The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Consumer Confidence Surveys: Can They Help Us Forecast Consumer Spending in Real Time? I by Dean Croushore n 1993, the Philadelphia Fed undertook a project to develop a real-time data set for macroeconomists, who can use these data in many ways — for example, when analyzing indexes of consumer confidence. Existing research indicates that consumer-confidence measures, though highly correlated with future spending, do not improve forecasts of future spending. But these studies used revised data that were not available to forecasters at the time they made their forecasts. In this article, Dean Croushore uses the real-time data set to investigate an important question: Does using data available to forecasters at the time — that is, real-time data — make measures of consumer confidence more valuable for forecasting? The Federal Reserve Bank of Philadelphia’s real-time data set for macroeconomists contains information on the data that a researcher or forecaster would have known at a date in the past. This data set, which Dean Croushore is an associate professor of economics and Rigsby Fellow at the University of Richmond. When he wrote this article, he was a visiting scholar in the Research Department of the Philadelphia Fed. This article is available free of charge at www.philadelphiafed. org/econ/br/index.html. www.philadelphiafed.org is available on the Philadelphia Fed’s website at www.philadelphiafed.org/ econ/forecast/reaindex.html, allows us to investigate a number of interesting economic and policy questions — one of which is the subject of this article. We will use the data set to investigate whether measures of consumer confidence help improve forecasts of consumer spending. For many reasons, people want to know how the economy is doing. They would like to answer questions such as: Are we in an economic expansion? Will the economic expansion continue? Are interest rates likely to rise or fall? To find answers to these questions, people read the newspapers, which often report on the forecasts of professional economists. The government and private-sector firms also report on a variety of economic data, which may include such items as a survey of consumer confidence. Several organizations take surveys of consumers to investigate what they say about the economy and their families’ finances. The survey responses are compiled and used to form an index of consumer confidence, which is reported in the news media. The consumer-confidence measures are correlated with changes in consumer spending, so they appear to capture useful information about consumers’ spending plans. But do they really help us forecast consumer spending in real time? In theory, the indexes should enable us to predict what consumers will spend in the future, and a glance at the data tells us that the consumerconfidence measures are, indeed, strongly correlated with consumer spending. But we are interested in seeing whether the consumer-confidence measures pass a tougher test: Do they tell us more than we already know from other economic data? If we look at the existing research, we see that the consumer-confidence measures, though highly correlated with future spending, do not improve forecasts of future spending made on the basis of knowing consumers’ incomes, past consumer spending, the interest rate, and the value of the stock market. However, that previous research (which we will discuss in more detail later) is flawed in one important aspect. The data used in those studies were not available to forecasters in real time, that is, at the time their forecasts Business Review Q3 2006 1 were made. Thoughtful researchers have long known that using such flawed data is not ideal, but they did not have a data set such as the realtime data set for macroeconomists until recently. The failure to use real-time data may be important because data are revised. For example, the Bureau of Economic Analysis (BEA), the government agency that releases data on consumer spending, revises the data many years after the fact. For example, when the BEA revises the data on consumer spending and income, it uses data from tax returns and Social Security records that no forecaster could have known earlier. These data are much more accurate than the government’s initial data on spending and income, which come from a very incomplete survey. If the revisions to the data on consumer spending and income are correlated with measures of consumer confidence, a forecaster in real time using measures of consumer confidence could make better forecasts than a forecaster who did not use measures of consumer confidence. So when previous researchers found that consumer-confidence indexes did not improve forecasts of consumer spending, they were not using the right data — no forecaster would have had the data they used. We will investigate the following question: If we used the data a forecaster would have had available in real time, would the measures of consumer confidence prove to be more valuable? Fortunately, the Philadelphia Fed’s real-time data set for macroeconomists allows us to undertake this exercise. That data set contains information on the data a researcher or forecaster would have known at a date in the past. As such, it contains exactly the data we need to investigate the realtime predictive power of consumerconfidence indexes. Q3 2006 Business Review DATA ON CONSUMER CONFIDENCE AND REAL-TIME DATA Consumer Confidence Surveys. The two most widely known surveys of consumer confidence are produced by the University of Michigan and the Conference Board. Both are similar in concept but implemented in different ways, and their use in forecasting models leads to somewhat different results. The University of Michigan’s survey contains about 50 questions, only five of which are part of its index of consumer sentiment. The survey, which began in 1946 on an occasional basis and has been taken monthly since 1978, is conducted with about 500 people via telephone. Consumers are asked five questions that reflect their sentiments about the economy and their family finances. Two questions reflect current economic conditions. The first question asks how people are getting along financially these days: Would you say that you (and your family living there) are better off or worse off financially than you were a year ago? The second question asks about the large items people buy, for example, furniture, appliances, or cars: Generally speaking, do you think now is a good or bad time for people to buy major household items? Three questions reflect future conditions: (1) Looking ahead, do you think that a year from now you (and your family living there) will be better off financially, or worse off, or just about the same as now? (2) Turning to business conditions in the country as a whole, do you think that during the next 12 months, we'll have good times financially or bad times, or what? (3) Looking ahead, which would you say is more likely: that in the country as a whole we'll have continuous good times during the next five years or so or that we will have periods of widespread unemployment or depression, or what? From the answers to these questions, the Michigan researchers create an index. For example, from question 1, they subtract the percentage of people who say they are worse off from the percentage of people who say they are better off. They calculate percentages in the same way for each of the other four questions. These percentages are averaged across all five questions then compared with the value in a base year (1966) that has been normalized to 100, and the result is the index of consumer sentiment. For our purposes in this article, we will call that index Michigan–overall. A separate index is created from the two questions about current conditions, which we will call Michigan–current, and an index is created from the three questions about future conditions, which we will call Michigan–future. The Conference Board creates its index of consumer confidence in a similar manner except the survey is mailed to 5,000 households, of which about 3,500 are returned. The survey has been conducted monthly since June 1977. As with the Michigan survey, the Conference Board’s survey asks five questions: two about current conditions and three about future conditions. Questions about current conditions are: (1) How would you rate the present general business conditions in your area? (2) What would you say about available jobs in your area right now? Questions about future conditions are: (1) Six months from now, do you think general business conditions will be better, the same, or worse? (2) Six months from now, do you think there will be more, the same, or fewer jobs available in your area? (3) How would you guess your total family income will be six months from now (higher, the same, or lower)? Again, similar to the University of Michigan, the Conference Board creates indexes, which we will call www.philadelphiafed.org CB–overall, from all five questions; CB–current, from the two questions on current conditions; and CB–future, from the three questions about future conditions. Although the Conference Board creates its index using a process similar to that used by Michigan, the base year for the Conference Board’s index is 1985, not 1966. Using Consumer Confidence to Forecast Consumer Spending. Figure 1 shows the values of the Michigan–overall and CB–overall indexes, plotted from January 1978 to December 2005.1 Gray bars indicate periods in which the economy was in a recession. As the figure indicates, the confidence indexes decline sharply at the start of recessions. Only for the 2001 recession did the confidence indexes decline several months before the recession began; that was the only time the indexes would have served as a leading indicator of a recession.2 Because the consumer confidence indexes do not appear to forecast recessions well, we examine their ability to forecast consumer spending instead. If measures of consumer confidence are able to forecast consumer spending, measures of consumer confidence should change before consumer spending does. The relevant data series for measuring consumer spending is known as personal consumption expenditures, which is collected by the Bureau of Economic Analysis as part Similar plots could be shown for the current and future indexes, but they are not included here to conserve space. For the same reason, Figures 2, 4, and 6 show only the CB—overall index. 1 The same is true for the future indexes, which are not shown, since they follow the same pattern as the overall indexes. 2 FIGURE 1 Consumer Confidence Indexes, Overall, January 1978 to December 2005 Index value 160 140 CB-Overall 120 100 80 Michigan-Overall 60 2005 2004 2003 2002 2001 2000 1999 1998 1997 1996 1995 1994 1993 1992 1991 1990 1989 1988 1987 1986 1985 1984 1983 1982 1981 1980 1979 1978 40 of the National Income and Product Accounts. The data we use are quarterly. Figure 2 plots the growth rate of consumption spending each quarter, measured as the amount of consumer spending within the quarter compared with the amount of spending in the same quarter of the previous year, along with the quarterly level of the CB—overall measure of consumer confidence. The graph indicates a fairly strong correlation between the growth rate of consumer spending and the measure of consumer confidence. Broadly speaking, consumer spending growth rises when consumer confidence rises, and vice versa. However, there are periods, such as 1987 to 1989, when the two variables appear to move in opposite directions. Nonetheless, it appears that the correlation is strong enough that we might be able to use consumer confidence to forecast consumer spending. Forecasting Model. We will construct a state-of-the-art forecasting model that has been used in previous research, and it is one that a forecaster could have used to predict consumer spending. Economic researchers have used this model in studies that have attempted to test whether consumer confidence indexes are helpful in forecasting. These studies include the paper by Jason Bram and Sydney Ludvigson and the one by Christopher Carroll, Jeffrey Fuhrer, and David Wilcox.3 We copy their forecasting model, which models the growth rate of consumer spending today as dependent on the growth of consumer spending in each of the last four quarters, the growth in people’s income in each of the last four quarters (because changes in income affect people’s decisions about how much they can spend), the change in the Date For a review of these and other studies, see Ludvigson’s 2004 paper. 3 Source: The Conference Board and the University of Michigan www.philadelphiafed.org Business Review Q3 2006 FIGURE 2 Conference Board Overall Index and Consumption Spending January 1978 to December 2005 Growth rate of consumption (percent) Index value 8 150 CB-Overall (left scale) 130 7 6 5 110 4 3 90 2 70 1 0 Growth Rate of Consumption (right scale) 50 -1 -2 2005 2004 2003 2002 2001 2000 1999 1998 1997 1996 1995 1994 1993 1992 1991 1990 1989 1988 1987 1986 1985 1984 1983 1982 1981 1980 1979 1978 30 Date Source: The Conference Board and Bureau of Economic Analysis interest rate (on three-month Treasury bills) in each of the last four quarters (higher interest rates induce people to save more and consume less), and the change in the value of the stock market in each of the last four quarters (increases in wealth induce people to consume more). Data on consumer spending, income, and the value of the stock market are in real terms; that is, they are adjusted for inflation. We will use this forecasting model as our baseline and then add a measure of consumer confidence to the model to see if we get improved forecasts.4 Data Revisions. One problem that an economic forecaster faces in practice is that data are sometimes incomplete and may be revised over For technical details on the forecasting models, see my 2005 paper, on which this article is based. 4 4 Q3 2006 Business Review time. To compare the models properly, we need to know what data a forecaster would have in real time. That is, to forecast what consumer spending would be during the first quarter of 1982, we must go back and examine the data a forecaster would have had available at that time, which may be quite different from what the data prior to the first quarter of 1982 look like today because of data revisions. To accomplish this task, we use the realtime data set for macroeconomists.5 Why are data revised? Mostly because the government makes an estiThe data set, available on the Federal Reserve Bank of Philadelphia’s website at philadelphiafed.org/ econ/forecast/reaindex. html, was first described in the Business Review article that I wrote with Tom Stark. See our other papers for further details on the data set and the implications of data revisions for economic research, forecasting, and monetary policy. 5 mate of the data before it has complete information. The government reports on many macroeconomic data series with a lag of just one month. For example, gross domestic product (GDP) for the first quarter of 2005 was first reported in April 2005. But the initial data release by the BEA is based on a very incomplete sample. Over time, the BEA gathers more information and revises the data, especially after people file their income tax returns. By July 2006, the BEA had a much clearer picture of what GDP was in the first quarter of 2005 than it did in April 2005. Thus, the revised data are significantly more accurate than the data that were initially released. But this poses a quandary for forecasters: Should they wait until the data have been revised, a process that takes over a year, or use what they have? The answer is clear for most situations: Forecasters need to forecast in the short run, and even the government’s initial release of the data is better than no data at all. Which variables do we need to worry about that might have data revisions? Consumer spending (more formally, real personal consumption expenditures) and income are revised over time by the government. In addition, we use the price index for personal consumption expenditures as our measure of inflation; so the real value of the stock market is revised when that price index is revised. The interest rate and the measure of consumer confidence are not revised. Thus, we need real-time data on consumer spending, income, and the price index, which are available in the real-time data set. How large are the revisions to the data series? Both consumer spending and income are revised substantially; however, the real value of the stock market is not revised very much. Figure 3 shows the revisions to consumer spending and income from when the data for each date were initially www.philadelphiafed.org FIGURE 3 Revisions to Real Consumption Growth and Real Income Growth Initial to February 2006 Database Percent (Annualized growth rate) 10 Consumption 5 0 -5 2005 2003 2001 1999 1997 1993 1991 1989 1987 1985 1983 1981 1979 1977 1975 1973 1971 1969 -15 1995 Income -10 Date Source: Author’s calculations from the real-time data set for macroeconomists released to the values recorded in the government’s database as of February 15, 2006. The numbers shown are the annualized growth rate6 for the quarter in the February 15, 2006, database minus the annualized growth rate for the quarter as reported by the government when the data were initially released. You can see that the data revisions can be large, reaching a magnitude of as much as 11.4 percentage points, and that revisions to income have generally been larger than revisions to consumption. An annualized growth rate is the growth rate from one quarter to the next, expressed at an annual rate so that comparisons with annual data can be easily made. For example, if GDP grew 0.6 percent from one quarter to the next, the annualized growth rate would be 2.4 percent — four times as large — because if GDP kept growing at the same pace for the entire year, it would grow 2.4 percent for the year. 6 www.philadelphiafed.org EVALUATING FORECASTS OF CONSUMER SPENDING Our model for forecasting consumer spending, as described above, uses data on past consumer spending, past income, past changes in the interest rate, and past changes in the real value of the stock market. At each date, beginning with the first quarter of 1982, we will imagine we are forecasters using the data available to us at the time. We will estimate our baseline model and generate a forecast for consumption spending in the quarter. Then, we will include a measure of consumer confidence in the model and generate another forecast. After following this procedure for the first quarter of 1982, we imagine stepping forward one quarter to the second quarter of 1982, with one additional quarter of data on which to base our forecasts. We will make forecasts for that quarter and then keep repeating this process through the fourth quarter of 2005. After following this procedure, we can show the forecasts for consumer spending each quarter, based on the baseline forecast with no consumer-confidence measure and the CB–overall forecast (Figure 4).7 As we can see in the graph, the forecasts are similar, but they also differ systematically at times; that is, the forecasts using the CB–overall index are higher than the baseline forecasts for many consecutive periods, such as most of the quarters from 1987 to 1990, and are lower than the baseline forecasts for most of the quarters from 1990 to 1991. How do we evaluate which forecast is better? To evaluate the forecasts for consumer spending, we will subtract the forecast made using the baseline model from the actual value of consumer spending in each quarter to calculate the baseline model’s forecast error. Next, we will do the same for the forecast made using