Interview with James Poterba

Interview conducted April 1, 2008.

It's often said that economists agree on little, but there was no divergence of opinion when the National Bureau of Economic Research, the nation's premier body of research economists, announced it had selected James Poterba as its new president and CEO as of July 1. Economists of all stripes were unanimous in their praise.

Left-of-center columnist Paul Krugman deemed Poterba "a great choice." Republican Martin Feldstein, NBER president for 30 years, declared Poterba "an excellent fit." Freakonomics co-author Steven Levitt summed it up: "The choice was so obvious that I can't say I'm surprised. Nonetheless, I couldn't be more delighted, both for Poterba and for the profession. This is wonderful news."

Poterba—a summa cum laude graduate of Harvard College, a Marshall scholar at Oxford University, a fellow of the American Academy of Arts and Sciences and the Econometric Society—personifies the highest standards of scholarship in economics. His work on the impact of taxation on household and firm behavior is fundamental in the profession, and his recent exploration of how taxes affect savings and portfolio decisions, and tax-deferred retirement plans in particular, has been pathbreaking.

But skills honed as head of the top-rated economics department at the Massachusetts Institute of Technology, as editor of the Journal of Public Economics, as director of the NBER's program on public economics and on countless policy committees over the years will be central to his future leadership of NBER. Colleagues refer often to his impartiality and his ability to bring fractious debate to cohesive conclusion. In the contentious world of research (and allocation of resources within), such skills will be invaluable.

In the following interview, that same respect for evidence and thoughtful consideration of alternatives is evident, but Poterba also promises to respond "nimbly" to the pressing economic challenges of our time. "One of the NBER's great strengths," he notes, "is its ability to bring a whole range of economic researchers to bear on important questions as they emerge."

TAX REFORM

Region: Let me throw you in the deep end. With the presidential election drawing near, tax policy is sure to become a topic of increasing debate. As a tax scholar and a member of the 2005 President's Advisory Panel on Tax Reform, what kind of advice would you offer to an incoming president on tax policy?

Poterba: We are approaching a period when tax reform will likely move out of academic discussion and attract serious attention from Washington policymakers, for two reasons.

One is the Alternative Minimum Tax, a feature of the tax code that was originally designed to affect very-high-income taxpayers but has over time come to affect a larger and larger fraction of the U.S. tax-paying population because the relevant provisions were not indexed for inflation. Congress has been fixing the AMT one year at a time by raising the thresholds at which it affects taxpayers. But the one-year fix strategy is becoming increasingly expensive because the compounding of inflation keeps lowering the levels at which statutory thresholds kick in. It's likely that Congress will have to do something to fix the AMT in a more permanent way—after the next election would be a natural time.

The other factor that will draw tax reform into the middle of the political debate is the scheduled expiration of the tax changes that were enacted in 2001 and 2003. The expiring provisions include reductions in marginal tax rates as well as the tax rate relief on dividends and capital gains. The estate tax will also revert to its pre-2001 form if Congress does not act in 2009 or 2010.

It's difficult to predict where tax reform is likely to go. When the president's tax panel looked at the AMT and other issues in 2005, we recognized that there was no easy solution to the challenge of tax reform. Fixing the AMT is a very expensive proposition. It costs more than a trillion dollars over a medium-term budget window to repeal the AMT, and if you're going to do that in a revenue-neutral way, you have to either change income tax rates or broaden the income tax base. The tax panel pushed toward the base-broadening approach, which I think has a great deal to recommend it.

Of course, supporting base-broadening is hardly novel. It's almost a reflexive action for most public finance economics. Marginal distortions associated with the tax code tend to rise with the square of the tax rate, so that as the tax rate gets into higher and higher territory, the marginal dead-weight losses tend to grow rapidly. Going from a 30 percent to a 40 percent marginal tax rate, for example, doesn't increase the marginal efficiency cost of taxation by a factor of 1.33, but by the ratio of 16 over 9—close to 75 percent.

Region: Thus the appeal of broadening the base.

