Survey

# Download Macroeconomic Expectations and the Stock Market

Document related concepts

Transcript

Macroeconomic Expectations and the Stock Market: The Importance of a Longer-Term Perspective Vanguard Investment Counseling & Research Executive summary. This paper aims to help long-term investors better understand how to evaluate economic information in forming expectations about the stock market. Given the tight long-run relationship between the stock market and the economy, investment commentary routinely suggests that the market’s short-run prospects have changed in response to observed changes in macroeconomic expectations. Of course, such logic presumes a predictive relationship: that today’s expectations for tomorrow’s economy are correlated with tomorrow’s stock returns. Is this valid? We begin by examining the link between macroeconomic expectations and the subsequent near-term performance of the U.S. stock market. Specifically, we run quarterly predictability regressions for excess stock returns on expectations for real growth and inflation derived from (1) economist surveys and (2) financial market indicators. We repeat these regressions for two stock-return spreads: (1) value minus growth and (2) small-cap minus large-cap. We show that the consensus view of future macroeconomic conditions—whether derived from professional forecasters or financial indicators—has explained virtually none of the short-term volatility in stock returns over the past 40 years. While this may be surprising to some given the considerable attention paid to such expectations, our results are consistent with the view of a stock market that is reasonably efficient (it tends to “price in” the mainstream macro outlook), that is forward-looking (as a leading indicator, it tends to anticipate economic shifts rather than lag them), and that can be quite volatile in the short run. 1 I would like to thank Glenn Sheay for excellent research assistance. Connect with Vanguard > www.vanguard.com Author Joseph H. Davis, Ph.D.1 In place of reliance on near-term forecasts, we urge macro-minded investors to adopt a more strategic perspective regarding the future outlook for the economy and the stock market. We provide a “game plan” for those who want to better integrate macroeconomics into their portfolio decision-making by focusing on the fundamental building blocks of the stock market’s long-term performance. The link between the economy and the stock market The value of a broad stock index reflects current and expected earnings and dividend growth and other systematic return and valuation factors. A simplified version of the well-known Gordon model, for instance, expresses the expected excess return of the stock market (over cash) as the dividend yield plus expected real earnings growth. All else being equal, the higher the earnings yield at purchase and the higher the expected long-term real earnings growth, the higher the expected stock return (Asness, 2003). Figure 1. The U.S. stock market, its fundamentals, and the U.S. economy over the long run Annual data for 1872–2007, in logarithmic scale 100,000 10,000 1,000 100 10 1 The building blocks of the stock market’s long-run performance—earnings and dividends per share— are ultimately co-integrated with the economy’s performance.2 Figure 1 illustrates that the trend of the broad U.S. stock market has historically tracked the productivity level of the U.S. economy, measured here as gross domestic product per capita. In the past, the trend growth rate in real corporate earnings has paralleled the economy’s real productivity growth, while dividend yields have been highly correlated with the level of inflation expectations. Of course, stock market risk premiums vary significantly in the short run. Given the tight longrun relationship between the stock market and the economy in Figure 1, it is perhaps not surprising that news stories and investment analysts focus keenly on expectations for an economy’s near-term performance to gauge the future prospects of its stock market. 0.1 1880 1900 1920 1960 1980 2000 GDP per capita ($000) S&P 500 Index earnings per share S&P 500 Index dividends per share S&P 500 Index Sources: Standard & Poor’s, U.S. Census Bureau, and Robert Shiller’s website, http://www.econ.yale.edu/~shiller. Recently in the United States, expectations for future economic growth have declined appreciably, in part as a result of the subprime-mortgage and financial crisis. At the same time, certain measures of inflation expectations have risen because of record-high commodity prices and the weak U.S. dollar. This unwelcome combination of lower growth and higher inflation expectations has led some analysts to warn of disappointingly low U.S. stock market returns in coming quarters. 