. In 2016, degrees from Lehigh University and is a member of the Council the board added NYU to the list and his Ph.D. from NYU. on Foreign Relations and the of institutions invited to nomi- In addition, at-large director Karen Mills Harvard Corporation. Mills nate NBER directors. Walter’s Don Conlan, retired president of earned an A.B. in economics principal areas of teaching and the Capital Group Companies, from Harvard and an MBA from research include international and Andrew Postlewaite of the Harvard Business School, where banking and capital markets, University of Pennsylvania, for- she was a Baker Scholar. corporate governance, and risk merly that university’s represen- Ingo Walter, Seymour management. He has authored tative on the board, were elected Milstein Professor of Finance, or co-authored papers in most to emeritus status.

Ingo Walter

Leah Boustan and William Collins Are New Co-Directors of the Development of American Economy Program Leah Boustan, a professor of economics at to provide new context and perspective for cur- Princeton University, and William Collins, the rent labor market outcomes. Her 2016 book, Terence E. Adderley Jr. Professor of Economics Competition in the Promised Land: Black Migrants at Vanderbilt University, are the new co-directors in Northern Cities and Labor Markets, has been of the Development of the American Economy widely celebrated. Boustan received her B.A. from (DAE) Program. Boustan, who joined the Princeton and her Ph.D. from Harvard. Princeton faculty this year, previously taught at Collins, who also is a professor of history at UCLA. She has been an NBER research associ- Vanderbilt and served as department chair 2011– Leah Boustan ate since 2012, and before that a faculty research 14, is currently editor of The Journal of Economic fellow since 2007. Collins, who became a faculty History. His research focuses on the evolution research fellow in 1999 and a research associate in of U.S. cities, racial wage differentials, and labor 2004, has been a member of the Vanderbilt faculty market dynamics. He recently co-edited, with since 1998. Robert Margo, the NBER volume Enterprising Boustan’s research focuses on U.S. labor mar- America: Businesses, Banks, and Credit Markets kets, with particular emphasis on the history of in Historical Perspective. Collins received his B.A. migration. She explores rich historical datasets and his Ph.D. from Harvard.

William Collins

30 NBER Reporter • No. 3, September 2017 Gita Gopinath and Pierre-Olivier Gourinchas Are New Co-Directors of International Finance and Macroeconomics Program

Gita Gopinath, the John the Harvard faculty, she taught from Lady Shri Ram College at Zwaanstra Professor of Inter­ at the University of Chicago. the University of Delhi and her national Studies and of econom- Gourinchas, an NBER research Ph.D. from Princeton. ics at Harvard, and Pierre-Olivier associate since 1998, was on the Gourinchas studies capi- Gourinchas, who holds joint faculty at the Stanford Graduate tal flows and global imbalances, appointments as a professor in School of Business and at debt crises, and, more gener- the economics department and Princeton University before mov- ally, international macroeco- Gita Gopinath as the S. K. and Angela Chan ing to Berkeley. nomic policy. He received the Professor of Global Management Gopinath’s research spans Bernácer Prize, recognizing the in the a range of topics, including best under-40 European mac- at the University of California, exchange rate pass-throughs, roeconomist, in 2007, and has Berkeley, are the new co-directors monetary unions, and sover- served as a member of the French of the International Finance and eign debt markets. She is man- Council of Economic Advisers. Macroeconomics (IFM) Program. aging editor of The Review of He is the editor of the Journal of Gopinath was appointed an Economic Studies and an eco- International Economics. A grad- NBER faculty research fellow in nomic adviser to the chief min- uate of the École Nationale des 2004, and became a research asso- ister of the Indian state of Ponts et Chaussées in Paris, he Pierre-Olivier Gourinchas ciate in 2011. Prior to joining Kerala. She received her B.A. received his Ph.D. from MIT.