the model that includes a measure of consumer confidence. Then, we will compare the forecast errors to see which model produces smaller errors. However, this raises a problem: What is the actual value of consumer spending? If we use today’s government database (in particular, the database as of February 15, 2006), we will probably find very large forecast errors in the earlier part of the sample period because of various changes to the definitions of the variables, changes in the base years for real variables, and so forth. This occurs because about every five years, the BEA modifies the methods it uses to construct the Forecasts for the other five measures of consumer confidence were also generated but are not shown here. 7 Business Review Q3 2006 5 FIGURE 4 Comparing Forecasts Over Time 1982Q1 to 2005Q4 Consumption growth rate Percent 14 12 CB-Overall 10 8 6 4 2 No Confidence Index 0 2005 2004 2003 2002 2001 2000 1999 1997 1998 1996 1997 1995 1996 1994 1993 1992 1991 1990 1989 1988 1987 1986 1985 1984 1983 -4 1982 -2 Date Source: Author’s calculations FIGURE 5 Alternative Actuals 1982Q1 to 2005Q4 Consumption growth rate 12 10 8 Feb. 15, 2006 Database 6 4 2 0 -2 Pre-Benchmark Date Source: Author’s calculations from the real-time data set for macroeconomists Q3 2006 Business Review 2005 2004 2003 2002 2001 2000 1999 1998 1995 1994 1993 1992 1991 1990 1989 1988 1987 1986 1985 1984 1983 -6 1982 -4 data in a process known as benchmark revision. It is hard to imagine that a forecaster working in early 1982 and making a forecast for consumer spending for the first quarter of 1982 could have anticipated the methods used by the government to calculate data on consumer spending as of early 2006. For that reason, we will not use the consumer spending data from the February 15, 2006, database as our measure of the actual value of the data. Instead, we will do the following: For each date for which a forecast is made, we will use as the actual value of the data the last value of the data before a benchmark revision. Benchmark revisions to the U.S. National Income and Product Accounts occurred in December 1980, December 1985, November 1991, January 1996, October 1999, and December 2003. Using the data just before a benchmark revision gives a better view of how accurate the forecasts are. How much does this choice matter? Figure 5 shows the data from the February 15, 2006, database compared with the data just before each benchmark release. Though the pattern of the growth rates of consumption spending is roughly the same, from 1982 to 1990 the pre-benchmark growth rate is almost always lower than the February 15, 2006, data. We would think that the forecasting model was making systematic forecast errors if we based our analysis on the most recent data instead of the pre-benchmark data. Figure 6 compares the forecast errors of the model that includes the CB–overall index with those of the baseline forecast. Since the graph shows there are times when each forecast error is higher or lower than the other, it is not obvious which forecast is worse. We need some way to compare the forecast errors over the entire period from 1982 to 2005. www.philadelphiafed.org relative RMSFE greater than 1, and a model with a lower RMSFE than the baseline model has a relative RMSFE less than 1. If a measure of consumer confidence was helpful in forecasting, its RMSFE would be less than 1. Table 1 also indicates whether the difference between the RMSFEs is statistically significant. None of the models has an RMSFE that is statistically significantly different from the baseline model. FIGURE 6 Comparing Forecast Errors Over Time 1982Q1 to 2005Q4 Forecast error (Percentage points) 6 4 2 0 -2 CB-Overall -4 -6 No Confidence Index 2005 2004 2003 2002 2001 2000 1999 1998 1997 1996 1995 1994 1993 1992 1991 1990 1989 1988 1987 1986 1985 1984 1983 -10 1982 -8 Date Source: Author’s calculations Economic theory provides a way to compare the forecast errors. We begin with the assumption that bigger forecast errors are substantially worse than smaller ones. A commonly accepted method of comparing forecast errors is to calculate the root-meansquared-forecast error (RMSFE). The RMSFE is found by squaring each forecast error (thus penalizing large errors more than small errors), adding the squared errors together, then taking the square root. The RMSFE is similar in concept to the standard deviation, which is commonly used in statistical analysis. The higher the RMSFE is, the worse the forecasts are. In addition, economists have developed tests for the statistical significance of differences in RMSFEs. For example, it could be that one forecasting model has a lower RMSFE than another, but www.philadelphiafed.org the difference between the two is so small that the result could have occurred by chance and, thus, does not mean that the one forecasting model is significantly better than the other. In each case, we will ask: Is the difference between the RMSFEs statistically significant? We compare the RMSFEs of the different forecasts in Table 1. As you can see, all of the forecasts using a measure of consumer confidence have higher RMSFEs than the baseline except for the forecast using CB–future. For ease of comparison, the table shows the relative RMSFE for each model, which is its RMSFE divided by the RMSFE of the baseline model with no consumer confidence measure. Thus, the baseline model has a relative RMSFE of 1, a model with a higher RMSFE than the baseline model has a ALTERNATIVE FORECASTING MODELS The results in Table 1 are discouraging. They suggest that none of the measures of consumer confidence help to significantly improve the forecasts, only one measure improves the forecasts at all, and the rest make the forecasts worse (though not significantly worse). However, our baseline model was based on models that other researchers in the literature had used. Those models were not necessarily designed to produce the best forecasts with real-time data. It might be possible to find a better forecasting model and then see if the measures of consumer confidence help improve the forecasts using that better model. One principle of forecasting is KISS (for example, see the references in Frank Diebold’s textbook on forecasting), which stands for Keep It Sophisticatedly Simple. In forecasting, this means that forecasters should use sophisticated models that capture the elements of the data that are essential to the process. But in comparing different sophisticated models, choose the simplest model that gets the job done. If a model is very complicated, it may suffer from data mining: Variables are included in the forecasting model because they help to explain a particular episode in the past, but they are of no value for forecasting the future and may, in fact, make such forecasts worse. Thus, we will try to simplify Business Review Q3 2006 7 TABLE 1 Root-Mean-Squared-Forecast Errors (RMSFE) Original Model, 1982Q1 to 2005Q4 Relative Significant Forecasting Model RMSFE RMSFE Difference? No confidence measure 2.16 1.000 --- M-overall CB-overall 2.28 2.17 1.055 1.004 no no M-future CB-future 2.28 2.13 1.055 0.988 no no M-current CB-current 2.40 2.26 1.114 1.048 no no TABLE 2 Root-Mean-Squared-Forecast Errors (RMSFE) Alternative Model, 1982Q1 to 2005Q4* Relative Significant Forecasting Model RMSFE RMSFE Difference? No confidence measure 2.11 1.000 --- M-overall CB-overall 2.18 2.19 1.033 1.035 no no M-future CB-future 2.22 2.18 1.051 1.033 no no M-current CB-current 2.23 2.25 1.053 1.065 no yes *Model uses changes in confidence indexes and fewer variables. the baseline model to see if we can make our forecasts better. One way to simplify the model is to eliminate some variables from the 8 Q3 2006 Business Review forecasting model. The only way to figure out the right variables to eliminate is by trial and error, and doing so results in slightly lower forecast errors. Essentially, all the information from the data on past income is already reflected in past consumption data, and the change in interest rates is simply not a very large factor affecting consumption. Therefore, we eliminate those two variables, and our forecasting model performs somewhat better. A second change that might help is to consider how the measures of consumer confidence should enter into our forecasting model. Following the previous researchers, we had initially used the level of consumer confidence in the forecasting model. But some people have suggested that what might be more helpful for forecasting is to note when there is a large change in consumer confidence, regardless of its level. A large increase in consumer confidence means people are likely to spend more, while a large decrease in consumer confidence means people are likely to spend less. We use only the change in a measure of consumer confidence in our model, not its level. We have simplified our forecasting model somewhat. The result, as shown in Table 2, is that our forecasts are slightly better — that is, the models generally have lower RMSFEs than those in Table 1 — except for CB—overall and CB—future. But the simplification of the model made the baseline model with no consumer confidence index slightly better. The result is that all of the measures of consumer confidence make the forecast worse, and one measure (CB–current) makes the forecasts significantly worse. The conjecture in the introduction suggested that by using real-time data, the measures of consumer confidence were more likely to be of help in forecasting than if we had used the revised data, for example, if we had pulled all the data out of the February 15, 2006, database. In fact, the use of real-time data did not make an appreciable difference in the forecasts www.philadelphiafed.org that used a consumer confidence index compared with the baseline model that did not. It appears that the use of realtime data did not rescue the consumerconfidence measures. chance of showing that measures of consumer confidence could prove useful in forecasting. After all, the measures of consumer confidence could reflect what people know that has not yet been captured by government statistical agencies. However, in trying to predict consumer spending, evidently the measures of consumer confidence reflect other events affecting the economy and do not sufficiently tell us what people know that government statistical agencies do not know. The bottom line: If you are forecasting consumer spending for the next quarter, you should use data on past consumer spending and stock prices and ignore data on consumer confidence. BR Bram, Jason, and Sydney Ludvigson. “Does Consumer Confidence Forecast Household Expenditure? A Sentiment Index Horse Race,” Federal Reserve Bank of New York Economic Policy Review (June 1998), pp. 59–78. Croushore, Dean, and Tom Stark. “A Funny Thing Happened on the Way to the Data Bank: A Real-Time Data Set for Macroeconomists,” Federal Reserve Bank of Philadelphia Business Review (Sept./Oct. 2000), pp. 15–27. Ludvigson, Sydney. “Consumer Confidence and Consumer Spending,” Journal of Economic Perspectives 18 (Spring 2004), pp. 29–50. Carroll, Christopher D., Jeffrey C. Fuhrer, and David W. Wilcox. “Does Consumer Sentiment Forecast Household Spending? If So, Why?” American Economic Review 84 (December 1994), pp. 1397–1408. Croushore, Dean, and Tom Stark. “A Real-Time Data Set for Macroeconomists,” Journal of Econometrics 105 (November 2001), pp. 111–30. SUMMARY The conjecture that began this article seemed sensible: The use of real-time data might have a better REFERENCES Croushore, Dean. “Do Consumer Confidence Indexes Help Forecast Consumer Spending in Real Time?” North American Journal of Economics and Finance 16 (December 2005), pp. 435–50. www.philadelphiafed.org Stark, Tom, and Dean Croushore. “Forecasting with a Real-Time Data Set for Macroeconomists,” Journal of Macroeconomics 24 (December 2002), pp. 507−31. Diebold, Francis X. Elements of Forecasting, 3rd edition. Cincinnati: South-Western, 2003. Business Review Q3 2006 9 A Review of Inflation Targeting in Developed Countries by michael dotsey I n the United States, inflation targeting has many advocates, but many others are skeptical about adopting such a policy. Given this debate and inflation targeting’s growing adoption around the world, now is a good time to review the economic performance of some inflation-targeting countries. In this article, Mike Dotsey examines five countries that have been targeting inflation for at least 10 years and whose inflation rates, though fairly well contained before inflation targeting, were nonetheless considered too high by policymakers. For purposes of comparison, he also looks at the economic performance of six noninflation-targeting countries. Many countries have debated the merits of inflation targeting, and some have adopted inflation targeting as a national policy. In an inflationtargeting framework, a central bank announces quantitative targets for inflation and specifies that controlling inflation is a long-run goal of monetary policy. Another common feature is a Mike Dotsey is a vice president and senior economic policy advisor in the Research Department of the Philadelphia Fed. This article is available free of charge at www. philadelphiafed.org/econ/br/index.html. 10 Q3 2006 Business Review specific policy for bringing inflation back to target in circumstances where the target has been missed. Also, inflation-targeting central banks have often adopted a more transparent policy that entails fairly detailed communications with the public.1 New Zealand first instituted this monetary policy framework in early 1990. Since that time, 22 countries have formally adopted inflation targeting, and no country that has adopted it has abandoned it. Although inflation targeting’s contribution to overall economic performance is still being debated, the general view is that it has had beneficial effects. Inflation and its volatility have generally declined in inflation-targeting countries and output growth has increased. At the same time, it appears that the volatility of output has decreased. Inflation targeting has many advocates in the United States. Those in favor base their argument on the economic benefits that ensue from low and stable inflation, the possibility that inflation targeting would enhance the FOMC’s credibility, and the increased flexibility that could come from increased credibility.2 Others, however, are not quite as enthusiastic, and their reservations largely involve concerns over a lack of flexibility that might result from inflation targeting, especially in situations where maintaining a tight rein on inflation could prove damaging to the economy. Critics also point out that U.S. monetary policy has performed quite well over the last 20 years without any formal reliance on an inflation target and that it may be a bit too early to fully evaluate the relative performance of inflation-targeting countries. Inflation targeting’s track record is rather short, and we may not yet have seen situations where adhering to an inflation target would be detrimental to economic performance. Given the status of the debate over inflation targeting, the recent interest in this topic in the United States, and its growing adoption by countries around the world, now seems For a representative viewpoint, see speeches by Anthony Santomero in Federal Reserve Bank of Philadelphia Business Review, Third Quarter 2003 and Fourth Quarter 2004. 2 See the book by Bernanke and co-authors for a formal description of inflation targeting. 1 www.philadelphiafed.org an opportune time to review the policies and economic performance of a number of inflation-targeting countries. To carry out this evaluation, I will examine a set of countries — New Zealand, Canada, Australia, the United Kingdom, and Sweden — that have been targeting inflation for at least 10 years and whose inflation rates were fairly well contained before they adopted inflation targeting. This choice helps avoid issues that arise if we choose countries that went from high inflation to low inflation after targeting, and the experience of the countries chosen is more relevant from the standpoint of the United States. For purposes of comparison, I will examine the economic performance of those five countries along with the performance of six noninflation-targeting countries: the United States, Germany, Japan, France, Italy, and the Netherlands. I will also summarize some recent empirical studies on inflation targeting. The view taken here is that inflation targeting has been modestly beneficial. Inflation has indeed declined in the five countries examined. Inflation volatility has also declined and, perhaps just as important, so has the inertia in the inflation process itself. Furthermore, expectations of inflation seem to be more stable in the inflationtargeting countries, lending credence to the assertion that inflation targeting enhances a central bank’s credibility. Also, there seem to be no negative consequences for economic activity. Output growth has tended to be stronger and less volatile over the time that these countries have targeted inflation. Moreover, central bankers in these five countries have publicly expressed enthusiasm for the framework. KEY FEATURES OF INFLATION TARGETING Inflation targeting establishes a www.philadelphiafed.org numerical objective for inflation, and the actual target is stated as either a specific point target or a range. Although the framework explicitly acknowledges that maintaining a low inflation rate is a primary policy objective, inflation need not be the sole objective of monetary policy. When inflation is not the central bank’s energy components of the headline measure. These less volatile measures are typically referred to as the core inflation measure of a price index. Also, should the target be a point or a range, and over what period should inflation be measured? Should monthly inflation be targeted, or should it be some long-run average of inflation? In addi- Importantly, in an inflation-targeting framework, monetary policy is delegated to an independent central bank. only concern, the other concerns are often stated in the central bank’s explicit mandate, or the central bank may communicate its concerns more informally to the public. Importantly, in an inflation-targeting framework, monetary policy is delegated to an independent central bank. By establishing numerical objectives that are to be met over specified periods, an inflation-targeting framework embeds accountability. The issue of accountability has led to very open communication between inflation-targeting central banks and the public. This increased openness is called transparency. The combination of transparency and specific numerical objectives makes monitoring inflation targeting easier for the public. This openness also helps central banks establish credibility because it is easier to judge whether they are meeting their commitments. Adherence to the general framework of inflation targeting requires the monetary authority to pay attention to a number of particular elements. One key element involves picking a particular measure of inflation. There are numerous measures of inflation, ranging from the headline measure of the consumer price index (CPI) to various less volatile measures of inflation, which typically exclude the food and tion, if the target is a point target, the central bank must decide how much of a deviation from its target it is willing to tolerate. For example, should the implicit or explicit range be just a few percentage points or wider? Other equally important considerations involve whether the central bank should have multiple goals, such as a target for output growth or the unemployment rate, and how a central bank can be held accountable for its actions. Communication and transparency become quite important in an inflation-targeting regime. But how transparent should the central bank be? The various approaches of the five inflation-targeting countries examined in this article are summarized in Table 1. As we can see, approaches to inflation targeting vary.3 However, although there are a number of differences, there are some key commonalities. Most have a point target but are content to let inflation vary within plus or minus 1 percent of the target. Also, most central banks currently target an annual average of the headline CPI, but they also report the behavior Excellent summaries of inflation targeting can be found in the book by Edwin Truman and the book by Bernanke and co-authors. 