2 For an excellent discussion of alternative ways to decompose long-run stock returns, see Ibbotson and Chen (2003). 2 > Vanguard Investment Counseling & Research 1940 Yet is the link between expected macro variables and future stock returns as strong as commonly suggested? In the following section, we examine this link in detail with the goal of helping long-term investors better understand the implications of macroeconomic expectations for the stock market. Do macroeconomic expectations predict near-term stock returns? Academic researchers (e.g., Chen et al., 1986; Fama and French, 1989; Schwert, 1990) have long argued that business conditions should be positively correlated with future stock returns, because an economy’s business cycle should affect corporate cash flows and discount rates.3 Of course, such logic assumes a causal relationship: that today’s expectations for tomorrow’s economy are strongly correlated with tomorrow’s stock returns. To test this logic, we run regressions to examine the correlation between macroeconomic expectations (for real growth and inflation) and the subsequent performance of the U.S. stock market over four decades. We consider two distinct sources for macro expectations: (1) professional surveys and (2) financial market indicators. We first focus on survey-based macro expectations. Consensus forecasts for real GDP growth and inflation come from the Federal Reserve Bank of Philadelphia’s Survey of Professional Forecasters (SPF). Conducted since 1968, the SPF is the oldest quarterly survey of macroeconomic forecasters in the United States. Each quarter, the survey asks a group of these professionals to provide quarter-by-quarter forecasts on a variety of macroeconomic variables.4 The median SPF forecast—hereafter the consensus forecast—has historically been highly correlated with the actual reported values for various macro variables.5 For example, the correlation coefficients between the quarter-ahead consensus forecast and actual value for both real GDP growth and the inflation rate have been close to 60% since the inception of the SPF program in 1968. Indeed, the professional forecasters, as a group, appear to incorporate relevant forward-looking information in their macroeconomic projections in that the consensus SPF forecast is more accurate than a random-walk forecast. To examine the correlation between macroeconomic expectations (for real growth and inflation) and nearterm U.S. stock returns, we estimate predictability regressions of the general form: R ts+1 – R tcash +1 = f (E t [SPFt +i ]), where next quarter’s excess return of the U.S. stock s cash market, R t +1 – R t +1 , is defined as the total return of the aggregate, market-cap-weighted stock market less the return of the 3-month U.S. Treasury bill. That is, the dependent variable in the regression is the equity risk premium, or the expected excess return of stocks over the return of a risk-free investment (i.e., cash). Table 1, on page 4, demonstrates that virtually all of the annual volatility in stock returns observed since the 1870s has come from changes in the equity risk premium. Notes about risk and performance data Investments are subject to risk. Investments in bonds are subject to interest rate, credit, and inflation risk. Prices of mid- and small-cap stocks often fluctuate more than those of large-company stocks. Foreign investing involves additional risks including currency fluctuations and political uncertainty. Past performance is no guarantee of future returns. The performance of an index is not an exact representation of any particular investment, as you cannot invest directly in an index. 3 More recently, however, certain academic studies have argued that expected excess stock returns are negatively correlated with the business cycle, because either systematic risk or risk-aversion increases during recessions. For an overview, see Andersen et al. (2005) and citations therein. 4 See Croushore (1993) for more details about the Survey of Professional Forecasters. 