Poterba: Exactly. Broadening the tax base reduces the need to impose high marginal dead-weight burdens associated with high marginal tax rates. The challenge of broadening the base—which I fear the tax panel members learned the hard way by proposing a lot of base-broadening and then seeing the proposals go nowhere in the political system—is that most of the things that erode the current income tax base are political sacred cows. These include the home mortgage interest deduction, the exclusion of employer-provided health insurance from the tax base for individuals, the deduction for state and local property taxes, and the relief we provide on medical expenditures of various kinds. They all have constituencies that find them very important and very attractive. There isn't a strong constituency at the moment, and probably never will be, for the efficiency gains associated with broadening the tax base and lowering rates.

So you're faced with one of the classic political economy problems of imposing burdens on an identifiable group that can tell they're going to have their oxes gored by the particular reform you've proposed, while generating broadly diffused benefits for the population at large in the form of a more efficient structure for raising revenue.

Region: Is there any hope that the political system can embrace broad-based tax reform?

Poterba: It's not impossible, and it's happened before.  In 1986, the tax system was reformed in a way that broadened the base and lowered marginal rates across the board. What that took was a combination of the political stars lining up in just the right way and the policymakers and policy advisers—the academic economists and the economists in the policy process—making a strong and articulate case for the benefits of an efficiency-improving tax reform.

In 1986, we'd been through a period with high inflation which had wrought havoc on the effective tax burdens on many investments. Inflation tends to raise the effective tax burden on bonds because the inflation premium in nominal bonds is taxed. The United States had higher tax marginal rates than many of our international competitors at that point. Policymakers recognized that these were serious concerns, and they managed, quite remarkably, to draw together in support of a major tax reform. I don't think it's impossible to do it again.

Region: The advisory panel also suggested a simpler code; is that a feasible goal?

Poterba: Absolutely. Simplification may be the easiest dimension of reform. If you look, for example, at the way we currently provide incentives for saving to finance retirement, education, medical expenditures—we have a dizzying array of programs: 401(k)s, IRAs, Roth IRAs, 403(b)s, 457s—all of these are basically ways of offering households what public finance economists call "consumption tax treatment" on their savings. By investing in these special forms, investors can earn the pretax return on their investments. But each of these account types has its own complicated set of rules, and many households don't know what they're eligible for and don't take full advantage of the opportunities that these programs present. Some simplification could surely be achieved by rolling some of those programs into a smaller number of more broadly available saving vehicles.

There probably are other opportunities for moving in the direction of a return-free system for a substantial number of low-income taxpayers who have relatively simple tax returns. If a taxpayer's only income source is wage income or that plus some very easily identifiable capital income, perhaps from a bank, the IRS could easily compute the household's tax liability and send out a bill.

The problem with simplicity is that while almost everybody thinks that a simpler tax code is an attractive ideal, in many cases the tax code is not simple because it's trying to deal with the complexity of economic life. You could make the tax code simpler, for example, by having one tax schedule for everybody and just saying that married couples and single individuals would file on the same schedule; no need to look up different rates or anything else. But of course that would offend some people's sense of fairness. They would say that a two-person household with a given income should not necessarily face the same tax burden as a five-person household with the same income, or as a single individual with that income. So you say, "Okay, we'll have different tax codes for married couples; we'll allow for some dependency deductions." Soon you're on the path to making the tax system more complicated. The same thing happens to an even greater degree when you're trying to track the income from investment property. A lot of the complexity really comes from the fact that people do complicated things, and we need to track them appropriately.

While it would be straightforward to simplify the tax code for quite a large number of taxpayers, some of the inherent complexity of the tax code that's associated with the measurement of capital income, with the tracking of business receipts and things like that, would be very difficult to get rid of without ending up with a tax system that many people would find objectionable.

TAX EXPENDITURES

Region: Going back to the issue of a broader tax base, I believe you've just hosted a conference on "tax expenditures." Can you tell us about it?