3 Business Review Q3 2006 11 TABLE 1 Inflation-Targeting Framework of Five Countries Transition period to reach final target Time frame to correct deviations Country Date Type of mandate New Zealand Dec. 1989 Price stability Range of 1-3% CPI over medium term Yes Not explicit Quarterly monetary policy statement No: Target set by agreement between government and bank Canada Feb. 1991 Multiple 2 pctage points of CPI with ±1% tolerance Yes 6-8 quarters Quarterly monetary policy report No: Target set by government and bank U.K. Oct. 1992 Hierarchy with price stability first 2 pctage points of CPI with ± 1% tolerance Yes Not specific, but required to set horizon each instance Quarterly inflation report No: Target set by government Sweden Jan. 1993 Price stability 2 pctage points of annual CPI with ±1% tolerance 1-2years ahead. No Yes 1-2 years Quarterly inflation report Yes Australia June 1993 Multiple Range of 2-3% CPI over medium term No No time frame Quarterly statement on monetary policy Yes Setting Communication Independence Source: Pooled from various materials (see references). of core measures. Many, but not all, report a time path for bringing inflation back to target if the target is missed. Finally, all of the inflation-targeting 12 Q3 2006 Business Review central banks are quite transparent and issue frequent and detailed communications concerning policy. The only major difference is with respect to independence. Although all independently set interest rates, only two out of the five have sole responsibility for setting the ultimate goals of policy. www.philadelphiafed.org TABLE 2 Inflation and Output Growth in Inflation-Targeting Countries, Before and After* Country Pre-inflation Targeting (10 years prior to adopting target; for dates see Table 1) inflation growth 11.4 1.8 2.9 2.7 2.1 3.0 1.8 2.4 Canada 5.7 2.8 2.9 2.9 2.0 2.7 1.3 2.1 U.K. 5.5 2.5 3.0 1.9 2.5 2.9 0.8 0.7 Australia 6.0 3.2 2.9 2.7 2.6 3.8 1.6 1.1 Sweden 6.7 1.9 2.9 2.3 1.5 2.5 1.1 1.6 Avg. IT 7.1 2.4 2.9 2.5 2.1 3.0 1.3 1.6 NZ s.d. inflation s.d. growth Post-inflation Targeting - 2004 (for dates see Table 1) inflation growth s.d. inflation s.d. growth Inflation and Output Growth in Noninflation-Targeting Countries, Comparison 1982-1992 1992-2004 U.S. 4.0 3.0 1.3 2.6 2.5 3.3 0.6 1.2 Japan 1.9 3.7 1.1 1.8 0.1 1.1 1.0 1.6 Germany 2.6 2.7 1.7 5.3 1.8 1.1 1.4 1.4 France 5.1 2.2 3.3 1.1 1.6 1.9 0.6 1.3 Neth. 2.6 2.5 2.3 2.2 2.4 2.4 0.8 1.6 Italy 8.3 2.2 4.6 1.3 3.0 1.5 1.3 1.4 Avg. NIT 4.1 2.7 2.4 2.4 1.9 1.9 1.0 1.4 * Inflation rates are annualized changes in the headline CPI and growth rates are annualized rates of growth in GDP. www.philadelphiafed.org Business Review Q3 2006 13 EXPERIENCE UNDER INFLATION TARGETING Now let’s compare the experience of the five inflation-targeting countries with that of the six noninflation-targeting countries. These six countries serve as a reference, preventing me from attributing various economic outcomes to inflation targeting when, in fact, these outcomes may be a result of global economic conditions. For example, output volatility declined in all 11 of the countries from 1992 to 2004, the years in the latter half of my sample. Attributing the entire decline in the inflation-targeting countries to inflation targeting would be erroneous. Inflation targeting should be viewed as helping to lower output volatility only if inflation-targeting countries experience a greater decline than noninflation-targeting countries. First, let’s look at data on inflation and output growth for both the five inflation-targeting countries and the six noninflation-targeting countries. For the inflation-targeting sample, we’ll use data for the 10 years before the adoption of inflation targeting and from adoption to the end of 2004. For the noninflation-targeting countries, the first sample of data covers 1982 to 1992, and the second sample covers 1992 to 2004. This methodology allows a visual comparison of the data before adoption of inflation targeting and after. The data are shown in figures 1 and 2 and summarized in Table 2. The first thing to notice is that with the exception of Italy, the inflation-targeting countries had higher inflation rates in the first part of the sample, while output growth was fairly comparable across the two groups. Therefore, it is evident that the U.S., Japan, Germany, and the Netherlands had less incentive to adopt inflation targeting, since their inflation rates were already fairly low. In the second half of the sample, both sets of coun- 14 Q3 2006 Business Review tries have similarly low rates of inflation, and high inflation is not deemed to be a problem for any of the countries.4 Thus, it appears that inflation targeting is associated with a lowering of inflation for all five countries that adopted it but that central banks can also achieve low inflation without explicitly targeting inflation. However, to gauge the effectiveness of inflation targeting, we want to examine the comparative differences in behavior of the two groups of countries over the two samples. Some of the the level of inflation because those country-specific factors are assumed to be the same across the two sample periods. By looking at differences across the two sample periods, we can remove that level effect. In that regard, the graphs show that both sets of countries saw a reduction in inflation, but, on average, the reduction was greater for inflation-targeting countries. In those countries, average annual inflation rates declined 5 percentage points as opposed to 2.2 percentage points for the noninfla- Noninflation-targeting countries tend to be larger countries and may be more immune to the effects of changes in international prices. noninflation-targeting countries may have specific circumstances that allow them to more easily keep inflation low. For example, the noninflationtargeting countries tend to be larger countries and may be more immune to the effects of changes in international prices. We do not want to conclude that inflation targeting is ineffective when two countries have similarly low inflation, one an inflation-targeting country and the other one not, since that outcome may occur because inflation targeting was helpful in the country that adopted it but was less needed in the country that didn’t. To avoid this confusion, I concentrate on differences in inflation and output growth across the two groups of countries and across the two sample periods. Doing so cancels out factors specific to a particular country that may affect Japan, on the other hand, suffered from deflation and a very sluggish economy in the second half of the sample. It is possible that Japan would have benefited from inflation targeting because it would have forced the country to have a more expansionary monetary policy. 4 tion-targeting countries. Also, output growth increased by an average annual rate of 0.6 percentage point in the inflation-targeting countries but actually declined 0.6 percentage point in the noninflation-targeting countries. Next, let’s look at the relative variability of inflation and output over the two sample periods (Figure 2). Excluding Italy, the variability of inflation, as measured by the standard deviation of annualized growth rates in the headline CPI, is greater for the inflationtargeting countries before the adoption of inflation targeting. After adoption, the average volatility of inflation fell a dramatic 1.6 percentage points for the inflation-targeting countries, but it also fell 1.4 percentage points for the noninflation-targeting countries. As in the case of the decline in the inflation rate, much of the decline in volatility in the noninflation-targeting countries occurred because France and Italy became part of the European Currency Union and one of the requirements for joining the union was a low and stable inflation rate. Thus, there was institutional pressure for France and Italy www.philadelphiafed.org FIGURE 1 Inflation Rates, Output Growth, and Inflation Targeting Inflation Inflation-Targeting Countries Noninflation-Targeting Countries Percent change in CPI Percent change in CPI 20 20 Australia Canada New Zealand Sweden UK 15 10 10 5 5 0 0 -5 -5 Q1 1982 Q1 1987 Q1 1992 Q1 1997 US Germany France 15 Q1 1981 Q1 2002 Q1 1986 Q1 1991 Japan Italy The Netherlands Q1 1996 Q1 2001 Output Inflation-Targeting Countries Noninflation-Targeting Countries Percent change in GDP Percent change in GDP 10 10 8 8 6 Japan Italy The Netherlands US Germany France 6 4 4 2 2 0 Australia Canada New Zealand Sweden UK -2 -4 -2 -4 -6 Q1 1982 0 Q1 1987 www.philadelphiafed.org Q1 1992 Q1 1997 Q1 2002 Q1 1982 Q1 1987 Q1 1992 Q1 1997 Q1 2002 Business Review Q3 2006 15 FIGURE 2 Inflation and Output Average Growth vs. Average Inflation Average percent change in CPI 12 10 8 IT countries 10 years pre-IT non-IT countries pre-1992 6 IT countries post-IT non-IT countries post-1992 4 2 0 0 1 2 3 4 5 Average percent change in GDP Standard Deviations of Growth & Inflation SD of percent inflation 5 4.5 4 3.5 3 to reduce their inflation rates in the second half of the sample.5 Output volatility also declined by nearly the same amount for both groups of countries, and both groups show a more favorable tradeoff between output and inflation volatility over the later sample (Figure 2). Just looking at the data in this way is informative, but it has limitations. It doesn’t control for many features of the economic environment that affect inflation and output and that may be unrelated to inflation targeting. The sample size is still very small, making any definitive conclusion statistically difficult.6 For example, countries may have experienced more favorable economic shocks after they adopted inflation targeting, and merely examining economic outcomes before and after adoption could overstate the benefits of inflation targeting.7 Using a control group of countries helps avoid attributing all of the improvement in inflation and output performance to the inflation-targeting framework, but it also brings its own set of interpretation problems. In particular, there are some important differences between the inflation-targeting countries and the control group. The inflation-targeting countries are small, open economies, whereas the control group contains IT countries pre-IT non-IT countries pre-1992 2.5 IT countries post-IT 2 non-IT countries post-1992 1.5 1 One of the convergence criteria of the Maastricht Treaty was for countries to have a rate of inflation that was less than a maximum of 1.5 percent above the average rate in the three EU countries with the lowest inflation. 5 For example, more sophisticated statistical work analyzing whether inflation targeting reduces the variability of inflation is inconclusive. 6 0.5 0 0 1 2 3 4 SD of percent GDP growth 16 Q3 2006 Business Review 5 6 An economic shock is a factor that causes unexpected changes in economic variables. Shocks can be unfavorable, such as the devastating economic effects of hurricanes, or favorable, such as an innovation in technology that increases productivity. 7 www.philadelphiafed.org some economically large countries.8 Thus, the inflation-targeting countries have a greater exposure to external economic disturbances. Also, the control group may have adopted better monetary policy over the most recent period, perhaps implicitly targeting inflation, and by doing so would diminish the benefits attributable to inflation targeting. It could also be the case that the countries that adopted inflation targeting just had relatively bad luck before adoption and now have experienced a more normal set of economic shocks. If that were the case, we might incorrectly attribute the benefits of a change in luck to inflation targeting. It also doesn’t allow us to examine another important aspect of inflation targeting, namely, its effect on inflation expectations. A MORE DETAILED EXAMINATION To more sharply assess the potential benefits of inflation targeting, let’s review the economic literature regarding the empirical effects associated with inflation targeting.9 In general, it appears that adopting inflation targeting has reduced the inflation rate and the persistence of inflation, has stabilized long-run expectations of inflaA country is said to have an open economy if it engages in significant trade with other countries. tion, and has not had any deleterious effects on output. However, inflation targeting has probably not had any significant effect on inflation volatility. The Inflation Rate. One of the primary reasons for moving to inflation targeting is to reduce inflation. From Figure 1 and Table 2, it is clear that inflation did decline after the adoption of inflation targeting. But it also declined for countries that did not adopt inflation targeting. Thus, distinguishing between the experience of countries that target inflation and those that do not requires more sophisticated statistical techniques. The basic message from such exercises is mixed. A number of studies indicate that inflation targeting was successful in reducing inflation, but that conclusion is sensitive to how the study controlled for the fact that many inflation-targeting countries had relatively high inflation before they introduced inflation targeting. It also depends on the countries in the particular study. An interesting but controversial study by Laurence Ball and Niamh Sheridan attributes all the statistically significant lowering of inflation to the fact that inflation-targeting countries initially had higher inflation, and therefore, all one sees is a regression to the mean.10 However, other studies 8 The results are largely taken from these five papers: David Johnson; Andrew Levin, Fabio Natalucci, and Jeremy Piger; Laurence Ball and Niamh Sheridan; Refet Gurkaynak, Andrew Levin, and Eric Swanson; and Marco Vega and Diego Winkelried; as well as evidence presented in the book by Edwin Truman. These studies were chosen because they are some of the most recent and therefore have the longest data sets. They also do a relatively good job of controlling for the experience of noninflation-targeting countries. All of the studies include the five inflation-targeting countries highlighted above, but many include a wider control group and, occasionally, other, more recently industrialized countries, such as Spain, Finland, and Norway, that have adopted inflation targeting more recently. 9 www.philadelphiafed.org This is an important but controversial result. Many of the noninflation-targeting countries, such as the United States, placed greater emphasis on controlling inflation during the 1990s, and a number of European countries aligned their monetary policy with Germany’s. Although Germany does not meet the strict definition of inflation targeting, it is widely recognized that inflation has always been a key concern of the Bundesbank. Accurately gauging the effects of inflation targeting requires significant independent variability across inflation-targeting countries. It is likely that we have insufficient data for making a sharp distinction between a simple regression toward the mean and the independent effects that inflation targeting has in lowering inflation in this set of countries. For a detailed exposition of this point, see Mark Gertler’s discussion of the Ball and Sheridan paper. 10 that included nonindustrialized countries have found significant benefits of inflation targeting in terms of lowering inflation in both industrialized and nonindustrialized countries.11 In his book, Edwin Truman notes that in this wider set of countries, there is little correlation between past inflation and the adoption of inflation targeting. The problem from the standpoint of the United States is that the experience of many of these countries may not be very relevant for understanding how inflation targeting would affect the United States. Fortunately, other types of data allow for a better understanding of the effects of inflation targeting. Expected Inflation. One measure that appears to behave differently between industrialized inflation-targeting countries and industrialized noninflation-targeting countries is expected inflation. A primary motivation for adopting inflation targeting is to both reduce and stabilize expectations of inflation. If inflation targeting can accomplish this, then, in theory, it can reduce the tradeoff between lowering inflation and the loss of output. A large set of economic models imply that when individuals expect a higher inflation rate than the rate the central bank is targeting, employment falls because higher expectations of inflation lead to higher wage demands. When these expectations are unrealistically high, the higher wages cause firms to hire fewer workers and employment falls. Firms may also set prices too high, thereby reducing the demand for their products. Both factors, which originate from erroneous views The work of Marco Vega and Diego Winkelried uses econometric techniques from the treatment literature as well as a wide sample of noninflation-targeting countries, while Truman tries to uncover the effects of inflation targeting by including a wealth of control variables in his analysis. 11 Business Review Q3 2006 17 of targeted inflation, reduce output. Examining the same set of countries as in this article, Johnson finds that inflation targeting has lowered inflation expectations. Thus, adopting inflation targeting helped those central banks coordinate the public’s expectations of inflation with the targeted rate.12 Interestingly, he also finds that up until the fifth year of pursuing inflation targets, the effect of inflation targeting on reducing expected inflation gets progressively larger with each year under the inflation-targeting regime. After adhering to inflation targeting for five years, the effect gradually dissipates. This result makes sense because the credibility of a country’s desire to lower inflation would be expected to increase over time. The longer a central bank sticks to inflation targeting, the more confident the public becomes that the change in policy is permanent. Further, the lowering of expected inflation induced by adopting inflation targets varies from country to country: New Zealand enjoys the largest decline and the U.K. the least. It is interesting that New Zealand has the strictest inflation contract with the Governor of the Reserve Bank, who is solely responsible for outcomes and subject to dismissal, while at the onset of inflation targeting, the Bank of England did not have operational independence. In complementary work that also sheds light on the behavior of inflation expectations, Andrew Levin and his co-authors examine the behavior of survey expectations of inflation in our group of inflation-targeting countries and those in a control group consisting of the U.S. and Japan and a European Johnson used survey measures of inflation. For all but the United States, they were taken from Consensus Forecasts. For the U.S., he used expectations from the Survey of Professional Forecasters, which is conducted by the Federal Reserve Bank of Philadelphia. 