5 We construct the median growth-rate forecast by aggregating the individual survey responses; our empirical results are nearly identical using the mean forecast. Vanguard Investment Counseling & Research > 3 Table 1. Summary statistics of annual U.S. stock market returns: 1872–2007 Average return Standard deviation Correlation to total return S&P 500 Index total return 9.0% 17.9 100.0% Cash return (“risk-free” asset) 4.2 2.7 –9.7 Equity risk premium (excess return of stocks over cash) 4.6 18.3 98.9 Note: The sum of the average equity risk premium and the average risk-free (cash) return does not equal the total return of the S&P 500 Index because of rounding and the omission of a small geometric interaction term. Source: Author’s calculations from data sources cited in Figure 1. The variables on the right hand side of the regression, (Et [SPFt +i ]) , include the median forecast surveyed in quarter t for real GDP growth and inflation for either the next quarter (i =1) or the next four quarters (i = 4). A statistically significant coefficient on any right-handside variable in this regression would provide evidence of in-sample stock market predictability, meaning that today’s expectations for tomorrow’s economy are correlated with tomorrow’s stock returns.6 Table 2 presents the future stock return regressions on the SPF macro expectations data. The sample period begins in the first quarter of 1969 (1969Q1), since the first available SPF observation is in 1968Q4. The regression coefficients in Table 2 reveal that none of the SPF consensus series significantly predict nextquarter stock returns. The correlations between today’s expectations for future macroeconomic conditions and tomorrow’s excess stock returns are not statistically significant for either real GDP growth or inflation. The regressions in Table 2 are so weak, in fact, that the average adjusted R 2 is negative. This is true whether one focuses on the median expectation for the next quarter or for the next year.7 Forecast dispersion, a proxy for disagreement or uncertainty among professional forecasters, is uncorrelated with future stock returns, too.8 Robustness checks One explanation for the weak empirical results could be that survey expectations—which tend to be quite smooth and persistent—may be distorted by a trend specific to the sample period.9 Table 3 thus repeats the stock-return regressions using changes in SPF macroeconomic expectations. A positive and statistically significant regression coefficient on, say, real GDP growth would indicate that an upward revision today to future growth expectations would be associated with higher stock returns next quarter. 6 Inoue and Kilian (2004), Campbell and Thompson (2005), and others argue that this type of stock-return regression is more meaningful than out-of-sample methods in investigating whether a right-hand-side variable (i.e., SPF) predicts future stock returns. 7 Although the regression coefficients are insignificant, they tend to be negative and countercyclical, as suggested by Andersen et al. (2005) and others: Stock returns are generally higher when either growth or inflation expectations are lower. 8 The cross-sectional dispersion among individual forecasts is a common proxy for forecaster uncertainty about future outcomes, although variation in predictions need not imply disagreement among forecasters. For details, see Mankiw et al. (2003). 9 Stambaugh (1999), for instance, argues that the regression coefficients in stock predictability regressions will be biased when the right-hand-side variables are highly persistent and correlated with future stock returns. Campbell and Yogo (2003) document the presence of “Stambaugh bias” for slow-moving series such as the dividend yield, and the year-ahead SPF variables appear to possess similar characteristics. 4 > Vanguard Investment Counseling & Research Table 2. Survey-based macro expectations and future stock returns: 1969Q1–2008Q1 Variable Model 1 Model 2 Model 3 Model 4 1.84 0.95 1.81 0.92 3.32 1.51 3.36 1.52 Expected real GDP growth for next quarter –0.13 –0.13 –0.13 –0.38 Expected inflation rate for next quarter –0.13 –0.10 –0.10 –0.19 –0.65 –1.48 –0.68 –1.47 0.01 0.04 –0.09 –0.17 Constant term Expected real GDP growth for next year Expected inflation rate for next year 0.84 0.15 Forecast dispersion (A proxy for forecast uncertainty) Adjusted R2 (in %) Standard deviation of dependent variable Standard error of model equation Model’s F-statistic p-value –1.2% 8.66 8.71 0.95 1.35 0.23 –1.9% 8.66 8.74 0.99 –0.4% 8.66 8.68 0.49 –1.0% 8.66 8.70 0.69 Notes: SPF data reflect the median survey response. The regression’s dependent variable is the quarter-ahead excess return of U.S. stocks over the 3-month T-bill return. Listed beneath the regression coefficients are the Newey-West-corrected t-statistics (in italics). No coefficients in the table are statistically different from zero at 10% level. Source: Author’s calculations. Table 3. Changes in survey-based macro expectations and future stock returns: 1969Q1–2008Q1 Variable Constant term Change in expected real GDP growth for next quarter Change in expected inflation rate for next quarter Model 1 Model 2 Model 3 Model 4 1.39* 1.93 1.39* 1.92 1.38* 1.94 