Poterba: Sure. I organized the conference to address some questions that arose during the tax panel's discussions. Those discussions, incidentally, were great. Serving on the panel was a remarkably interesting and educational experience. The other members brought a variety of interesting and insightful backgrounds to bear on the questions of tax reform, and they worked incredibly hard on the issues we faced. We also had a fabulous professional staff of both lawyers and economists who tackled a lot of the difficult issues that tax reform raises.

As I worked with that very distinguished group, I realized that a key issue that is likely to emerge next time we have a big tax reform discussion concerns how to broaden the income tax base. This will require good estimates of the economic effects of various current tax code provisions that narrow the base. The research projects that were presented at the recent conference focused on this issue. Under the auspices of the National Bureau of Economic Research, I brought together 11 research teams working on various features of the current tax code that cause deviations from a broad-based income tax and lead to revenue reductions.

Region: Revenues forgone by the government as a result of tax exemptions, credits or exclusions?

Poterba: Exactly. Over time there's been an increasing use of the tax code for social engineering through various kinds of deductions and credits, most of which would fall into the tax expenditure category. For example, we now encourage various kinds of energy activities with tax credits or tax deductions. We encourage the construction of low-income housing with tax credits. These incentives are joining the more traditional and bigger ticket items like the tax exemption for employer-provided health insurance and for property taxes, the home mortgage interest deduction, and the exemption for state and local interest payments on tax-exempt bonds.

I organized the research project that was summarized at the conference to focus on a set of tax expenditures that would be regarded as narrowing the tax base both by those who support an income tax and by those who support an expenditure tax, a tax based on consumption rather than income. An income tax taxes the return to capital income; a consumption tax does not.

The research presented at the conference did not consider the tax expenditures for private pensions, 401(k)s, IRAs and other types of tax-deferred savings. These provisions of the tax code reduce revenues if the benchmark is an income tax, but not if it is a consumption tax. The conference brought new empirical work to bear on these issues and also provided detailed information based on tax returns and household surveys on the distribution of benefits associated with the current tax expenditures. Some of the researchers also looked at the behavioral consequences of tax expenditures.

To give you one example, my work with Todd Sinai at the Wharton School at the University of Pennsylvania focused on the tax treatment of owner-occupied housing. Not all homeowners have mortgages; the elderly often have fully paid-up houses, so they don't take advantage of the mortgage interest deduction. By contrast, younger households, particularly high-income younger households, have a lot of mortgage debt. Also, because house prices vary a lot from place to place—they're much higher in Boston than in Minneapolis and higher still in San Francisco—the geographical diversity of the benefits of the home mortgage interest deduction is substantial. Another source of variation is that not all households with mortgages itemize on their tax returns. Those who do not cannot claim a mortgage interest deduction.

Todd and I found that the mortgage interest deduction is a benefit skewed toward young households in high-income areas on the East and West coasts. It is also more beneficial to higher-income taxpayers because the benefit is proportional to the taxpayer's marginal tax rate. So somebody who's at a 35 percent bracket saves more than somebody who's in a 25 or 15 percent bracket. That's an important feature of many tax expenditures—their value is increasing with marginal tax rates. It is not obvious that one would design a tax subsidy so that its value increases with a household's income or marginal tax rate.

Region: You've also been looking at the nontaxation of imputed rent of owner-occupied housing, I believe.

Poterba: Yes. As a homeowner, you are buying a long-lived durable asset and then renting it to yourself each period. Think of it as two transactions: In one, you're a consumer who pays the landlord, while in the other, you're the landlord who's collecting the rent. It just happens that the landlord and the tenant are the same person.

The income tax doesn't tax the implicit rental income that a homeowner receives in the landlord capacity. If the same taxpayer owned a rental building and collected rent from a tenant, that income would be taxed. The nontaxation of implicit rent means that owner-occupied housing is taxed less heavily than many other assets in the economy. The result is that we have much more owner-occupied housing in the U.S. economy than we would otherwise. The distortion is likely to be larger for high-income taxpayers who are in higher tax brackets.