David Autor and Alex Mas Are New Co-Directors of Labor Studies Program

David Autor, the Ford Professor of Economics of Economic Perspectives, and a member of the at MIT, and Alex Mas, a professor of econom- American Economic Association’s executive com- ics and public affairs in the economics department mittee. He received his B.A. from Tufts University and the Woodrow Wilson School at Princeton and his Ph.D. from Harvard. University, are the new co-directors of the Labor Mas has studied the impact of unions on pay Studies Program. Autor joined the NBER as a fac- and productivity, the consequences of credit mar- ulty research fellow in 1999 and became a research ket disruptions for labor demand by small firms, the associate in 2006. Mas, who taught at the University role of fairness considerations, norms, and social David Autor of California, Berkeley before joining the Princeton interactions in wage-setting, and the links between faculty, was appointed a faculty research fellow in residential segregation and labor market outcomes. 2006 and a research associate in 2009. He is director of the Industrial Relations Section at Autor’s research spans a wide range of issues in Princeton, and the editor of the American Economic labor economics, including the employment effects Journal: Applied Economics. He served as chief econ- of trade, the impact of technological change on omist of the U.S. Department of Labor in 2009–10, labor demand and the wage distribution, and the and as associate director for economic policy and effect of disability insurance on labor market par- chief economist in the U.S. Office of Management ticipation. He is director of the NBER’s Disability and Budget, 2010-11. Mas is a graduate of Macalaster Alex Mas Research Center, a former editor of the Journal College; he received his Ph.D. from Princeton.

NBER Reporter • No. 3, September 2017 31 Conferences

CEPRA/NBER Conference on Aging and Health

The NBER’s Conference on Aging and Health, supported by the Center for Performance and Research Analytics, took place in Lugano, Switzerland, on June 1–3. Fabrizio Mazzonna of Università della Svizzera Italiana, NBER Program on Aging Director Jonathan S. Skinner of Dartmouth College, and Massimo Filippini of ETH Zurich and Università della Svizzera Italiana organized the meeting. These researchers’ papers were presented and discussed:

• Mathilde C. M. Godard, GATE-LSE, CNRS, University of Lyon, and Pierre Koning and Maarten Lindeboom, Vrije Universiteit Amsterdam, “Screening Disability Insurance Applications and Targeting”

• Corina D. Mommaerts, University of Wisconsin at Madison, “Long-Term Care Insurance and the Family”

• Fabrizio Mazzonna and Osea Giuntella, University of Pittsburgh, “Sunset Time and the Economic Effects of Social Jetlag: Evidence from U.S. Time Zone Borders”

• Marika Cabral, University of Texas at Austin and NBER, and Mark R. Cullen, Stanford University and NBER, “Estimating the Value of Public Insurance Using Complementary Private Insurance” (NBER Working Paper No. 22583)

• Pieter Bakx, Institute of Health Policy & Management, Erasmus University Rotterdam; Bram Wouterse, CPB Netherlands Bureau for Economic Policy Analysis; Eddy Van Doorslaer, Erasmus School of Economics, Erasmus University Rotterdam; and Albert Wong, National Institute for Public Health and the Environment, “The Health Effects of a Nursing Home Admission”

• Teresa Bago d’Uva and Owen O’Donnell, Erasmus University Rotterdam, and Eddy Van Doorslaer, Erasmus School of Economics, Erasmus University Rotterdam, “Who Can Predict Their Own Demise? Heterogeneity in the Accuracy of Longevity Expectations”

• Itzik Fadlon, University of California, San Diego and NBER, and Torben Heien Nielsen, University of Copenhagen, “Family Health Behaviors”

• Nicole Maestas, and NBER, and Kathleen Mullen and David Powell, RAND Corporation, “The Effect of Population Aging on Economic Growth, the Labor Force, and Productivity” (NBER Working Paper No. 22452)

• Amitabh Chandra, Harvard University and NBER, and Douglas Staiger, Dartmouth College and NBER, “Predicting the Impact of Hospital Closures on Patient Outcomes”