12 18 Q3 2006 Business Review average composed of Germany, France, Italy, and the Netherlands. They find that expectations of long-run inflation (five years and 10 years) are not influenced by current inflation (measured by an average of inflation over the last three years) in inflation-targeting countries, whereas long-run inflation expectations respond to changes in actual inflation in the control group. Thus, inflation expectations seem better anchored under inflation targeting. increase inflation in the near term may not be offset in the future. In that case, current economic surprises could affect expectations of long-run inflation. Unfortunately, there are only three countries for which this type of experiment can be carried out: the U.S., the U.K., and Sweden. Based on data from 1999-2005, when all three countries’ inflation rates had fairly well stabilized, only in the U.S. are expectations of long-run inflation sensitive to Some economists prefer measures of inflation expectations derived from financial market data to those derived from surveys. Some economists prefer measures of inflation expectations derived from financial market data to those derived from surveys. Investors in bonds face serious losses if they misjudge future inflation. Refet Gurkaynak and colleagues at the Federal Reserve Board carried out research using an alternative measure of inflation expectations. They looked at the difference in yields between long-term government bonds and long-term government bonds indexed for inflation. The difference in yields gives a market expectation of inflation. They tested to see how sensitive long-run expected inflation is to unexpected economic developments, which are measured as the difference between actual reports of various economic statistics and survey forecasts of those statistics a few days before their release. If investors are confident that the central bank will adhere to its inflation objectives, the fluctuations in current economic data should not influence investors’ beliefs about longrun inflation. However, if investors believe that the monetary authority is not committed to controlling long-run inflation, economic disturbances that unexpected economic news. Thus, inflation expectations appear to be better anchored in the U.K. and Sweden than in the U.S. An interesting result in their study is that before 1997, when the Bank of England obtained operational independence from the government, long-run expectations of inflation in the U.K. were also sensitive to economic news. Whether this result is due to operational independence providing more credibility or the fact that the Bank of England had established more credibility over time is an open question. The Persistence of Inflation. One effect of inflation targeting should be to reduce the persistence of deviations in inflation from its target because any deviations of inflation from target are gradually offset, whereas there is no explicit requirement that a noninflation-targeting central bank do so. This potential benefit of inflation targeting finds support in a couple of studies.13 However, the extent to which persistence is diminished varies across the two studies. One indicates that The relevant papers are the ones by Vega and Winkelried and Levin and co-authors. 13 www.philadelphiafed.org the effect of inflation targeting is quite large, while the other finds it to be rather small. The difference in results could be attributable to the use of different noninflation-targeting countries in the two studies, but more work and perhaps better data are needed as well. Output. While it appears that inflation targeting in general has had economically beneficial effects on the behavior of inflation, it would be difficult to find political support for inflation targeting if, at the same time, it had deleterious effects on output. Truman finds that industrialized inflationtargeting countries experience both an increase in output growth and a reduction in output volatility relative to the experience of noninflation-targeting countries. His first analysis finds that inflation targeting raises growth and lowers the variance of growth rates. His second experiment directly tests whether the changes in relative growth rates over the two samples (pre- and post-inflation targeting) between inflation-targeting and noninflationtargeting countries are significantly different. He finds that the increase in growth in inflation-targeting countries was significantly higher than the increase in growth in noninflationtargeting countries. Similarly, he finds that the decrease in the volatility of www.philadelphiafed.org output growth was significantly greater for the inflation-targeting countries. Conclusions A number of countries have implemented inflation targeting, and it has been in effect in a few of these countries for more than 10 years. The exact nature of the inflation-targeting framework differs across countries, and in most countries, it has evolved over time. As expressed in their testimony and speeches, monetary policymakers in the five inflation-targeting countries examined in this article all seem to be pleased with the results and have found the framework flexible enough to allow consideration of economic performance. There is no indication that inflation targeting has diminished economic performance in countries that have adopted it relative to the performance of other industrialized countries. Indeed, there is some evidence that inflation targeting has been associated with a reduction in inflation and that expectations of inflation are more stable in countries that have adopted inflation targeting. Further, inflation targeting appears to be compatible with robust economic activity. While the empirical evidence on the effects of inflation targeting is encouraging, we must acknowledge that the data that lend themselves to this optimistic view are limited. The experiment of inflation targeting has proceeded for a fairly short time, and thus, it has probably not been subject to all the vagaries that economies can experience. However, the testimony of central bankers who have been responsible for guiding monetary policy in the five inflation-targeting countries has been overwhelmingly positive.14 Many cannot envision departing from their current practices and returning to regimes that were less explicit about underlying inflation goals. They point to numerous instances where having an inflation target both focused monetary policy and made it easier to conduct. BR Examples of the enthusiasm that inflationtargeting central banks have for inflation targeting can be found in several places: See the comments pertaining to the Canadian experience by the Governor of the Bank of Canada, Gordon Thiessen, and those describing the Australian experience by the Governor of the Australian Reserve Bank, Ian J. Macfarlane. Also, a favorable opinion of inflation targeting can be found in a speech by the Governor of the Reserve Bank of New Zealand, Donald Brash, delivered at the AEA meetings in 2002. Mervyn King, the Governor of the Bank of England, has also eloquently discussed the benefits of inflation targeting. For comments by members of the Riksbank, who have viewed their experience with inflation targeting favorably, see the article by Claes Berg. 14 Business Review Q3 2006 19 REFERENCES Archibald, Joanne. “Independent Review of the Operation of Monetary Policy: Final Outcomes,” Reserve Bank of New Zealand Bulletin, 64, 3, pp. 4-14. Ball, Laurence, and Niamh Sheridan. “Does Inflation Targeting Matter?” in The Inflation Targeting Debate. National Bureau of Economic Research Studies in Business Cycles, 32, Chicago: University of Chicago Press, 2005. Bank of England. Monetary Policy Framework, available at: www.bankofengland. co.uk/monetarypolicy/framework.htm Barker, Kate. “Monetary Policy in the U.K.,” speech, National Association for Business Economics, Washington, D.C., March 21, 2005. Berg, Claes. “Inflation Forecast Targeting: The Swedish Experience,” Sveriges Riksbank Quarterly Review, 3, 1999, pp. 44-70. Berg, Claes. “Experience of Inflation Targeting in 20 Countries,” Sveriges Riksbank Quarterly Review, 1, 2005, pp. 20-47. Bernanke, Ben S., Thomas Laubach, Frederic Mishkin, and Adam Posen. Inflation Targeting: Lessons from the International Experience. Princeton: Princeton University Press, 1999. Brash, Donald T. “Inflation Targeting 14 Years On,” Reserve Bank of New Zealand Bulletin, 65, 1, pp. 58-70 (speech, American Economic Association, January 5, 2002). Dodge, David. “Inflation Targeting: A Canadian Perspective,” speech, National Association for Business Economics, March 21, 2005 (www.bankofcanada. ca/en/speeches/2005/sp05-2.html). Gertler, Mark. “Comments,” in The Inflation Targeting Debate. National Bureau of Economic Research Studies in Business Cycles, 32, Chicago: University of Chicago Press, 2005. 20 Q3 2006 Business Review Gurkaynak, Refet, Andrew T. Levin, and Eric T. Swanson. “Inflation Targeting and the Anchoring of Long-Run Expectations: International Evidence from Daily Bond Yield Data,” manuscript, Board of Governors of the Federal Reserve System, June 2005. Johnson, David R. “The Effect of Inflation Targeting on the Behavior of Expected Inflation: Evidence from an 11 Country Panel,” Journal of Monetary Economics, 49 (November 2002), pp 1521-38. Levin, Andrew T., Fabio M. Natalucci, and Jeremy Piger. “The Macroeconomic Effects of Inflation Targeting,” Federal Reserve Bank of St. Louis Review, 86 (July/August 2004), pp. 51-80. King, Mervyn. “The Monetary Policy Committee: Five Years On,” speech to the Society of Business Economists, available at: www.bankofengland.co.uk/ publications/speeches/2002/speech172.pdf Kuttner, Kenneth N. “A Snapshot of Inflation Targeting in Its Adolescence,” paper, Reserve Bank of Australia, available at: www.rba. gov.au/PublicationsAndResearch/ Conferences/2004/Kuttner.pdf Macfarlane, Ian J. “Australia’s Experience with Inflation Targeting,” in Stabilization and Monetary Policy: The International Experience, Banco de Mexico (November 2000). Mishkin, Frederic. “From Monetary Targeting to Inflation Targeting: Lessons from the Industrialized Countries,” in Stabilization and Monetary Policy: The International Experience, Banco de Mexico (November 2000). Reserve Bank of New Zealand. “What Is the Policy Targets Agreement?” Fact Sheet No.3, RBNZ, available at: www.rbnz.govt. nz/monpol/pta/0127027.html. Santomero, Anthony M. “Flexible Commitment or Inflation Targeting for the U.S.?,” Federal Reserve Bank of Philadelphia Business Review, Third Quarter 2003. Santomero, Anthony M. “Monetary Policy and Inflation Targeting in the U.S.,” Federal Reserve Bank of Philadelphia Business Review, Fourth Quarter 2004. Sherwin, Murray. “Institutional Framework for Inflation Targeting,” speech, Bank of Thailand symposium on “Practical Experiences on Inflation Targeting,” October 20, 2000 (http://www.rbnz.govt. nz/speeches/0097459.html). Stevens, Glenn. “Inflation Targeting: A Decade of Australian Experience,” Reserve Bank of Australia Bulletin (April 2003), pp. 17-29. Svensson, Lars E.O. “Independent Review of the Operation of Monetary Policy in New Zealand: Report to the Minister of Finance,” February 2001. Thiessen, Gordon. “The Canadian Experience with Inflation Targeting,” in Stabilization and Monetary Policy: The International Experience, Banco de Mexico (November 2000), pp. 85-90. Truman, Edwin M. Inflation Targeting in the World Economy. Institute for International Economics, Washington, D.C., 2003. Twaddle, James. “The Reserve Bank of New Zealand Amendment Act 2003,” Reserve Bank of New Zealand Bulletin, 67, 1, pp. 14-33. Vega, Marco, and Diego Winkelried. “Inflation Targeting and Inflation Behavior: A Successful Story?,” manuscript, February 2005. www.philadelphiafed.org Residential Mortgage Default by ronel elul A dramatic expansion of mortgage credit in recent years, coupled with a rapid run-up in house prices, has focused the attention of pundits and policymakers on the risks of home mortgage lending. In this article, Ronel Elul discusses the models that economists have developed to help us understand the default risk inherent in home mortgages and how default risk and house prices are related. He also applies these models to show how falling house prices would affect mortgage default rates today and explores the impact that rising default rates would have on financial institutions and other participants in the mortgage market. Although default rates on residential mortgages have been relatively low in recent years, policymakers and economists should still be concerned about mortgage default for several reasons. First, while the foreclosure rate in the U.S. has averaged only 1 percent over the past 20 years, there have been dramatic swings in regional default rates over this period. For example, in the early 1990s foreclosure rates in California rose fivefold, from less Ronel Elul is a senior economist in the Research Department of the Philadelphia Fed. This article is available free of charge at www. philadelphiafed. org/econ/br/index. html. www.philadelphiafed.org than 0.4 percent to nearly 2 percent. In addition, this jump in default rates coincided with a 25 percent drop in house prices in California. One reason to be concerned about mortgage default is the prominent role that mortgages play in our financial system. First, home mortgages represent the bulk of credit extended to consumers. According to data collected by the Federal Reserve, mortgages make up over $8 trillion of the $10 trillion in consumer debt outstanding. Second, defaults on mortgages affect not only homeowners but also the holders of the mortgages. These obviously include the original lenders, which are primarily banks and thrifts. In addition, however, mortgage-backed securities (MBS) distribute this risk throughout the entire economy; indeed, some estimates show that one- quarter of all mortgages are ultimately held by investors in MBS.1 In addition, the risk of default is currently of particular concern because of the rapid run-up in house prices in recent years. Although many scenarios are feasible, one possible outcome is a significant decline in many housing markets across the U.S. Both policymakers and market participants certainly need to be able to quantify the effect of falling house prices on mortgage default rates. Fortunately, economists have developed optiontheoretic models that permit us to understand the default risks inherent in home mortgages and how they relate to house prices.2 According to these models, homeowners simply compare their house value to their remaining debt when deciding whether to default. While the simplified view of the world that option-theoretic models present provides useful insights, in practice, other considerations also influence a household’s decision about whether or not to default on its mortgage. THE OPTION-THEORETIC APPROACH TO MORTGAGE DEFAULT The Ability to Default on a Source: Mortgage Market Statistical Annual (2004). 1 In addition to facing default risk, an investor in mortgages also faces prepayment risk. This is the risk that a borrower will pay a mortgage before its maturity, and the investor will have to find a new place to invest his funds. Since prepayment often occurs through refinancing the mortgage at a lower rate, it is usually disadvantageous for the lender. Not surprisingly, the primary factor that determines prepayment risk is the current level of interest rates relative to the rate when the mortgage was issued. 2 Business Review Q3 2006 21 Mortgage Can Be Viewed as a Put Option. One way to think about the risk that a homeowner will default on his mortgage is to view default as an option available to the homeowner. In general, an option is a contract in which one party obtains the right to buy or sell some underlying asset to another party for a prespecified price, known as the “strike,” or exercise, price. When the party has the right to buy the asset at a fixed price, the contract is known as a call option; if he has the right to sell the asset, it is a put option. The most prominent example is a stock option (Figure 1). Consider the case of a put option on IBM stock with a strike price of $75. If IBM is trading at $50 per share, exercising such an option would give the holder the right to sell a share of IBM for $75, for a profit of $25. When the exercise of the option is profitable, the option is said to be in the money.3 By contrast, it would not be profitable to exercise a put option with a strike price of $75 if IBM were trading at $80, since the strike price is below the current market price. In such a case the option is said to be out of the money. Figure 1 plots the profit an investor would earn from this put option as a function of the price of IBM stock, assuming that a rational investor would not exercise the option when it is out of the money. In the case of a mortgage, the homeowner’s ability to default can also be viewed as a put option. Should the homeowner default, he is in effect “selling” the house to the lender for the current mortgage balance. When the house value is lower than the mortgage balance (commonly termed negative equity), the borrower gains financially if he stops paying the mortgage, surrenders the house to the lender, and buys a similar house for less than the mortgage balance. This corresponds to “selling” the house to the lender for the mortgage balance, since the borrower essentially gains the difference between the mortgage balance and the value of the house.4 What We Learn from the Option-Theoretic Approach. Setting the default decision in this sort of framework is very fruitful because economists know a lot about how to value options. Indeed, the pioneering work of Fisher Black and Myron Scholes and that of Robert Merton developed a methodology that enables us to calculate a precise numerical value for very general types of options. One Michael Asay was the first to formally model mortgage default as an option. For an overview of more recent literature, see the article by James Kau and Donald Keenan. 4 appeal of their approach is that it leads to a formula that depends on only a few variables, which can be measured. In the case of the mortgage default option, these variables are the current loan-to-value (LTV) ratio, the mortgage amortization schedule (that is, the monthly schedule of how the mortgage balance is paid down), the volatility of house prices, and interest rates. Lenders can use option-pricing formulas to determine how high an interest rate they must charge in order to compensate them for the risk of default. Investors in mortgage-backed securities can also use these formulas to determine how much these securities are worth. Finally, regulators and economists interested in mortgage default can use these formulas to gauge the risk that a given drop in house prices might pose to lenders. We will perform an exercise of this type later. Viewing the right to default as an option also gives us qualitative insights FIGURE 1 Payoff to the Holder of a Put Option Payoff 75 50 25 0 Of course, it may be preferable to wait longer to exercise the option in the hope that the stock price and the profit from exercising the option go even higher before the option expires. 