This is a standard example of how the tax code distorts the economy, which in turn leads to a less efficient allocation of capital than in a no-tax world. In many general equilibrium models built to study taxes and other issues, when there are different tax burdens on different kinds of assets, the level of output is substantially lower than in an economy in which all assets are taxed the same way.

HOUSING AND THE WEALTH EFFECT

Region: Let me ask about another issue related to housing. In 2000, in a paper reviewing evidence on the wealth effect, you concluded that the "rising stock market has surely contributed to rising consumer spending in the 1990s" but that there is "at best a weak link between house price changes and nonhousing consumption."

Since that time, of course, both equity and housing markets have undergone significant boom and collapse cycles. What is your sense right now of the impact of the housing market collapse on consumer spending?

Poterba: Great question. Much of the discussion of the wealth effect seems to assume that there's a question of whether it exists. I think that's a misapprehension. The simple logic of budget constraints for multiperiod consumers and households tells you that if you reduce the value of the assets that households own at a given point in time, the present discounted value of their consumption stream must be reduced correspondingly. So the real question is simply, How does a household, or the household sector in aggregate, spread the loss of wealth from a drop in equity values or house values over various consumption in various future periods?

Region: So it's not whether, but when and how big.

Poterba: Right, and once we say that, we are drawn to the question of how precisely we can measure these effects. This is a particular challenge since in the aggregate data, if asset prices have changed precipitously over some time period, it's typically not the case that the only shock buffeting the economy is the movement in asset prices. If we see a significant move in stock prices, for example, it's possible that consumers might also have changed their future expectations of interest rates or other discount factors; risk premia may have changed in the economy; the expected wage path the consumer thinks he or she may be able to achieve by supplying labor in the future may also have changed at about the same time.

All of those variables potentially can affect the level of consumer spending, so teasing out the part that can be identified with the wealth effect of movements in the stock market becomes very difficult. The same is true when you pose the question with respect to housing, particularly in the context of a large house price movement like the one we've seen in the last couple of years. When we look back at the current period, it may be hard to identify the effect of house price movements as distinct from changes in lending practices and other aspects of the financial environment.

Region: But scholars have made the effort nonetheless. What have you found?

Poterba: Yes, there have been some attempts to look at cross-sectional differences in the exposure of households to different kinds of wealth. For example, some households hold more stock than others and consequently would be more exposed to a revaluation of stock market equities, so one can study those households to see if they are particularly sensitive to stock price fluctuations.

Equity ownership is concentrated in the United States. Households in the top 1 percent of the wealth distribution own more than half of the corporate stock which is held outside of retirement accounts. Even if you include retirement account wealth, the concentration is still very high. With housing equity, by the way, there's a more egalitarian distribution. The top 1 percent owns about 15 percent of housing. In the portfolios of the richest households, housing equity is less important than other sources of financial wealth.

One thing that researchers have tried is to see if there is evidence that households with lots of equity exposure tend to cut their consumption back more when the stock market tends to decline. Unfortunately, we don't have very good measures of consumer spending at the household level linked up with high-quality data on the household balance sheet, especially for the best-off households.

The Federal Reserve System sets the gold standard for household balance sheet information with the Survey of Consumer Finances. This survey is administered every three years. The most recent version was 2007, and we should have the data from that survey later this year.

Region: But there are definite limitations to SCF data, including a small sample of upper-income families, right?

Poterba: It's true that the SCF doesn't have a very large sample of the upper-wealth households, but it is still by far the best source of information about this group. But the real problem for studying wealth effects is that the SCF doesn't have any data on consumer spending. This is not because the consumption data would not be of great value—it is just because there is a trade-off between getting respondents to answer the survey and the length of the survey, and adding questions about consumption might lengthen the questionnaire and depress the response rate. Also, the SCF is a snapshot. It's a one-time cross section each time it's administered, so researchers cannot study a given household and see how its wealth changed over time.

The data sets that do have a lot of information on consumer spending either have very little information on assets or have very little information on the high-wealth households who are likely to be the key holders of corporate stock.