• Florian Heiss, University of Düsseldorf; Daniel L. McFadden, University of California, Berkeley and NBER; Lauren Scarpati, University of Southern California; and Joachim Winter and Amelie C. Wuppermann, University of Munich, “The Housing Crisis of the Late 2000s and Causal Paths between Health and Socioeconomic Status”

Summaries of these papers are at: www.nber.org/confer/2017/CAHs17/summary.html

32 NBER Reporter • No. 3, September 2017 26th NBER-TCER-CEPR Conference

The 26th NBER-TCER-CEPR Conference, “Corporate Governance,” took place in Tokyo on June 22. This meeting was spon- sored jointly by the Centre for Economic Policy Research in London, the NBER, the Tokyo Center for Economic Research, the Center for Advanced Research in Finance, and the Center for International Research on the Japanese Economy. Franklin Allen, of Imperial College London, CEPR, and NBER, Shin-ichi Fukuda of Tokyo University, and Takeo Hoshi of Stanford University and NBER organized the meeting. These researchers’ papers were presented and discussed:

• Randall Morck, University of Alberta and NBER; M. Deniz Yavuz, Purdue University; and Bernard Yeung, National University of Singapore, “State-run Banks, Money Growth, and the Real Economy”

• Franklin Allen; Elena Carletti, Bocconi University and CEPR; and Yaniv Grinstein, Interdisciplinary Center Herzliya and European Corporate Governance Institute, “International Evidence on Firm Level Decisions in Response to the Crisis: Shareholders vs. Other Stakeholders”

• Benjamin E. Hermalin, University of California, Berkeley, “Biased Monitors: Corporate Governance When Managerial Ability is Mis-Assessed”

• Naoshi Ikeda, Kotaro Inoue, and Sho Watanabe, Tokyo Institute of Technology, “Enjoying the Quiet Life: Corporate Decision-Making by Entrenched Managers”

• Hideaki Miyajima and Ryo Ogawa, Waseda University, and Takuji Saito, Keio University, “Changes in Corporate Governance and President Turnover: The Evidence from Japan”

• Takeo Hoshi, “Decline of Bank-Led Restructuring in Japan: 1980–2010”

Summaries of these papers are at: www.nber.org/confer/2017/TRIO17/summary.html

East Asian Seminar on Economics

The NBER, the Australian National University, the Peking University China Center for Economic Research, the Chung- Hua Institution for Economic Research (Taipei), the Hong Kong University of Science and Technology, the Korea Development Institute, the National University of Singapore, the Tokyo Center for Economic Research, and Tsinghua University (Beijing) jointly sponsored the NBER’s 28th Annual East Asian Seminar on Economics, “Inequality.” It took place in Manila, Philippines, on June 29–30. Research Associates Takatoshi Ito of and Andrew K. Rose of the University of California, Berkeley, organized the conference. These researchers’ papers were presented and discussed:

• Nicholas Bloom, Stanford University and NBER; Fatih Guvenen, University of Minnesota and NBER; David Price, Stanford University; Jae Song, Social Security Administration; and Till M. von Wachter, University of California, Los Angeles and NBER, “Firming up Inequality” (NBER Working Paper No. 21199)

• Bo Chen, Shanghai University of Finance and Economics; Miaojie Yu, Peking University; and Zhihao Yu, Carleton University, “Measured Skill Premium and Input Trade Liberalization: Evidence from Chinese Firms”

• Pi Chen and Suling Peng, Chung-Hua Institution for Economic Research, “Wage Inequality in Taiwan and Its Causes”

• Masayuki Inui and Nao Sudo, Bank of Japan, and Tomoaki Yamada, Meiji University, “Effects of Monetary Policy Shocks on Inequality in Japan”

NBER Reporter • No. 3, September 2017 33 • Yukinobu Kitamura, Hitotsubashi University; Takeshi Miyazaki, Kyushu University; and Taro Ohno, Shinshu University, “Income Tax Reforms and Intra-Generational Redistribution: Evidence from Japan”