3 22 Q3 2006 Business Review 0 25 75 50 100 125 Stock Price www.philadelphiafed.org into mortgage default that might not otherwise be apparent. For example, options are more valuable when the underlying asset is more volatile. Consider the case of an investor holding a put option. Such an option will be in the money, i.e., profitable to exercise, when the asset price is below the strike price. When the asset price is more volatile, it is more likely to take both high and low values. This means that the option is more likely to be in the money (and by larger amounts). However, the greater likelihood of a very high asset price doesn’t lead to a counteracting loss because the holder of the put option will choose not to exercise the option when the asset price is higher than the strike price. Thus, viewing the right to default on one’s mortgage as a put option suggests that more volatile house prices should be associated with both a greater incidence and a greater severity of default. The study by James Kau and Donald Keenan has confirmed this. Finally, the option-theoretic model also serves as a useful conceptual framework for extending our knowledge further. By testing this model, economists are able to assess the extent to which it accurately describes homeowners’ behavior and, when it does not, to determine ways in which the model may be improved. EMPIRICAL TESTS OF THE OPTION-THEORETIC MODEL As we have discussed, one appeal of the simple option-theoretic approach is that it is parsimonious: Only a few factors play a role, most notably home equity.5 Empirical testing of the option-theoretic model has confirmed the important role played by home Home equity is defined here as the difference between the value of a house and that of all loans secured by the house. 5 www.philadelphiafed.org equity.6 It has also provided evidence that the homeowner’s option is more complex than the simple model suggests. In addition, empirical work has uncovered evidence that default decisions also depend on factors outside the framework of an option-theoretic model. Economists Extend the Model in Light of Empirical Findings. One important finding uncovered by testing of the option-theoretic model is that homeowners do not appear to default as soon as their equity becomes negative. In their 1985 study, Chester Foster and Robert Van Order found One appeal of the simple option-theoretic approach is that it is parsimonious: Only a few factors play a role, most notably home equity. For the most part, empirical work has focused on fixed-rate mortgages, in particular, those made to borrowers with good credit histories, known as prime loans. As the name suggests, the payment on these mortgages is fixed (in nominal terms) over the life of the mortgage. In addition, the borrower is typically permitted to refinance (prepay) the mortgage, for example, if interest rates drop.7 See, for example, the article by Yongheng Deng, John Quigley, and Robert Van Order. 6 While a perfectly general analysis would take into account other types of mortgage products — most notably adjustable rate mortgages (ARMs) and subprime mortgages (which are loans made to riskier borrowers with poor credit histories) — we can still learn a lot by restricting our attention to prime fixed-rate loans. Despite the recent growth of other types of mortgages (particularly subprime loans), prime fixed-rate loans still represent approximately two-thirds of all outstanding mortgages, and models for subprime loans are in an earlier stage of development. In addition, the main factors affecting default risk in prime fixed-rate mortgages are shared by other types of mortgages as well. For example, the risk from falling prices affects all types of mortgages. Nonetheless, we should be cautious in drawing general conclusions about the mortgage market as a whole from studies of prime fixed-rate mortgages alone because other types of mortgages have additional risk factors. For example, borrowers with ARMs are also exposed to the risk that interest rates will rise in the future, causing their required monthly payment to go up. Subprime borrowers are at greater risk for job loss than prime borrowers, which puts them at greater risk of default in response to a regional downturn that affects both housing prices and labor markets. 7 that even when the LTV rises to as much as 110 percent, only 4.2 percent of borrowers in their data set default. They suggest that this is evidence against a simple option-theoretic model in which homeowners default as soon as the equity in their house is negative. Other researchers have argued, however, that homeowners’ behavior is still well described by the option pricing model if we extend the simple model to account for the panoply of options available to the homeowner. In particular, some economists point out that the mortgage default option is essentially an “American” option, which the holder can exercise at any time up to its maturity. In contrast, a European option can be exercised only at a single prespecified date. We have already observed that it may not be optimal to exercise a put option on a stock as soon as the stock price dips below the strike price; one may prefer to wait in case it falls further. Similarly, if the house price is slightly below the mortgage balance, a fully rational homeowner may prefer to wait to default in order to give house prices a chance to fall further, making default even more profitable. Kau, Keenan, and Taewon Kim construct plausible numerical examples that show that it may be optimal to wait to default until the house price is as much as 15 Business Review Q3 2006 23 percent below the mortgage value. Another reason that a rational homeowner may not default when it may appear to be optimal is that he actually has another option: prepaying his mortgage (for example, by refinancing).8 This option may be viewed as a call option on the mortgage, since in prepaying the mortgage, the homeowner is taking the opportunity to buy back his outstanding debt by paying the remaining balance.9 These two options interact. If someone has already prepaid his mortgage, he obviously cannot default. Similarly, someone who anticipates that he will refinance his mortgage shortly might decide that it is not worthwhile to default, since he does not plan to pay on the current mortgage for much longer. A recent paper by Yongheng Deng, John Quigley, and Robert Van Order tests the extent to which mortgage default is driven solely by negative equity. They find that although negative equity is indeed an important determinant of default behavior, the existence of a prepayment option does have a statistically significant impact on the default decision. That is, a homeowner who is very likely to prepay his mortgage (for example, if his mortgage interest rate is much higher than current rates) is also less likely to default. Similarly, they also find that the default option has a significant impact on the exercise of the prepayment option; that is, households likely to default tend to prepay less often. Empirical Work Also Points to Factors Outside the Option-Theoretic Framework. Other economists argue, however, that the reason homeowners do not default as soon as their equity turns negative is that defaulting involves significant transaction costs. For example, defaulting on a mortgage entails moving and losing one’s home.10 The impact that a default has on a borrower’s reputation (for example, his credit score) may also be viewed as a form of transaction cost, since the defaulter sends a negative signal to potential lenders, a situation that makes any future borrowing more costly and difficult.11 Finally, some borrowers may also have moral qualms that make them more reluctant to default. All of the borrower is self-employed, also help explain default behavior.12 By contrast, recall that in the option-theoretic model, only variables directly related to the mortgage or house value should matter.13 These findings are consistent with the plausible hypothesis that at least some homeowners are liquidity constrained; that is, a borrower cannot borrow freely against his expected future income or wealth.14 Consider the example of a homeowner who loses his job but knows he is likely to find a new one in the near future. Suppose that he would like to continue paying Researchers have also found evidence that variables that capture crisis or “trigger” events for households, such as unemployment rates and divorce rates, all seem to lead to defaults. these may be viewed as factors outside the option-theoretic framework, which assumes that homeowners optimize in a perfectly frictionless manner or, at least, that transaction costs are small enough to be ignored. Researchers have also found evidence that variables that capture crisis or “trigger” events for households, such as unemployment rates and divorce rates, all seem to lead to defaults. Similarly, personal characteristics of the homeowner associated with greater income risk, such as whether his mortgage so as to retain his home but that he has no equity in the house against which to borrow. If he could find a lender willing to lend on his assurances that he will find a new job, and if he could commit to repay the loan from this as yet unrealized future income, he would be able to borrow enough to continue making his mortgage payments during this temporary spell of unemployment. In practice, however, it is likely to be difficult to find a lender willing to lend under these circumstances, and the home- In their 1984 study, Foster and Van Order were the first to find evidence that these costs have an impact on the default decision. 12 10 See the article by Kerry Vandell and Thomas Thibodeau. While it is fairly straightforward to test for the impact of trigger events empirically, incorporating them into a theoretical model requires a framework that focuses on consumer decisions, rather than a simple modification of the option pricing approach. See the paper by Peter J. Elmer and Steven A. Seelig for an example. 13 This would imply that borrowers with lower credit scores, who thus have less of a reputation to protect, would be likelier to default. This has been confirmed by several studies, for example, the one by Anthony Pennington-Cross. But note that low credit scores are also associated with less access to credit and riskier income; so this evidence is also consistent with theories (discussed below) that relate default to credit constraints. 11 Note that while the prepayment option is nearly universal for prime mortgages, this is not necessarily the case for subprime loans. 8 In addition, he would also have to pay any costs associated with prepaying, for example, closing costs, if he were to refinance his mortgage. 9 24 Q3 2006 Business Review Many studies find evidence of liquidity constraints in other arenas; see, for example, the article by Tullio Jappelli. 14 www.philadelphiafed.org owner may well be forced to default. Further support for the existence of liquidity constraints can be found in the paper by Deng, Quigley, and Van Order. First, these authors confirm that high state unemployment and divorce rates are associated with a higher incidence of default. Second, they find that higher initial loan-to-value ratios are associated with greater default risk. This finding is also consistent with the existence of liquidity constraints, since borrowers who have less wealth available for a down payment are likelier to be constrained. Last, these authors also find support for the existence of transaction costs that discourage homeowners from defaulting. Finally, in addition to transaction costs and liquidity constraints, state laws may also affect homeowners’ default behavior. (See State Laws and Mortgage Default.) EMPIRICAL MODELING OF MORTGAGE DEFAULT Competing Risks Models: An Empirical Framework for Modeling Mortgage Default. One framework researchers use to test the option-theoretic model of mortgage default and to assess the significance of additional variables is the proportional hazard model. D.R. Cox first applied this model in the biomedical sciences,15 where it was used to study the effect of various treatments on patients’ survival. The proportional hazard model explains the likelihood of exiting the sample in the next instant of time, given that the patient has survived up to this time. For example, it has been used to explain mortality from cancer, given the patient’s age, gender, treatment history, and whether the patient is a smoker. Proportional hazard models have also been applied extensively to explain mortgage default. 15 See the book by D.R. Cox and D. Oakes. www.philadelphiafed.org State Laws and Mortgage Default I n principle, the existence of state laws governing mortgage default (in particular, those laws that govern deficiency judgments) may also impede the free exercise of homeowners’ default option. Some states prohibit lenders from pursuing deficiency judgments, which means that they cannot try to collect any deficiency between the value of the house and the mortgage balance from the homeowner’s other assets. In principle, this would make defaulting on a mortgage more attractive for a homeowner with negative equity. Despite considerable effort, economists have uncovered little evidence that laws that prohibit deficiency judgments make homeowners more likely to default. The reason may be that deficiency judgments are rare even when they are permitteda because the defaulting homeowner is unlikely to have many assets aside from his house and because even in states where deficiency judgments are permitted, the homeowner may often protect himself against them by filing for bankruptcy.b a See the article by Charles Capone. For more on the empirical significance of these laws, see the article by Karen Pence and the one by Terrence Clauretie and Thomas Herzog. b As we have discussed above, however, the homeowner typically has another option as well, which is to prepay his mortgage. In light of this, the model by Deng, Quigley, and Van Order uses an extension of the proportional hazard model with two “competing risks”: default and prepayment. In this case, the mortgage will terminate when the borrower either prepays or defaults, whichever occurs first. This extension allows them to study the interaction between default and prepayment and to estimate the relative significance of trigger events such as unemployment and divorce rates.16 Predicting Default Rates in a Hypothetical Housing Market Downturn. One immediate application of the models we have presented is to forecast default rates in a hypothetiThis model is also used in other areas of economics. For example, someone may leave unemployment either because he finds a job or because he drops out of the labor force altogether. 16 cal downturn in the housing market. This is obviously of interest to policymakers. The scenario we consider is motivated by the work of Joshua Gallin. He argues that, based on an analysis of historical rent-price ratios, housing is currently overvalued by more than 20 percent. One way to understand this is to note that given today’s house prices and rents, a savvy homeowner could profit by selling his house, investing the money in a relatively safe asset such as long-term Treasury bonds, and using the interest income to rent a comparable house.17 He would profit because at today’s inflated prices, his interest income would exceed his rent. This process may not necessarily be as straightforward as we describe. In particular, it is not always easy to find comparable rental accommodation. Indeed, Gary Smith and Margaret H. Smith argue that if one carefully matches owner-occupied and rental housing, prices do not appear to be out of line relative to rents in most cities. 17 Business Review Q3 2006 25 Such selling pressure would tend to lower house prices, and the increased demand for rental units might also raise rents. This process would continue until all such opportunities for easy profits are exhausted. At this point, the market would be in equilibrium.18 Gallin finds that when prices are high relative to rents — as in the past few years — there has indeed been a tendency for this equilibrium relationship to be re-established. Figure 2 shows the rent-price ratio since 1970.19 Observe first that in late 2005 this index was at its lowest level since 1970; in addition, periods in which this ratio moved away from its long-run mean (roughly 100) appear to be followed by reversals. Gallin also shows that this adjustment process generally involves both rents rising more rapidly than usual and prices rising more slowly (or even falling). In particular, assuming that housing is overvalued by 20 percent, Gallin’s work predicts that over the next three years, real rents20 should rise about 1.2 percent per year faster than usual, and real house prices should rise 3.4 percent per year more slowly than usual. Gallin’s argument that the housing market is out of equilibrium is statistical; that is, he compares the rent-price ratio to its historical average. He makes no conjectures as to why the market moves out of equilibrium in the first place. Furthermore, although Gallin finds evidence that this adjustMore precisely, according to this argument, the equilibrium price of a house should be roughly equal to the present value of the expected future income one could earn by renting out the house, after adjusting for taxes and maintenance. 18 The rent-price ratio is constructed by dividing the rent index from the CPI-W (reported by the Bureau of Labor Statistics), by Freddie Mac’s conventional mortgage home price index; we make several minor adjustments as suggested by Gallin. ment has taken place in the past, this does not necessarily mean that it is certain to occur in the future because the equilibrium house-price–rent ratio may have permanently changed for various reasons. In particular, an argument often made is that the current high level of prices (relative to rents) can be justified because financial innovations have made borrowing easier and cheaper. For example, increased subprime lending allows households to buy homes when they would previously have been forced to rent. This increased demand for owner-occupied housing should raise house prices relative to rents.21 To examine the potential impact of price declines on mortgage default, we will consider a more extreme trajectory for house prices than the one suggested by Gallin. We will begin with a benchmark case in which prices increase at a steady 4 percent a year.22 However, rather than stagnation in prices, as Gallin suggests, we then consider the impact of an immediate 20 percent drop in prices (followed by 4 percent growth thereafter). While such a scenario is admittedly extreme,23 it nevertheless provides useful insights by establishing bounds on the possible impact of mortgage default. We also consider a more conservative scenario. We use the empirical model of Yongheng Deng and John Quigley to generate forecasts of mortgage deThis is consistent with the average real rate of increase in house prices over the past 30 years. That is, adjusting for inflation, the average rate of increase has been 1.5 percent a year. Given a current inflation rate of roughly 2.5 percent, we arrive at a 4 percent nominal rate of increase. 22 Indeed, the homeownership rate in 2005 was at a historical high of 69 percent. This view was articulated by Janet Yellen, president of the San Francisco Fed, in a speech on October 21, 2005. 21 However, there were drops of roughly this magnitude in New England and California in the early 1990s. 23 FIGURE 2 Rent-Price Ratio: 1970-2005 Rent-price ratio (1996: Q1 = 100) 120 110 100 90 80 19 20 That is, after adjusting for inflation. 