It may be possible, using other data sets, to get a better fix on the wealth effect of house price movements. But there is another complication. Movements in house values in regional and local economies are tied up with employment prospects and many other factors. In some of the boom housing markets of the last decade—Southern California, Las Vegas, South Florida, for example—construction became a very important part of the economy. When house prices decline and the construction sector contracts, the effect on consumer spending cannot be attributed only to a simple wealth effect. It is very hard to design empirical tests which can distinguish wealth effects from other associated shocks.

These serious concerns notwithstanding, when one looks at the historical correlations between aggregate wealth movements and aggregate consumer spending, controlling as well as possible for other shocks to the economy, one finds something between 4 cents on the dollar, 6 cents on the dollar, of change in consumer spending within about 12 months of a stock market revaluation. That's not very different from the estimate of the consumption effect you would develop if instead of looking at data, you used a standard life-cycle model with plausible parameter values and worked through the potential effects of a wealth shock on consumer spending.

With regard to measuring the housing wealth effect, I worry that historical correlation patterns between price changes and consumer spending may be misleading because the financial services environment has changed in ways that may alter the underlying behavior we're interested in. The changes in lending practices that took place between 1990 and 2002 made it easier for households to tap into their home equity and provided home equity loans on a broader scale than in the past. Related changes altered the loan-to-value ratios of first-time home buyers in a way that could change the link between house prices and consumer spending. I suspect that the innovations in the financial services area increased the linkages between changes in house prices and changes in consumer spending. We saw those on the upside in a period of rising house prices, as households withdrew equity from their homes. The current discussions of foreclosures and trying to find ways to keep people in their homes suggest that we're just in the early stages of finding out what those dynamics are going to be on the downside.

SAVING FOR RETIREMENT

Region: With Joshua Rauh, Steven Venti and David Wise, you've conducted a great deal of research on the shift from defined-benefit to defined-contribution retirement plans, one of the major changes in the private pension landscape. Can you tell us about that transformation?

Poterba: Over the last two decades, we have seen a radical change in the role of the individual in taking responsibility for retirement in the U.S. economy.

Twenty-five or 30 years ago, the predominant form of employer-provided retirement benefit was a defined-benefit pension plan. A defined-benefit, or DB, pension plan is a liability of the employer which gives the worker an annuity when he or she retires. In most cases DB plans provided a guaranteed nominal income stream for as long as the worker, or the worker and the worker's spouse, remained alive. The worker never had to think much during the working phase about the way DB plan assets were invested by the firm. Most workers just knew that there was a formula, often complicated, which related their benefits after retirement to their last wage, years of service at the firm and the age at which they retired.

In the late 1960s and early 1970s, several large firms went bankrupt and their workers were left with empty pension promises. Their DB plans had not been funded, and this exposed a set of risks associated with the DB paradigm. In 1974, Congress passed ERISA, the Employee Retirement Income Security Act. It regulated DB plans along a number of dimensions and provided federal insurance for retiree benefits. The regulatory structure for DB plans was more stringent than that for defined-contribution plans, to which firms and employers contributed but with no benefit promise at retirement. Subsequent legislative changes raised the regulatory burden, particularly on firms with DB plans.

Firms responded to the changing regulatory environment, and to growing worker interest in defined-contribution plans in the 1990s, by shifting away from the historical DB model and toward DC plans. Very few new defined-benefit plans have been started in the last decade. There has been very rapid growth in defined-contribution plans, primarily the 401(k) type of plans. With these plans, the individual takes responsibility for deciding how much to contribute, how to invest the assets, what to do with the assets at retirement. The retirement options include leaving the assets in the plan, annuitizing them or taking them out as a lump-sum distribution. Those choices are important ones, and they all involve difficult decisions.