• Joshua Aizenman, University of Southern California and NBER, and Yothin Jinjarak and Ilan Noy, Victoria University of Wellington, “Vocational Education, Manufacturing, and Income Distribution: International Evidence and Case Studies”

• Ronald Mendoza and Miann Banaag, Ateneo School of Government (Manila), “Political and Economic Inequality: Insights from Philippine Data on Political Dynasties”

• Dan Liu, Shanghai University of Finance and Economics, and Christopher M. Meissner, University of California, Davis and NBER, “Geography, Income, and Trade When Income Inequality Matters”

• Jakob B. Madsen, Monash University (Melbourne), “Piketty’s Third Law of Capitalist Economics and the Dynamics of Inequality in Britain, 1210–2013”

Summaries of these papers are at: www.nber.org/confer/2017/EASE17/summary.html

International Seminar on Macroeconomics

The NBER’s 40th International Seminar on Macroeconomics, hosted by the Bank of Lithuania, took place in Vilnius, Lithuania, on June 30–July 1. The seminar was organized by Research Associates Jeffrey Frankel of Harvard University and Hélène Rey of London Business School. These researchers’ papers were presented and discussed:

• Thomas Drechseland Silvana Tenreyro, London School of Economics, “Commodity Booms and Busts in Emerging Economies”

• Jonas Heiperz, Paris School of Economics; Amine Ouazad, HEC Montreal; Romain Rancière, University of Southern California and NBER; and Natacha Valla, European Investment Bank and Paris School of Economics, “Balance-Sheet Diversification in General Equilibrium: Identification and Network Effects” (NBER Working Paper No. 23572)

• Antonio Fatás, INSEAD, and Lawrence H. Summers, Harvard University and NBER, “The Permanent Effects of Fiscal Consolidations” (NBER Working Paper No. 22374)

• Wenxin Du and Joanne Im, Federal Reserve Board, and Jesse Schreger, Harvard University and NBER, “The U.S. Treasury Premium”

• Cristina Arellano, Federal Reserve Bank of Minneapolis and NBER; Yan Bai, University of Rochester and NBER; and Gabriel Mihalache, Stony Brook University, “Default Risk, Sectoral Reallocation, and Persistent Recessions”

• Marc Flandreau, University of Pennsylvania, “Sovereign Debt Enforcement: Historical Evidence on the Role of Financial Engineering”

• Bartosz Mackowiak and Marek Jarocinski, European Central Bank, “Monetary Fiscal Interactions and the Euro Area’s Malaise”

• Ambrogio Cesa-Bianchi, Bank of England; Andrea Ferrero, University of Oxford; and Alessandro Rebucci, Johns Hopkins University and NBER, “International Credit Supply Shocks”

Summaries of these papers are at: www.nber.org/confer/2017/ISOM17/summary.html

34 NBER Reporter • No. 3, September 2017 40th Annual NBER Summer Institute

The NBER hosted its 40th annual Summer Institute during a three-week period in July. The 2,965 registered participants took part in 52 distinct meetings led by more than 100 organizers. About one in six participants — 608 researchers — were attending their first Summer Institute. There were 99 graduate student participants. More than two thirds of the participants were not NBER affiliates. Lord Mervyn King, former Governor of the Bank of England and an NBER research associate, delivered the 2017 Martin Feldstein Lecture on “Uncertainty and Large Swings in Activity.” He examined the capacity of standard macroeconomic models to explain sharp declines in economic activity, such as those associated with the Great Depression or the 2008 global financial cri- sis. He recommended that researchers be open to new models, and that they recognize the inherent uncertainty associated with vir- tually all modeling exercises in macroeconomics. An edited text of the lecture appears earlier in this issue of The NBER Reporter. The 2017 Methods Lectures focused on data linking. Research Associates John Abowd of Cornell University and the U.S. Bureau of the Census, Martha Bailey of the University of Michigan, and Joseph Ferrie of Northwestern University described statisti- cal tools and matching algorithms for merging large administrative data sets. They also presented various applications of these tools, describing linkages between decennial Census data files and between firm and employee data sets. In addition, Jonathan Schwabish of the Urban Institute lectured on “Data Visualization,” presenting a range of new tools and techniques for the visual display of data. All of the presentations — the Feldstein Lecture, the Methods Lectures, and the presentation on data visualization — have been videotaped and can be accessed through the NBER Videos tab on the left side of the NBER homepage.