26 Q3 2006 Business Review 70 1970 1973 1977 1980 1984 1987 1991 1994 1998 2001 2005 Source: Bureau of Labor Statistics and Freddie Mac www.philadelphiafed.org fault rates under these scenarios. We consider a representative homeowner who has just taken out a mortgage at an interest rate of 6 percent (which we assume is also the current market interest rate) and who has an initial LTV of 80 percent. According to data from the 2004 Survey of Consumer Finances, the fraction of homeowners with LTVs at or below 80 percent is 80 percent.24 Further detail on the distribution of LTVs is presented in Figure 3. Aside from the contemporaneous loan-to-value ratio, which we can calculate from the initial LTV and the interest rate, the other variables used in the model are the volatility of house prices, state unemployment rates, and state divorce rates. We also assume that interest rates are constant; so given that the mortgage is taken out at the market interest rate, there is no reason for homeowners to prepay their mortgages.25 For the benchmark scenario of an 80 percent LTV mortgage, the risk of default over the 360-month life of the mortgage is about 1.8 percent.26 Figure 4 plots the cumulative default rate as a function of time (in months). Now consider an instantaneous drop in house prices of 20 percent just after the mortgage has been taken out, so that this mortgage now has an LTV of 100 percent. Over the life of the mortgage, the default rate, at 6 percent, is over three times as high as the benchmark scenario because even a small drop in house prices in the future will lead to negative equity. As can be seen in the figure, most of the acceleration in default rates comes in the early years of the mortgage, before amortization lowers the LTV significantly. Once the LTV has fallen, the value of the option to default declines substantially. It is also useful to explore what happens for less dramatic scenarios. If house prices decline only 10 percent, for example, lifetime default rates increase from 1.8 percent to 3 percent. So a decline in prices that is twice as large (20 percent as compared to 10 percent) results in default rates that are three times as large. In other words, drops in housing prices have a nonlinear effect on default rates, with large declines increasing default rates more than proportionally. This nonlinearity can also be seen in Figure 4; observe that the default rates corresponding to a 10 percent drop are much closer to those with no drop than they are to those when prices drop 20 percent. Since we saw in Figure 1 that the price of an option does not have a linear relationship to the price of the underlying asset (because of the option holder’s right to not exercise the option when prices fall), it is not surprising that drops in house prices have a similarly nonlinear effect on default rates. The reason is that if the value of a house falls only slightly below the outstanding mortgage balance, the homeowner will be unlikely to default on his mortgage, since there is a significant FIGURE 3 LTV Distribution for Those with Mortgages These figures for LTVs include first mortgages as well as home equity loans. 24 We assume house price volatility of 11.5 percent, following the study by John Campbell and Joao Cocco. Unemployment and divorce rates are set at roughly current U.S. levels: 5 percent and 4.8 percent, respectively. We calculate the default rates by simulating many paths for house prices under our assumptions, use these paths to calculate the probability of negative equity in every period, and then apply the model in Deng and Quigley to these simulated probabilities. Deng and Quigley’s model is closely related to that in the published paper by Deng, Quigley, and Van Order; it has the advantage for us that only publicly available data are required to generate predictions. 25 This is higher than the 1 percent foreclosure rate we reported at the start of the paper. The difference may be attributed to the fact that actual prices have risen more rapidly in the past than our scenario specifies, as well as the fact that we impose assumptions that rule out prepayment. 26 www.philadelphiafed.org Fraction of borrowers 0.9 0.8 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.11 0.1 0.08 0.01 0 LTV < 80% 80%< LTV < 90% 90% < LTV < 100% 100% < LTV Source: 2004 Federal Reserve Board Survey of Consumer Finances Business Review Q3 2006 27 likelihood that the house’s value will rise above the mortgage balance in the near future. By contrast, for large drops in prices, default will be much likelier, since equity will still be negative even if prices go up in the future.27 Although homeowners gain financially when they exercise their option to default in the face of falling house prices, this gain obviously comes at the expense of other market participants. The incidence of losses is also of interest to economists and regulators. (See Who Is Hurt When Homeowners Default?) SUMMARY One of the risks to mortgage lending is that the homeowner will default on his promise to continue making payments. One of the primary drivers of mortgage default is declines in house prices. Economists have developed option-theoretic models that can quantify the impact that falling prices have on mortgage default. These models have had some success in explaining homeowners’ defaults; however, there is evidence that they fail on three dimensions. First, they do not recognize that default is costly, which makes homeowners more reluctant to stop paying. Second, they do not account for the fact that some homeowners are credit constrained, so that if they experience a “trigger event,” such as a job loss, they may not be able to continue paying on their mortgage even if they expect to find new employment in the near future; this increases the risk of default. Finally, homeowners may be less reluctant to default than is suggested by the option-theoretic models because they also have another option: prepaying their mortgage. As a result, economists have developed empirical models that seek to account for mortgage default through a combination of explanatory variables, both ones related to home equity and ones that account for transaction costs, trigger events, and the prepayment option. We have seen that such models can be used to predict the effect that falling prices would have on mortgage default rates. Further research is needed on the determinants of default for newer mortgage products, such as subprime loans, as well as the impact of default on other market participants, particularly investors in MBS. BR FIGURE 4 Mortgage Default Rates for Three Scenarios Default rate 0.07 0.06 0.05 0.04 0.03 0.02 0.01 Default rates are similarly nonlinear in LTVs. For example, a 20 percent drop in prices would have a negligible effect on a borrower with an initial LTV of 60 percent, raising his lifetime default rate from 1.1 percent to 1.3 percent. By contrast, for a borrower with an initial LTV of 100 percent, the default rate would rise from 23 percent to nearly 100 percent. 27 28 Q3 2006 Business Review 0 0 100 200 300 Months 100% 90% 80% www.philadelphiafed.org Who Is Hurt When Homeowners Default? T here are four main parties that are exposed to the risk of homeowners defaulting on their mortgages. Banks and thrifts hold approximately 30 percent of all home mortgages. Although banks would obviously take significant losses if prices fell dramatically, and some might even find themselves under severe stress, the banking sector as a whole is currently well capitalized and could sustain a drop of the magnitude we considered in the text. In particular, depository institutions have approximately $850 billion in capital, against liabilities of $9.6 trillion. Of these liabilities, no more than $2.75 trillion are nonguaranteed mortgage loans of some sort (first mortgages, home equity loans, and private mortgagebacked securities). To determine the impact of falling prices on banks, we need information on the LTVs of the mortgages in their portfolios; we will make the simple assumption that the distribution of LTVs for those loans held by banks is roughly the same as that for the population of mortgages as a whole (see Figure 3 in the text). In this case, an application of our model allows us to conclude that the default rates that banks experience on their mortgage portfolios would rise roughly 2 percent (over and above the current U.S. foreclosure rate of 1 percent) within one year of a 20 percent price decline. Given the currently sound state of banking institutions, this would not appear to pose a dramatic risk to the stability of this sector.a Of those mortgages not held by depository institutions, the vast majority are packaged into mortgagebacked securities (MBS). Most are “agency MBS”: They are backed by a government-sponsored enterprise (GSE), most notably Fannie Mae and Freddie Mac. So investors in these securities are protected against default. The GSEs themselves bear very little credit risk, however, because they require private mortgage insurance (PMI) for borrowers with LTVs above 80 percent. Thus, the vast majority of the default risk on agency MBS falls on the PMI industry, which insured approximately 13 percent of all conventional mortgages issued in 2004.b In addition, approximately one-quarter of all MBS are “private-label MBS,” which are not backed by any agency. Although these securities feature some sort of credit enhancement to mitigate the risk of default, this protection is typically incomplete, so that investors generally end up bearing some default risk. These investors include hedge funds, life insurance companies, pension funds, and private individuals. The extent to which these participants are exposed to mortgage credit risk and the degree to which this risk is concentrated in a few entities are unknown, and further research on this issue would be instrumental for policymakers. This has not always been the case. In particular, some people have suggested that declines in the value of banks’ real estate portfolios led to a “credit crunch” that aggravated the recession of the early 1990s. See the article by Joe Peek and Eric Rosengren. a Source: micanews.com and HMDA data. This represents a trend down from earlier years, since financial innovations such as “piggyback loans” have reduced the importance of PMI. b www.philadelphiafed.org Business Review Q3 2006 29 REFERENCES Asay, Michael. “Rational Mortgage Pricing,” Board of Governors of the Federal Reserve System, Research Papers in Banking and Financial Economics, 30 (1978). Black, Fisher, and Myron Scholes. “The Pricing of Options and Corporate Liabilities,” Journal of Political Economy, 81 (May/June 1973), p. 637. Campbell, John Y., and Joao F. Cocco. “Household Risk Management and Optimal Mortgage Choice,” Quarterly Journal of Economics, 188 (November 2003), pp. 1449–94. Capone, Jr., Charles A. “Providing Alternatives to Foreclosure: A Report to Congress,” U.S. Department of Housing and Urban Development (March 1996). Capozza, Dennis R., and Thomas A. Thomson. “Optimal Stopping and Losses on Subprime Mortgages,” Journal of Real Estate Finance and Economics, 30 (March 2005), pp. 115-31. Clauretie, Terrence M., and Thomas Herzog. “The Effect of State Foreclosure Laws on Loan Losses: Evidence from the Mortgage Insurance Industry,” Journal of Money, Credit and Banking, 22 (May 1990), pp. 221-33. Cox, D. R., and Oakes, D. Analysis of Survival Data. London: Chapman and Hall, 1984. Deng, Yongheng, John Quigley, and Robert Van Order. “Mortgage Terminations, Heterogeneity, and the Exercise of Mortgage Options,” Econometrica, 68 (March 2000), pp. 275-307. 30 Q3 2006 Business Review Deng, Yongheng, and John Quigley. “Woodhead Behavior and the Pricing of Residential Mortgages,” University of Southern California Lusk Center for Real Estate, Working Paper 2003-1005 (revised February 2004). Elmer, Peter J., and Steven A. Seelig. “Insolvency, Trigger Events, and Consumer Risk Posture in the Theory of SingleFamily Mortgage Default,” Journal of Housing Research, 10 (1999), pp. 1-25. Foster, Chester, and Robert Van Order. “An Option-Based Model of Mortgage Default,” Housing Finance Review, 3 (October 1984), pp. 351-72. Foster, Charles, and Robert Van Order. “FHA Terminations: A Prelude to Rational Mortgage Pricing,” AREUEA Journal, 13 (1985), pp. 273-91. Gallin, Joshua. “The Long-Run Relationship Between House Prices and Rents,” Board of Governors of the Federal Reserve System, Finance and Economics Discussion Series 2004-50 (2004). Jappelli, Tullio. “Who Is Credit Constrained in the U.S. Economy?” Quarterly Journal of Economics, 105 (February 1990), pp. 219-34. Kau, James B., Donald C. Keenan, and Taewon Kim. “Default Probabilities for Mortgages,” Journal of Urban Economics, 35 (May 1994), pp. 278-96. Merton, Robert C. “Theory of Rational Option Pricing,” Bell Journal of Economics and Management Science, 4 (1973), pp. 41-83. Peek, J., and Eric Rosengren. “Capital Crunch: Neither a Borrower Nor a Lender Be,” Journal of Money, Credit and Banking 27 (1995), pp. 625-38. Pence, Karen M. “Foreclosing on Opportunity: State Laws and Mortgage Credit,” Review of Economics and Statistics, 88 (February 2006), pp.177-82. Pennington-Cross, Anthony. “Credit History and the Performance of Prime and Nonprime Mortgages,” Journal of Real Estate Finance and Economics, 27 (November 2003), pp. 279-301. Smith, Gary, and Margaret H. Smith. “Bubble, Bubble, Where’s the Housing Bubble?” Preliminary draft prepared for the Brookings Panel on Economic Activity, March 30-31, 2006. Vandell, Kerry D., and Thomas Thibodeau. “Estimation of Mortgage Defaults Using Disaggregate Loan History Data,” AREUEA Journal, 13 (1985). Yellen, Janet L. “Housing Bubbles and Monetary Policy,” presentation to the Fourth Annual Haas Gala, San Francisco, October 21, 2005. Kau, James B., and Donald C. Keenan. “Patterns of Rational Default,” Regional Science and Urban Economics, 29 (November 1999), p. 765. www.philadelphiafed.org Fiscal Imbalance: Problems, Solutions, and Implications A Summary of the 2005 Philadelphia Fed Policy Forum “F by Loretta J. Mester iscal Imbalance: Problems, Solutions, and Implications” was the topic of our fifth annual Philadelphia Fed Policy Forum held on December 2, 2005. This event, sponsored by the Bank’s Research Department, brought together economic scholars, policymakers, and market economists to discuss and debate the implications of fiscal imbalance for the U.S. economy. Our hope is that the 2005 Policy Forum will serve as a catalyst for both greater understanding and further research on the fiscal challenges facing the U.S. economy. At the current pace of spending and revenue generation, the U.S. faces a worsening budget position over the coming years. While the problems with the Social Security program have garnered most of the headlines, financing health care and the Medicare system poses the greatest challenge. The size of the problem, longer-term implications of fiscal imbalance, and potential solutions were the focus of the 2005 Philadelphia Fed Policy Forum. While there is general agreement Loretta J. Mester is a senior vice president and director of research at the Federal Reserve Bank of Philadelphia. This article is available free of charge at www.philadelphiafed.org/econ/br/index.html. www.philadelphiafed.org that budget imbalance is one of the important challenges facing the U.S. economy over the medium and longer run, there is considerably less agreement on what should be done to meet those challenges. Alan Greenspan, then Chairman of the Federal Reserve Board, opened the conference. In his view, the deficit-reducing actions necessary to stem the worsening budget position will be difficult to implement unless procedural restraints on the budget-making process, like limits on discretionary spending and the PAYGO requirements, are restored. He said that reinstating the structure in the Budget Enforcement Act of 1990 and coupling it with provisions for dealing with unexpected budget outcomes would be beneficial. But it would not be enough to solve the problem. The fundamental issue is making choices among budget priorities, especially since the number of retirees is increasing. Greenspan pointed out that currently 3.25 workers contribute to the Social Security system for each beneficiary. By 2030, the number of beneficiaries will have doubled and the ratio of covered workers to beneficiaries will have fallen to 2. At the same time, spending per Medicare beneficiary is expected to increase as the cost of medical care rises. In fiscal year 2005, federal outlays for Social Security, Medicare, and Medicaid totaled about 8 percent of gross domestic product (GDP). Office of Management and Budget projections suggest this share will rise to 9.5 percent by 2015 and to 13 percent by 2030. While productivity growth can help alleviate some of the strain on the budget, it won’t be the whole answer. Growing budget deficits could drain resources from private investment and thereby hurt the growth of living standards. As Greenspan noted, some of the parameters needed to scale the problem are known. For example the size of the adult population in 2030 is fairly easy to estimate since most of that population has already been born. But other parameters, such as the amount of future medical spending, are very difficult to estimate. Medical technological innovations can improve the quality of health care and lower the cost of existing treatments, but they can also expand treatment possibilities and life expectancy, and both of these can mean higher spending. Greenspan said that he fears the U.S. may have already committed more resources to the baby boomers than it can deliver. If so, making changes to those promises should come Business Review Q3 2006 31 sooner rather than later – a theme echoed by other speakers at the Policy Forum – so people can plan their work, savings, and retirement spending accordingly. Although he believes closing the budget gap depends on changes to both the spending and tax sides, he thinks that most of the change should come on the outlay side, and he suspects that we may need to make significant structural changes to U.S. retirement and health programs. Solving the Medicare problem is more difficult than Social Security because of the difficulties in estimating the trend in medical expenditures. Greenspan concluded by saying that doing nothing to solve the budget imbalance could have severe consequences for the U.S. economy, but addressing the issue in a timely and sound fashion could produce lasting benefits. SOCIAL SECURITY AND MEDICARE: SCALING THE PROBLEM AND PROPOSED SOLUTIONS* This session took up the problem of how to scale the fiscal deficit problem and what can be done to solve it. Our first speaker, Robert Shiller, of Yale University, spoke on the underlying life-cycle issues involved in the Social Security and Medicare deficits. He expanded on many of Chairman Greenspan’s themes but from the perspective of behavioral economics. One of the fundamentals underlying the government budget deficit and the low personal saving rate is the problem people and society have in planning for the distant future. In Shiller’s view, these behavioral considerations justify government interventions in a broader set of circumstances than those suggested by the traditional economic * Many of the presentations reviewed here are available on our website at www.philadelphiafed.org/econ/conf/ forum2005/ program.html. 32 Q3 2006 Business Review theory of public goods or externalities. Shiller listed several of the concepts from behavioral economics that are important to understanding how we think about the future. One of these is hyperbolic discounting, which refers to the tendency to behave inconsistently over time: We tend to be impulsive and put more value on today than tomorrow. Psychologists are documenting that people think about the present in concrete terms but the future in more abstract terms, and this may underlie why people place more importance on the present than the future. Another concept is that of framing. People may behave inconsistently, depending on how a situation is described to them; they react to the names things are given and the context. Psychological research has also shown that some of the biggest errors people make are errors of attention: Something else has caught their attention, and they don’t get around to thinking about saving. In addition, what people tend to think about is what other people think about (a kind of herding). There