Financial economists will recognize that the planning problem for a household thinking about financial accumulation over the life cycle in a world with uncertain labor income, uncertain asset returns and uncertain mortality is difficult—even for someone with sophisticated analytical tools. Ordinary households confront a pretty daunting task in trying to make these decisions. The DC-based retirement system to which we have shifted offers households the opportunity to save more than they typically could in the defined-benefit world and to make different choices for retirement depending in part on their preferences and attributes. The DC system allows an individual who is in poor health to choose a lump-sum payout rather than an annuity, but it also allows an uninformed but healthy individual who does not recognize the chance of a very long retirement to inadvertently endanger his late-life resources by making the same choice.

Region: What does this mean for the future of retirement? Are 401(k)s likely to provide greater financial security for retirees than defined-benefit plans did?

Poterba: We are not quite far enough into the 401(k) experience yet to see how people do in the drawdown phase. Even though many retirees today are reaching retirement with some assets in a 401(k) plan, most of those people joined 401(k) plans relatively late in their working career.

The retirees who will hit retirement in 2025 or 2030 will have worked for most of their life under a 401(k) regime. They will have accumulated a lot more assets in that part of the retirement system, and the decisions they make will be much more consequential for how their retirement plays out than the decisions of today's retirees because many of today's 401(k) retirees also have a DB pension. A lot of 401(k) wealth is invested in equities. As we move forward, the future retirement accumulations for people in the 401(k) sector are going to be quite dependent on the equity return path.

If equities on average deliver returns like their historical returns since the 1920s, then there's every reason to think that accumulations in 401(k) plans will be larger by a substantial margin than the wealth values that were built up in traditional defined-benefit plans. On the other hand, there is a chance that equity markets will perform less well in the future than in the past—and a small chance of a much weaker performance. If they perform dramatically less well than they have historically, we are likely to end up with households with significantly lower levels of retirement wealth than previous generations of retirees. Of course, it is hard to know how a more traditional defined-benefit system would perform if equity markets generated weak returns for a prolonged period, since this would place heavy demands for plan contributions on the sponsoring firms.

One of the challenges that the shift toward the 401(k) model places on households is the need to become more educated about the financial decisions that they face. It places a responsibility on both public policymakers and firms' human resource departments to find relatively straightforward heuristics that will help people who in most cases don't really want to know the details of financial engineering.

Region: What kinds of heuristics would be useful in this context?

Poterba: Brigitte Madrian, David Laibson and their collaborators have done research on behavioral economics in the context of retirement and have found that many of the investment decisions people make are influenced by factors in the workplace. One of the best documented facts is what happens when a firm shifts from having a voluntary participation 401(k) plan where, when a worker is hired, he or she has to check a box and say, "Yes, I'd like to join the 401(k) plan and contribute X percent of my salary to this plan" to an "opt-out" plan in which the firm's hiring forms say, "We have a 401(k) plan. You will be automatically enrolled, and the firm will contribute 5 percent of your salary to your account."

Region: Unless you opt out.

Poterba: Right. You have to check a box to say "no contribution." Neoclassical economists would say the voluntary participation and the opt-out settings are identical. The budget sets, the returns, everything else is identical in these two problems, so we would expect households to make the same choices. In fact, however, participation in "opt-out" 401(k) plans is sometimes 40 percentage points higher than participation in voluntary participation plans.

Region: That much!?

Poterba: It's a huge difference. The effect tends to slowly damp out over time. The effects are largest in the first year or so when people are hired. Even in a voluntary program, people who have been with the firm longer do tend to migrate into the 401(k) plan over time, but even five years after workers are hired, there are still some effects of the automatic enrollment.

What the evidence suggests is that there are ways to affect participation in defined-contribution plans. I don't think the ways necessarily have to come from behavioral economics. Standard price-based approaches, such as matching contributions to plans, also work in increasing 401(k) participation and contribution rates.

Region: You've looked at life-cycle allocation strategies too.

Poterba: We've looked at how people might allocate their 401(k) savings over the course of their lives. The rise of life-cycle funds is a good example of how the financial services sector and retirement planners are trying to find ways of making this simpler for households and giving them ways to basically put the retirement saving process on autopilot. The designers of these funds are trying to recognize that many households do not want to think about financial decisions and are offering them ways to reduce the potential costs of financial inertia.