Japan Project

The NBER held a meeting on the Japanese economy in Tokyo on July 31. The seminar was organized by Shiro Armstrong of the Australian National University, Research Associate Charles Horioka of the Asian Growth Research Institute (Kitakyushu), Research Associate Takeo Hoshi of Stanford University, Tsutomu Watanabe of the University of Tokyo, and Research Associate David Weinstein of Columbia University. These researchers’ papers were presented and discussed:

• Koichiro Ito, University of Chicago and NBER; Takanori Ida, ; and Makoto Tanaka, GRIPS (National Graduate Institute for Policy Studies, Tokyo), “Information Frictions, Inertia, and Selection on Elasticity: A Field Experiment on Electricity Tariff Choice”

• J. Mark Ramseyer, Harvard University, and Eric B. Rasmusen, Indiana University, “Outcaste Politics and Organized Crime in Japan: The Effect of Terminating Ethnic Subsidies”

• Kozo Ueda, Waseda University; Kota Watanabe, Meiji University; and Tsutomu Watanabe, “Product Turnover and Deflation: Evidence from Japan”

• Makoto Saito, Hitotsubashi University, “On Large-Scale Money Finance in the Presence of Black Markets: A Case of the Japanese Economy during and Immediately after World War II”

• Kentaro Nakajima, Hitotsubashi University, and Kensuke Teshima, Instituto Tecnológico Autonomo de México, “Identifying Neighborhood Effects among Firms: Evidence from Location Lotteries of the Tokyo Tsukiji Fish Market”

• Wataru Miyamoto, Bank of Canada; Thuy Lan Nguyen, Santa Clara University; and Dmitriy Sergeyev, Bocconi University, “Government Spending Multipliers under the Zero Lower Bound: Evidence from Japan”

Summaries of these papers are at: nber.org/confer/2017/JPMs17/summary.html

NBER Reporter • No. 3, September 2017 35 Economic Dimensions of Personalized and Precision Medicine

The NBER’s Conference on Economic Dimensions of Personalized and Precision Medicine was supported by the USC Schaeffer Center for Health Policy & Economics and Columbia University. The conference took place in Santa Monica, CA, on September 13–14. Research Associates Ernst R. Berndt of MIT and Dana Goldman of the University of Southern California, and John Rowe of Columbia University organized the meeting. These researchers’ papers were presented and discussed: • Ernst R. Berndt, and Mark Trusheim, MIT, “The Information Pharms Race and Competitive Dynamics of Precision Medicine” • Manuel I. Hermosilla, Johns Hopkins University, and Jorge A. Lemus, University of Illinois at Urbana-Champaign, “Therapeutic Translation in the Wake of the Genome” • John A. Graves and Josh Peterson, Vanderbilt University, “Rational Integration of Genomic Health Care Technology: Evidence from PREDICT” • Kristopher Hult, University of Chicago, “Measuring the Potential Health Impact of Personalized Medicine: Evidence from MS Treatments” • Rachel Lu, Chang Gung University (Taiwan); Karen Eggleston, Stanford University and NBER; and Joseph T. Chang, Chang Gung Memorial Hospital, “Economic Dimensions of Personalized and Precision Medicine in Asia: Evidence from Breast Cancer Treatment in Taiwan” • Mark Pauly, University of Pennsylvania and NBER, “Cost Sharing in Insurance Coverage for Precision Medicine” • Frank R. Lichtenberg, Columbia University and NBER, and Rebecca A. Pulk, Marc S. Williams, and Eric Wright, Geisinger Health System, “The Social Cost of Suboptimal Medication Use and the Value of Pharmacogenomic Information: Evidence from Geisinger” • Amitabh Chandra, Harvard University and NBER; Craig Garthwaite, Northwestern University and NBER; and Ariel Dora Stern, Harvard University, “Characterizing the Drug Development Pipeline for Precision Medicines” • David H. Howard, Emory University; Jason Hockenberry, Emory University and NBER; and Guy David, University of Pennsylvania and NBER, “Personalized Medicine When Physicians Induce Demand” • Philippe Gorry, University of Bordeaux, “Empirical Economic Analysis of Orphan Drug Innovation” Summaries of these papers are at: www.nber.org/confer/2017/PPMf17/summary.html