is also a wishful thinking bias: People believe what they want to believe. This makes people tend to underestimate risk. Indeed, psychologists have hypothesized that people have certain pathways in their brains to deal with risk, but they are not suited to the modern world. For example, being in a crowded room when a wild animal escapes would cause your scared reflexes to engage, but being told you are not saving enough for the future doesn’t. The last behavioral concept Shiller discussed was the instinct for people to believe those in authority: People have high expectations for government authorities and tend to believe them. In Shiller’s view, we need to incorporate these concepts of behavioral economics into our thinking about ways to solve the government budget problems, and he believes this is beginning to happen. This approach need Robert Shiller, Yale University (left), and Peter Diamond, Massachusetts Institute of Technology. www.philadelphiafed.org not imply a “big government” solution. People are in a better position to know what they need and should be allowed to express it, but they will make mistakes that have to be dealt with. He concluded his talk by discussing some of the potential solutions to the Social Security and health-care problems. Shiller has been critical of the Bush Social Security plan, although he acknowledges it had some creative elements. In particular, Shiller thinks the life-cycle part of the plan was unique in that it would automatically put people who chose the account plan into a life-cycle portfolio at the age of 47. Having this as a default option is in accord with some of the recent principles of behavioral economics – i.e., you cannot expect people to make active choices. On the other hand, he also criticizes the Bush plan, likening it to a “margin loan” whereby people could borrow against their Social Security benefits and put the money into stock portfolios. For people already saving with a diversified portfolio, it is not of much use. For people who are not saving, the plan is risky, since the stock market is volatile. Shiller said that the Medicare Part D prescription and health savings accounts have had some implementation problems that can be viewed through a behavioral economics lens. The prescription plans afford people so many options that it is a daunting task to make an optimal choice. In Shiller’s view, there is a creative idea behind the health savings accounts, namely, insure people for catastrophic events but have them manage a budget to cover their other health spending. Unfortunately, not many people have signed up for these plans, suggesting they don’t know how to prepare for health risks. (Later Forum speakers pointed out that consumers may not have the information they need on prices and quality to make optimal www.philadelphiafed.org choices regarding health care.) Shiller thinks this should give government and private initiatives motivation to help people deal with these complicated issues, and he pointed to a few examples of private initiatives. The “save more tomorrow” plan of Richard Thaler and Shlomo Benartzi offers employees the choice to funnel future pay raises automatically into a savings account. People have a tendency not to save for today, since While poverty among the elderly has fallen, it is still at fairly high levels, especially among divorced women. it means taking away some of today’s spending. But they are willing to sign up to save more tomorrow, and the plan has been shown to increase savings. Firms are also beginning to change their 401K plans so that the default is that the employee is in the plan rather than out of it – a simple change that takes into account human behavior. Shiller views this as a time of experimentation in which our way of thinking about basic economic problems is changing. Our solutions to the savings and health-spending problem can be more creative, since they will be based on new discoveries about the ways people make decisions. Peter Diamond, of the Massachusetts Institute of Technology, continued the discussion with a summary of the size of the Social Security and Medicare deficit problems and a critique of proposed solutions. According to the 2005 Annual Report of the Trustees of Social Security, the Social Security trust fund will be exhausted in 2041. At that point, benefits would be cut by about 25 percent to match revenues. Using a 75-year horizon, the unfunded portion of promised benefits is 1.8 percent of taxable payrolls as of January 1, 2005. Comparing this to the current Social Security payroll tax of 12.4 percent shows that there is a problem, but it is not an enormous one when compared with the Medicare problem. Certain groups rely more on Social Security for a larger part of their retirement income than other groups. For example, a fifth of the elderly get all of their income from Social Security, and two-thirds get 50 percent or more. Particularly vulnerable groups include long-career low earners, widows and widowers with low benefits, disabled workers, and surviving children. While poverty among the elderly has fallen, it is still at fairly high levels, especially among divorced women. In Diamond’s view, Social Security is a part of addressing the country’s poverty issue. Diamond was critical of the Bush Social Security plan and some of the others on the table. Diamond agreed with Shiller that to the extent that people’s private retirement plans are moving from defined benefit to defined contribution, with some investment in the stock market, he doesn’t see the individual accounts in the Bush plan as being that valuable. Moreover, the private accounts would exhaust the trust fund about a decade earlier than under the current program. Some of the other plans actually go in the wrong direction and make the 75-year Social Security trust fund shortfall larger rather than smaller. Provisions such as price indexing, which in Diamond’s view is a misnomer for real-wage deflating, and raising the age at which full benefits start result in large reductions in benefits. Regarding Medicare, the issue in Diamond’s view is how to combine universal coverage Business Review Q3 2006 33 Kent Smetters, Wharton School, University of Pennsylvania with quality and cost containment. Diamond said that without universal coverage, cost containment would have some unintended consequences. The final speaker in the first session, Kent Smetters, of the Wharton School, University of Pennsylvania, addressed some of the measurement issues in budget accounting, arguing that the traditional budget approach worked fine when government programs were more of a bricks and mortar type; it works less well for programs with long-term liabilities, such as Medicare and Social Security. The federal budget substantially underestimates the government’s liabilities by ignoring long-term liabilities. It tracks them separately, off budget. So the budget gives an incomplete picture of the country’s fiscal imbalance. The traditional budget accounting also makes it hard to evaluate the impact of program reforms. If the benefit of the reform is off budget but the cost is on budget, the reform will look like it increases the fiscal imbalance. Smetters proposes a new budgetary framework that includes two integrated components: a fiscal imbalance component 34 Q3 2006 Business Review that equals debt held by the public plus the present value of all future outlays minus the present value of all future revenues, and a generational imbalance component that measures the proportion of the fiscal imbalance due to spending by past and current generations relative to what they have paid into the system. Different reform proposals for Social Security will have different effects on the generational imbalance, depending on how they affect taxes and benefits now and in the future. Under the assumptions made by the Office of Management and Budget, the Department of the Treasury, and the Council of Economic Advisers, Smetters estimates that the total fiscal imbalance in Social Security in 2004 was $8 trillion. Past and living generations have gotten about $9.5 trillion more from Social Security than they paid into it, and under current law, future generations will pay $1.5 trillion more into the program than they will get out of it, for a net total imbalance of $8 trillion. Medicare has a much larger imbalance of $61 trillion, with $24 trillion due to past and living generations and $37 trillion due to future generations. The rest of the federal government is in a surplus. Thus, the total fiscal imbalance was $63 trillion in 2004, and it is growing significantly each year. This represents 18 percent of all future payrolls and is a very large problem. For example, Social Security and Medicare benefits would have to be cut by over half to close the imbalance. Alternatively, the combined employer-employee payroll tax would have to rise from 15.3 percent to over 32 percent and the payroll tax ceiling would have to be removed. Given these dire numbers, why haven’t the capital markets reacted? Smetters says it could be behavioral, along the lines Shiller discussed; that is, they don’t understand the magni- tudes. But it also could be that capital markets believe the government is going to solve the problem mainly by cutting benefits rather than raising taxes. Smetters thinks this is a somewhat irrational view, given the aging of the median voter. He ended on an optimistic note by pointing out that 50 percent of U.S. households don’t hold any equities either directly or indirectly in employer-sponsored defined contribution plans. Thus, the component of the Bush plan that puts people into a life-cycle portfolio plan automatically by default is an important innovation in Smetters’ view. He is less concerned than Shiller and Diamond that people will make wrong choices. The Policy Forum’s keynote luncheon speaker was Katherine Baicker, a member of the Council of Economic Advisers, who spoke about the important fiscal challenges the U.S. faces over the coming years on both the spending and revenue sides of the federal balance sheet and her views of what steps should be taken to meet those challenges. Baicker pointed out that while over the last 40 years spending and revenues have been relatively stable, there have been important changes in the composition of both that will help determine future stability if nothing is done to entitlement programs, the largest of which are Social Security, Medicare, and Medicaid. Without changes in those entitlement programs, Baicker says that a decade from now, government spending as a share of GDP will begin to rise swiftly, with potentially dire consequences for the U.S. economy. On the expenditure side, federal spending as a fraction of GDP since 1962 has been relatively stable at about 20.4 percent, but the share of GDP devoted to entitlement spending has tripled, while the share of spending going to defense and other government spending, such as highways, educa- www.philadelphiafed.org tion, and national parks, has fallen. In 1962, entitlement spending was primarily Social Security, and it was 2.5 percent of GDP and 13 percent of the federal budget. Medicare and Medicaid were introduced in the 1960s, and in 2005, the three programs together accounted for 8 percent of GDP and made up 40 percent of the federal budget (not including the substantial contributions to Medicaid made by the states). The revenue side of the federal budget also shows stability, with total federal revenues averaging 18.2 percent of GDP since 1962. Payroll taxes, which are used to fund Social Security and Medicare, have doubled over the period, from about 3 percent of GDP to a bit over 6 percent. Personal income tax collections have been relatively stable, while excise tax and corporate income tax collections have declined. Comparing the revenue and expenditure sides shows that the federal government has been running a deficit of about 2.2 percent of GDP a year. In 2005, the deficit was somewhat higher at 2.6 percent of GDP. But Baicker pointed out that the stability of the fiscal situation in the U.S. over the past 40 years is in jeopardy, since the first part of the baby boomers will reach retirement age in 2008. Over the next 40 years, the costs of the three entitlement programs will rise from about 8 percent of GDP today to over 15 percent of GDP in 2045. This trend suggests that without a change in the programs, either taxes must increase substantially or spending outside of entitlements must be nearly eliminated – both poor choices in Baicker’s view. Baicker agreed with the earlier speakers that solving the Medicare/Medicaid problem was more challenging than solving Social Security, because she was optimistic that the President’s plan of progressive indexing of benefits of higher-earning www.philadelphiafed.org workers to prices would be an important step toward permanent solvency. To control the cost of government-financed health care, Baicker said we need to address the costs of health care in the private sector as well. In her view, much of the spending on health care – both publicly and privately financed – is not being efficiently allocated. To alleviate this, she said it is most important to create incentives for high-value care. For example, Baicker Katherine Baicker, Council of Economic Advisers said that the current tax code subsidizes employerincrease competitiveness in these provided health insurance relative to markets, leading to lower prices and other forms of compensation and to improved quality. At the same time, individually purchased health insurBaicker acknowledged in the question ance. This leads to insurance coverage and answer period that several difficulof routine and predictable health-care ties would need to be solved before expenditures rather than paying for moving to what she calls “consumerthose out-of-pocket and insuring The stability of the fiscal situation in the U.S. over the past 40 years is in jeopardy, since the first part of the baby boomers will reach retirement age in 2008. against catastrophic and unexpected expenditures. Baicker says capping the employer exclusion of health insurance premiums is one step that could be taken to increase the sensitivity of the use of health care to its cost. She is also in favor of expanding health savings accounts, which allow people to pay for health care with pre-tax dollars as long as their health insurance policy includes a sufficiently high deductible and catastrophic coverage. She believes steps like these would help ensure that health-care resources were allocated to uses with higher value, and she thinks this could also driven health care.” One of these is a lack of transparency. For example, it is difficult to decipher the pricing of services from the bills you receive from health-care providers and to obtain information on the quality of providers. Without price and quality information, rational health-care decisions are severely hampered (even aside from the behavioral aspects of decision-making Shiller spoke about). Our second session turned to two budget experts for their views on the current budget deficit and prognosis. Doug Holtz-Eakin, then director of the Congressional Budget Office, said Business Review Q3 2006 35 he viewed the U.S. fiscal situation to be the single most important economic policy challenge we face if the current programs are not reformed. In his view, adhering to the promises to spend as under current law will fundamentally impair the economic success of the U.S. It will result in a larger federal government, higher tax rates, and more reliance on mandates and regulations to achieve policy aims rather than on the budgetary process. In the CBO’s summer update to the budget outlook, the federal budget was projected to move back to baseline trends and become better in balance over the next five years. But there were several risks to that projection, for example, the path of defense spending and possible changes to the alternative minimum tax. Moreover, the hurricanes, which occurred after that update, affected the budget in three ways. They changed the cost of ongoing programs, but not by large amounts. They led to direct appropriations for relief and recovery, but the spending associated with those generally takes place over time. They might also lead to permanent changes in the law; for example, 12 pieces of legislation with hurricane relief provisions passed quickly. Holtz-Eakin explained that the spending and tax reconciliation is now an important part of the budget process. For the first time in eight years, Congress has used these procedures to cut spending in mandatory programs that are relevant to the long-term budget outlook. In HoltzEakin’s view the important thing is not the amounts but the fact that Congress now understands that each year the mandatory programs need to be on the table and that the process of reconciliation will be part and parcel of the process of legislating. While he suggested there are some things the government could do to improve the formulation of the budget 36 Q3 2006 Business Review – such as incorporating an average level of funds in anticipation of natural disasters that recur repeatedly like hurricanes, wild fires, and droughts – Holtz-Eakin said this is not the key Doug Holtz-Eakin said he viewed the U.S. fiscal situation to be the single most important economic policy challenge we face if the current programs are not reformed. to solving our budget problems. Rather, the key is addressing the long-term cost of our mandatory spending problems. It is important that the relatively benign near-term budget outlook not seduce us into ignoring the long-term problems. In Holtz-Eakin’s view, policy decisions rather than the course of the economy are central to the long-term budget outlook. Alice Rivlin, of the Brookings Institution, continued the discussion by pointing out that it has been decades since the U.S. has seen as rapid a reduction in revenues as a percent of GDP as has occurred in the past five years. The CBO projections are based on current law. Thus, they assume the tax cuts of 2000, 2001, and 2003 expire. If instead they continue, the budget imbalance is much worse. Rivlin said we experienced a similar situation in the 1990s. Back then there was bipartisan agreement that something needed to be done about it. There wasn’t bipartisan agreement about what should be done, but rules were put in place to control spending, control entitlement spending, and control tax cuts, and the strong economy operated to reduce the deficit and turn it into a surplus. Rivlin said this time we do not have consensus that there is a problem, even though the future budget imbalance is very large as spending on Social Security, and especially on Medicare and Medicaid, increases rapidly. Echoing the earlier Policy Forum speakers, Rivlin believes the Social Security problem is manageable; Medicare and Medicaid are far larger problems. She thinks per capita health spending, both nationally and in these programs, will continue to rise 2.5 percent faster than GDP as it has over the last four decades; she is skeptical of the Medicare trustees’ assumption that it will decelerate to 1 percent faster than GDP growth and eventually to the same pace as GDP growth. Under Rivlin’s assumption, there is no tax rate that will bring back fiscal balance, and if the deficit problem isn’t Alice Rivlin, Brookings Institution www.philadelphiafed.org solved, interest expenditures will rise to over 20 percent of GDP by 2050, so borrowing is not a sustainable option either. According to Rivlin, to solve the budget imbalance problem, we need to slow the rise in health-care spending in the federal budget. And we need to do that in a way that will slow the per capita spending on health care not only by the government but also by the