INSURANCE PRICING

Region: You've developed a significant body of work on asymmetric information in insurance markets, much of it with Amy Finklestein. In a recent paper with Amy and Casey Rothschild, "Redistribution by Insurance Market Regulation," you estimate the efficiency costs and distributive impact of regulations that prohibit insurers from looking at buyer characteristics in determining prices. Are the cost and impact quantitatively significant? In an age when genetic testing is growing ever more sophisticated, what is the policy import of such findings?

Poterba: I am fascinated by insurance markets and the contracts that are available to individuals. This strikes me as an important and somewhat understudied area. My work with Amy and Casey tries to understand what happens when regulations preclude insurance companies from using some information to set insurance prices.

There are many examples of such regulations. Many states limit the data that firms can use to price automobile insurance. We consider the market for retirement annuities. In the United States, a firm cannot offer different pension payouts to men and women with the same salary history and years of service, retiring at the same age, even though life expectancy for women is several years longer than that for men at typical retirement ages.

In most settings, a ban on using some information to price insurance transfers resources toward those whom this information would show to be high-risk insurance buyers. In the market for annuities, someone who is expected to live a long time is more expensive to insure than someone who is unhealthy and has a high mortality risk. An insurance company could offer to pay higher monthly benefits to those who are ill or infirm if it could identify them. Sometimes insurance regulations make this difficult or impossible and therefore represent a transfer from one mortality risk group to another.

One of the intriguing questions is the extent to which insurance companies can induce policy buyers to reveal something about their underlying mortality type through the creative design of insurance policies. In Britain, insurers offer both inflation-indexed and nominal annuities. The households who buy the inflation-indexed products, which deliver more of their value at advanced ages than nominal products, tend to live longer than those who buy nominal annuities.  This enables insurers to partly distinguish their client base. The extent to which such distinctions can be made is likely to vary across markets and settings.

NBER

Region: Let me ask you a last question, if I may, about the NBER. You're about to become president and CEO, stepping into the shoes of Martin Feldstein, who held the position for 30 years. Can you tell us your thoughts about directions that economic research should take in coming decades and where you plan to steer the Bureau?

Poterba: Let me start by saying it is an incredible honor and a humbling opportunity to be asked to lead an organization which has had such an enormous impact on the course of empirical research within the economics profession. It's an exciting opportunity and one I'm looking forward to.

The NBER under Marty has been a great place for economists from different universities, sometimes from different subfields within the field of economics, to come together for discussion and collaboration. The topics have ranged widely, and they have been of broad interest to economic policy analysts whether they're in academe, government or business. I hope that the NBER will continue to deliver top-quality research that will be of general interest and that we'll manage to respond nimbly to the interesting economic developments of the day. One of the NBER's great strengths is its ability to bring a whole range of economic researchers to bear on important questions as they emerge.

I hope to follow an opportunistic strategy in directing research. Sometimes this involves looking into the crystal ball and identifying issues and questions that are likely to become more important going forward. But other times it involves looking in the rearview mirror and saying, "What's just happened that's really interesting? Should we look more deeply into this?"

I would expect, for example, in the very near term I will encourage research on the role of risk in financial markets, the interplay between regulatory structure and risk-bearing by financial institutions, and the interplay between the financial services sector and the broader real economy. The economics profession needs to devote significant effort to understanding what happened during the financial market gyrations of late 2007 and 2008, to modeling the linkages between different parts of the financial sector and to explaining why contagion spread from one market to another in the way it did.

Part of what's important here is to make sure that policy analysts and researchers in universities are brought up to speed and learn what was happening in the financial markets so they're able to think cogently about the episode we've been through and focus their research on the right issues.

Once we've got that knowledge base, I think the next question is "what are the lessons?" as we try to think about the potential consequences of different regulatory structures. The NBER does not make policy recommendations. It tries to identify the consequences of policies, to think through "if you do this, then this might happen" without saying whether a particular policy is a good or a bad idea.

Region: Thanks so much for your time.

Interview conducted April 1, 2008.