Economics of Artificial Intelligence

The NBER’s Conference on Artificial Intelligence took place in Toronto on September 13–14. Research Associates Ajay K. Agrawal, Joshua Gans, and Avi Goldfarb of the University of Toronto organized the meeting. These researchers’ papers were pre- sented and discussed: • Iain M. Cockburn, Boston University and NBER; Rebecca Henderson, Harvard University and NBER; and Scott Stern, MIT and NBER, “The Impact of Artificial Intelligence on Innovation” • Erik Brynjolfsson, MIT and NBER; Daniel Rock, MIT; and Chad Syverson, University of Chicago and NBER, “Artificial Intelligence and the Modern Productivity Paradox: A Clash of Expectations and Statistics” • Paul Milgrom, Stanford University, and Steven Tadelis, University of California, Berkeley and NBER, “Market Design” • Susan Athey, Stanford University and NBER, “Impact on Economics” • Ajay K. Agrawal, Joshua Gans, and Avi Goldfarb, “Prediction, Judgment, and Complexity”

36 NBER Reporter • No. 3, September 2017 • Catherine Tucker, MIT and NBER, “Privacy”

• Daniel Trefler and Avi Goldfarb, University of Toronto and NBER, “Trade”

• Colin Camerer, California Institute of Technology, “Behavioral Economics”

• Jeffrey D. Sachs, Columbia University and NBER, “Income Distribution”

• Philippe Aghion, College de France; Benjamin Jones, Northwestern University and NBER; and Charles I. Jones, Stanford University and NBER, “Artificial Intelligence and Economic Growth”

• Joel Mokyr, Northwestern University, “Historical Context and the Long Run”

• Carl Shapiro, University of California, Berkeley and NBER, and , University of California, Berkeley, “Machine Learning, Market Structure, and Competition”

• Joseph E. Stiglitz, Columbia University and NBER, and Anton Korinek, Johns Hopkins University and NBER, “Artificial Intelligence, Worker-Replacing Technological Change, and Income Distribution”

• David Autor, MIT and NBER, and Anna Solomons, Utrecht University, “Robocalypse Now: Does Productivity Growth Threaten Employment?”

Summaries of these papers are at: www.nber.org/confer/2017/AIf17/summary.html

Tax Policy and the Economy

The NBER’s Conference on Tax Policy and the Economy took place in Washington, DC, on September 14. Research Associate Robert A. Moffitt of Johns Hopkins University organized the meeting. These researchers’ papers were presented and discussed:

• Alex Rees-Jones, University of Pennsylvania and NBER, and Dmitry Taubinsky, University of California, Berkeley and NBER, “Taxing Humans: Normative Implications of Biased Responses to Taxes”

• James Andreoni, University of California, San Diego and NBER, “The Benefits and Costs of Donor Advised Funds”

• Andrew Samwick, Dartmouth College and NBER, “Means-Testing Federal Health Entitlement Benefits” (NBER Retirement Research Center Paper No. NB 12-16)

• Jeffrey Clemens, University of California, San Diego and NBER, and Benedic N. Ippolito, American Enterprise Institute, “Implications of Medicaid Financing Reform for State Government Budgets”