private sector, because otherwise it is just shifting the expenditures. Rivlin pointed out that the U.S. has a very expensive health-care system compared with other countries; while the rates of growth in per capita spending are similar in developing countries, our level of spending is higher. While there have been cost-saving innovations in providing health care, these innovations also tend to increase demand for the service. But there are ways to increase cost effectiveness. For example, the practice of medicine continues to be paper-based; improvements in information systems could probably reduce costs and might also result in fewer treatment errors. Because the Medicare system is an almost universal system for the over-65 population, it holds the potential for learning about which treatments are cost-effective – provided its data can be analyzed. The next issue would be what to do with this information. Rivlin pointed out that one strategy was suggested by Baicker: to give consumers the information and let them make the choices through, for example, health savings accounts. The other strategy is to change the reimbursement system to reward effective medical care and not pay for ineffective and excessive medical care – although Rivlin admitted we don’t know how to do that yet. She suggested a companion step is to use federal government research dollars to push for innovations likely to be cost saving, www.philadelphiafed.org especially for diseases like cancer where the innovations are unlikely to lead to expanded treatment, since these diseases already always get treated at some stage. There are political obstacles that would need to be overcome. In Rivlin’s view, these include the power of insurers, pharmaceutical companies, and providers, who have been fairly negative on change. Rivlin concluded Richard Fisher, President, Federal Reserve Bank of Dallas by pointing out that the U.S. is not alone in this problem, which she says is a problem tion. He believes it is very important of prosperity. In the U.S. and in other to consider how the forces of globalizasuccessful economies people are living tion affect U.S. fiscal deficits. Glolonger and better, and part of that livbalization means a nation’s economic ing better is better medical care. potential is no longer defined by its geographic boundaries. In a global economy, goods, services, capital, and labor can migrate to where they can be most efficiently used and where there are fewest obstacles to putting them to efficient use. So countries need to compete for these resources. In Fisher’s view, businesses have come to grips with globalization, and globalization has helped discipline central bankers around the world to focus on keeping inflation low. Fisher believes that globalization is also exerting some discipline on fiscal policymakers, and the U.S. is in better shape than most of its competitors. One of the ways The Policy Forum’s final session globalization has a beneficial effect took up the broader implications of on fiscal decision-making is via tax fiscal imbalance for the macroeconocompetition. Fisher pointed out that my. Richard Fisher, president of the average tax rates are falling in the Federal Reserve Bank of Dallas, agreed world’s most open economies. Also, to that the magnitude of the projected the extent that young people can move budget deficits is of great concern to escape high Social Security taxes, and said that, left unchecked, they it is more difficult to sustain a system have the potential of harming U.S. based on intergenerational transfers. economic prosperity and undermining In theory, globalization should exthe progress we have made on inflaert a similar discipline on the spending In a global economy, goods, services, capital, and labor can migrate to where they can be most efficiently used and where there are fewest obstacles to putting them to efficient use. Business Review Q3 2006 37 side, but Fisher says we have yet to see such deficit-reduction pressures. Nonetheless, when investors are considering where to allocate their capital, it is the relative position of one country vs. another that matters. In Fisher’s view the U.S. has been able to finance its spending via foreign capital because we are doing better in terms of fiscal policy compared with other countries. Fisher provided some numbers: According to OECD data, public-sector spending (including federal, state, and local government spending) was projected to be 3.7 percent of GDP in the U.S. in 2005, compared with 6.5 percent in Japan, 4.3 percent in Italy, and 3.9 percent in Germany. He thinks that the demographic challenges regarding Social Security and Medicare in the U.S. are not as severe as those facing Japan and Germany. But while the U.S. may be better off than other countries, Fisher believes following “least-bad” policy is risky, since it is never clear whether our advantages will last, especially if a rising deficit erodes U.S. economic performance. He believes that to secure our advantages we should put our fiscal house in order before our competitors put theirs in order. Fisher pointed out that monetary policymakers cannot be indifferent to the thrust of fiscal policy because poor fiscal policies create pressure for poor monetary policies, e.g., monetizing the debt and fueling inflation. But he emphasized that the solution to the U.S. fiscal imbalance rests with fiscal policymakers and not the central bank. Robert Barro, of Harvard University, took up the theme of monetary policy touched on by 38 Q3 2006 Business Review Fisher. In Barro’s view, in the last 25 years there has been a major triumph in terms of central banks around the world achieving low and stable rates of inflation. He said he is not certain why monetary policy has worked as well as it has in the U.S. and abroad. His analysis indicates that Fed policy under former Chairman Greenspan could be characterized as a reaction function, with the federal funds rate reacting to the inflation rate and the real economy as embodied in employment growth and the unemployment rate. The analysis suggests that the Fed does not respond to changes in real GDP that are due to productivity growth. The Fed’s policy is also characterized by gradualism: It moves interest rates gradually. Barro said it was not clear that the Fed’s reacting to the real economy and gradualism are beneficial. Nonetheless, in Barro’s view the Fed’s triumph over high inflation is a remarkable achievement. Alan Auerbach, of the University of California at Berkeley, formulated his talk around the policy changes we should expect in response to the fiscal situation we face and the economic effects we should expect as people anticipate these policies. He agreed with earlier speakers that we face a rising imbalance that gets much larger with every year something isn’t done to solve it. Auerbach said the problems policymakers need to address (in order of importance) are health-care spending and the federal contributions to Medicare and Medicaid; general revenue taxes, i.e., taxes not associated with entitlement programs and not payroll taxes; and Social Security. In Auerbach’s view, most discussions of the macroeconomic effects of fiscal imbalance have focused mainly on the effect of current fiscal policy on the economy. These would include possible crowding out of private investment by government spending, higher interest rates, and current account deficits. There has been little discussion of the effects (left to right): Tony Santomero, former President, Federal Reserve Bank of Philadelphia; Robert Barro, Harvard University; Alan Auerbach, University of California at Berkeley; and Doug Holtz-Eakin, former Director, Congressional Budget Office. www.philadelphiafed.org of the necessary policy changes on the economy. Given the size of the future fiscal imbalance and the fact that federal taxes as a share of GDP are lower now than at any time since the 1960s, an eventual tax increase of 4 percent of GDP, through a combination of broadening the tax base and increasing marginal tax rates, would not be implausible in Auerbach’s view. Economic models suggest that higher future tax burdens should induce people to increase effort today to be able to pay future taxes and to save. Thus, we’d expect higher labor-force participation, higher employment, and higher private saving to pay for future taxes. The higher marginal tax rates might also encourage more work today if people plan to retire earlier than otherwise as a result of the tax change and decide to work harder now to save enough to retire. However, higher marginal tax rates would also induce lower private saving, since those savings would be taxed at a higher rate. Auerbach doesn’t think much progress on the Social Security and Medicare problems will be made until there is a crisis. At that point, the problem will be too large to be solved by increased payroll taxes alone, but politically it will be nearly impossible to make sizable benefit cuts for less affluent retirees. Thus, Auerbach believes there will be means testing of entitle- www.philadelphiafed.org ments in the future. Means testing has mixed effects on incentives to accumulate wealth. If you are so wealthy that you know you are going to be hit by the means test, you’ll have an incentive to accumulate even more wealth, since your retirement and health-care benefits have just been reduced. But if your wealth is near the level where benefits are phased out by the means markets. Until that uncertainty is resolved, the equity premium should be higher. At least, this should occur when people realize the current fiscal situation is not sustainable. A resolution of policy uncertainty would make us better off, and Auerbach suggested that the costs of adjustment that we know must come at some point would be lower if we adopted more gradual There was agreement that difficult policy choices will have to be made and that the time for making them is now, not later, if we want to reduce the impact of the fiscal imbalance on the U.S. economy. test, you could have a strong incentive to save less so that you would qualify for benefits. And since you are saving less, you’ll work less as well. Auerbach pointed out two other potential macroeconomic effects as the economy adjusts to changes in fiscal policy. Trade deficits will shrink and turn into trade surpluses in the future. As that occurs, the composition of U.S. GDP will change toward more trade-sensitive industries. Until we know how the fiscal imbalance will be handled – how much taxes increase, how much marginal tax rates increase, how much the tax base broadens, how much benefits are cut – there will be substantial uncertainty in financial systemic plans to address the fiscal imbalance. However, Auerbach said he was not encouraged by recent policy actions. SUMMARY The 2005 Policy Forum generated lively discussion among the program speakers and audience on the challenges facing the U.S. in dealing with its increasing fiscal imbalance. Although there was no agreement on particular solutions, there was agreement that difficult policy choices will have to be made and that the time for making them is now, not later, if we want to reduce the impact of the fiscal imbalance on the U.S. economy. BR Business Review Q3 2006 39 The Philadelphia Fed POLICY FORUM December 1, 2006 We will hold our sixth annual Philadelphia Fed Policy Forum on Friday, December 1, 2006. This year’s topic is “Economic Growth and Development: Perspectives for Policymakers.” At right is the program. The Policy Forum brings together a group of distinguished economists and policymakers for a rousing discussion and debate of the issues. For further information, please contact us at PHIL.Forum@phil.frb.org. Photo by B. Krist for GPTMC FEDERAL RESERVE BANK OF PHILADELPHIA The Philadelphia Fed Policy Forum Economic Growth and Development: Perspectives for Policymakers December 1, 2006 Federal Reserve Bank of Philadelphia, 6th and Arch Streets Continental Breakfast Opening Remarks Charles I. Plosser, Federal Reserve Bank of Philadelphia Economic Growth and Development: An Overview of Issues and Evidence Moderator: Michael Dotsey, Federal Reserve Bank of Philadelphia Roberto Zagha, The World Bank Xavier Sala-i-Martin, Columbia University Discussion and Audience Participation Policy Responses: Trade and Foreign Aid Moderator: Kei-Mu Yi, Federal Reserve Bank of Philadelphia Elhanan Helpman, Harvard University William Easterly, New York University Discussion and Audience Participation Lunch Financial Markets and Growth Moderator: Loretta J. Mester, Federal Reserve Bank of Philadelphia Jeffrey M. Lacker, President, Federal Reserve Bank of Richmond Robert M. Townsend, University of Chicago Discussion and Audience Participation Institutional Arrangements and Economic Growth and Development Moderator: George Alessandria, Federal Reserve Bank of Philadelphia Dani Rodrik, Kennedy School of Government, Harvard University Ross Levine, Brown University Daron Acemoglu, Massachusetts Institute of Technology Discussion and Audience Participation Reception and Informal Discussion www.philadelphiafed.org Business Review Q3 2006 41 Research Rap Abstracts of research papers produced by the economists at the Philadelphia Fed You can find more Research Rap abstracts on our website at: www.philadelphiafed.org/econ/resrap/index. html. Or view our Working Papers at: www.philadelphiafed.org/econ/wps/index.html. Nontraded Goods and the Behavior of Exchange Rates Empirical evidence suggests that movements in international relative prices (such as the real exchange rate) are large and persistent. Nontraded goods, both in the form of final consumption goods and as an input into the production of final tradable goods, are an important aspect behind international relative price movements. In this paper, the authors show that nontraded goods have important implications for exchange rate behavior, even though fluctuations in the relative price of nontraded goods account for a relatively small fraction of real exchange rate movements. In their quantitative study, nontraded goods magnify the volatility of exchange rates when compared to the model without nontraded goods. Cross-country correlations and the correlation of exchange rates with other macro variables are closer in line with the data. In addition, contrary to a large literature, standard alternative assumptions about the currency in which firms price their goods are virtually inconsequential for the properties of aggregate variables in the authors’ model, other than the terms of trade. Working Paper 06-9, “Nontraded Goods, Market Segmentation, and Exchange Rates,” 42 Q3 2006 Business Review Michael Dotsey, Federal Reserve Bank of Philadelphia, and Margarida Duarte, Federal Reserve Bank of Richmond Interpreting the Link Between Technology and Human Capital The positive correlations found between computer use and human capital are often interpreted as evidence that the adoption of computers has raised the relative demand for skilled labor, the widely touted hypothesis of skill-biased technological change. However, several models argue that the skill intensity of technology is endogenously determined by the relative supply of skilled labor. The authors use instruments for the supply of human capital coupled with a rich data set on computer usage by businesses to show that the supply of human capital is an important determinant of the adoption of personal computers. Their results suggest that great caution must be exercised in placing economic interpretations on the correlations often found between technology and human capital. Working Paper 06-10, “Labor Supply and Personal Computer Adoption,” Mark Doms, Federal Reserve Bank of San Francisco, and Ethan Lewis, Federal Reserve Bank of Philadelphia www.philadelphiafed.org Using the National Income Accounts to Quantify Economic Activity This article presents a brief overview of the national income accounts. It summarizes the main parts of accounts and situates them within the efforts of economists to quantify economic activity and economic well-being. The author argues that these statistics are necessarily provisional and imperfect but nevertheless extremely useful. Some current directions for economic research seeking to extend the accounts are also discussed. Working Paper 06-11, “National Income Accounts,” Leonard Nakamura, Federal Reserve Bank of Philadelphia Understanding the Great Depression What caused the worldwide collapse in output from 1929 to 1933? Why was the recovery from the trough of 1933 so protracted for the U.S.? How costly was the decline in terms of welfare? Was the decline preventable? These are some of the questions that have motivated economists to study the Great Depression. In this paper, www.philadelphiafed.org the authors review some of the economic literature that attempts to answer these questions. Working Paper 06-12, “Monetary and Financial Forces in the Great Depression,” Satyajit Chatterjee, Federal Reserve Bank of Philadelphia, and Dean Corbae, University of Texas at Austin Extending the Job Matching Model In the U.S. labor market, the vacancy-unemployment ratio and employment react sluggishly to productivity shocks. The authors show that the job matching model in its standard form cannot reproduce these patterns due to excessively rapid vacancy responses. Extending the model to incorporate sunk costs for vacancy creation yields highly realistic dynamics. Creation costs induce entrant firms to smooth the adjustment of new openings following a shock, leading the stock of vacancies to react sluggishly. Working Paper 06-13, “Job Matching and Propagation,” Shigeru Fujita, Federal Reserve Bank of Philadelphia, and Garey Ramey, University of California, San Diego Business Review Q3 2006 43 ANNOUNCEMENT AND CALL FOR PAPERS The Federal Reserve Bank of Philadelphia, Rutgers University, and the University of Richmond Real-Time Data Analysis and Methods in Economics April 19-20, 2007 Philadelphia, Pennsylvania T he Research Department of the Federal Reserve Bank of Philadelphia, the Economics Department at Rutgers University, and the Robins School of Business at the University of Richmond are sponsoring a conference on Real-Time Data Analysis and Methods in Economics to be held at the Federal Reserve Bank of Philadelphia on April 19-20, 2007. The purpose of the conference is to bring together leading researchers interested in all areas of real-time data analysis, including but not limited to topics such as real-time macroeconometrics, finance, forecasting, and monetary policy. Those interested in presenting a paper at the conference are encouraged to send a completed paper or detailed abstract by November 1, 2006, to Tom Stark at tom.stark@phil.frb.org. Discussions are underway with a number of journals, including the Journal of Business and Economic Statistics, about the possibility of publishing a special conference volume (though authors would not be compelled to publish their paper in such a volume), and a variety of leading researchers in the area have expressed interest in taking part. Additionally, a summary of the conference will be published in the Philadelphia Fed’s Business Review. We will provide some travel expenses for paper presenters and discussants, following Federal Reserve guidelines. Conference details will be posted in due course on the websites of the conference organizers. Questions or comments should be directed to one of the conference organizers: Dean Croushore: Economics, University of Richmond............................dcrousho@richmond.edu Tom Stark: Federal Reserve Bank of Philadelphia....................................... tom.stark@phil.frb.org Norman R. Swanson: Economics, Rutgers University........................ nswanson@econ.rutgers.edu 44 Q3 2006 Business Review www.philadelphiafed.org