• Bruce D. Meyer, University of Chicago and NBER, and Wallace K. C. Mok, Chinese University of Hong Kong, “Disability, Taxes, Transfers, and the Economic Well-Being of Women”

• Caroline M. Hoxby, Stanford University and NBER, “Online Postsecondary Education and the Higher Education Tax Benefits: An Analysis with Implications for Tax Administration”

Summaries of these papers are at: www.nber.org/confer/2017/TPE17/summary.html

NBER Reporter • No. 3, September 2017 37 Program Meeting

Economic Fluctuations and Growth

The NBER’s Program on Economic Fluctuations and Growth met in Cambridge on July 15. Research Associates Mark Bils of the University of Rochester and Gita Gopinath of Harvard University organized the meeting. These researchers’ papers were pre- sented and discussed:

• Nuno T. Coimbra, Paris School of Economics, and Hélène Rey, London Business School and NBER, “Financial Cycles with Heterogeneous Intermediaries” (NBER Working Paper No. 23245)

• Philippe Aghion, College de France; Antonin Bergeaud, Banque de France; Timo Boppart, IIES, Stockholm University; Peter J. Klenow, Stanford University and NBER; and Huiyu Li, Federal Reserve Bank of San Francisco, “Missing Growth from Creative Destruction”

• David W. Berger and Ian Dew-Becker, Northwestern University and NBER, and Stefano Giglio, University of Chicago and NBER, “Uncertainty Shocks as Second-Moment News Shocks”

• Emmanuel Farhi, Harvard University and NBER, and David Baqaee, London School of Economics, “The Macroeconomic Impact of Microeconomic Shocks: Beyond Hulten’s Theorem” (NBER Working Paper No. 23145)

• Daron Acemoglu, MIT and NBER, and Pascual Restrepo, Boston University, “Robots and Jobs: Evidence from U.S. Labor Markets” (NBER Working Paper No. 23285)

• Òscar Jordà, Federal Reserve Bank of San Francisco; Katharina Knoll, Dmitry Kuvshinov, and Moritz Schularick, University of Bonn; and Alan M. Taylor, University of California, Davis and NBER, “The Rate of Return on Everything, 1870–2015”

Summaries of these papers are at: www.nber.org/confer/2017/EFGs17/summary.html

38 NBER Reporter • No. 3, September 2017 NBER Books

Measuring Entrepreneurial Businesses: Current Knowledge and Challenges (Studies in Income and Wealth, Volume 75) John Haltiwanger, Erik Hurst, Javier Miranda, and Antoinette Schoar, editors Cloth $130 Start-ups and other entrepre- mists and data analysts to discuss neurial ventures make a significant the most recent research covering contribution to the U.S. economy, three broad themes. The first chap- particularly in the tech sector, where ters isolate high- and low-performing they comprise some of the largest entrepreneurial ventures and analyze and most influential companies. Yet their roles in creating jobs and driv- for every high-profile, high-growth ing innovation and productivity. The company like Apple, Facebook, next chapters focus on specific chal- Microsoft, and Google, many more lenges entrepreneurs face and how fail. This enormous heterogeneity these have varied over time, includ- poses conceptual and measurement ing over business cycles. The final challenges for economists concerned chapters explore core measurement with understanding their precise issues, with a focus on new data proj- impact on economic growth. ects under development that may Measuring Entrepreneurial improve our understanding of this Businesses brings together econo- dynamic part of the economy.

For information on ordering and electronic distribution, see www.press.uchicago.edu/books/orders.html or to place an order you may also contact the University of Chicago Press Distribution Center, at Telephone: 1-800-621-2736 Email: [email protected]

NBER Reporter • No. 3, September 2017 39 Nonprofit Org. NBER U.S. Postage Reporter PAID NATIONAL BUREAU OF ECONOMIC RESEARCH National Bureau of Economic Research 1050 Massachusetts Avenue Cambridge, Massachusetts 02138-5398 (617) 868-3900

Change Service Requested