1.38* 1.93 –0.52 –0.70 –0.52 –0.68 1.37 0.89 1.37 0.87 Change in expected real GDP growth for next year Change in expected inflation rate for next year –1.09 –0.69 0.32 0.14 0.26 0.11 0.30 0.05 Change in forecast dispersion (A proxy for forecast uncertainty) Adjusted R2 (in %) Standard deviation of dependent variable Standard error of model equation Model’s F-statistic p-value –1.11 –0.69 0.3% 8.68 8.67 0.29 –0.3% 8.68 8.70 0.48 1.86 0.33 –0.5% 8.68 8.70 0.54 –1.1% 8.68 8.73 0.71 Notes: SPF data reflect changes in median survey response versus previous survey. The regression’s dependent variable is the quarter-ahead excess return of U.S. stocks over the 3-month T-bill return. Listed beneath the regression coefficients are the Newey-West-corrected t-statistics (in italics). Bolded coefficients with an asterisk are statistically different from zero at 10% level. Source: Author’s calculations. Vanguard Investment Counseling & Research > 5 The pattern in Table 3 parallels that for Table 2: None of the SPF consensus series significantly predict next-quarter stock returns. Quarter-to-quarter revisions to the median SPF forecasts for current growth and inflation expectations are not significantly correlated with next-quarter excess stock returns. This is true whether the revised forecasts are for the next quarter or the next year. As in Table 2, the average regression-adjusted R 2 is slightly negative: Changes in consensus macroeconomic expectations explain none of the volatility in future stock returns. We checked the results in Tables 2 and 3 against other expectation variables derived from the Survey of Professional Forecasters, including (but not limited to): (1) corporate profits, (2) industrial production, (3) nominal GDP growth, and (4) housing starts. The regressions were similarly weak for all four of these variables at various SPF forecast horizons. The “expectations spreads” between the next-quarter and next-year median SPF forecasts were insignificant, too. In addition, we verified the robustness of our results using a monthly sample of macroeconomic expectations provided by the Blue Chip Survey. Although the Blue Chip sample does not begin until 1977, the monthly predictability regressions parallel our results displayed in Tables 2 and 3.10 Finally, we ran regressions of annual stock market returns on the previous year’s median SPF expectation. The results were similarly weak, and the adjusted R 2 was again slightly negative. Macro expectations: Economist surveys versus financial market proxies Another potential explanation for the zero correlation documented in Tables 2 and 3 could be that economist surveys are either “stale” or biased in that they do not incorporate all information available in the financial market.11 If that were true, then we would expect the stock market to lead (1) changes in the median SPF forecast (particularly for real GDP growth) and (2) “surprises” to the median SPF forecast, meaning outcomes that the forecast did not predict. These two patterns should be related, since announcements of surprising economic data often lead to revisions in economist forecasts. Figure 2 reveals the correlation between “surprises” to the median SPF forecasts for two values—real GDP growth and inflation— and past, present, and future excess stock returns. Here, we define “surprises” as the difference between what was actually reported for those values and the consensus expectation for them in the prior quarter’s survey. Perhaps not surprisingly, the cross-correlation patterns demonstrate that the U.S. stock market anticipates “shocks” to real GDP growth at least two quarters ahead of the SPF survey. Figure 2 shows that excess stock returns at a given point in time are positively correlated (at a statistically significant 30%) with real GDP growth surprises announced three quarters later. The stock market does a somewhat poorer job of anticipating inflation surprises, although stock returns are likelier to lead SPF inflation surprises than to lag them. Indeed, Granger causality tests reveal that excess stock returns predict changes in survey-based macroeconomic expectations, not vice versa.12 10 We should note that our SPF survey results contradict Campbell and Diebold (2005), who find that consensus real GDP growth expectations from the biannual Livingston Survey significantly predict future stock returns (controlling for other variables). For a number of reasons, it is likely that the Campbell and Diebold results suffer from Stambaugh bias and thus are overstated. For one, the time series for real GDP growth expectations derived from the biannual Livingston Survey is highly persistent; tests fail to reject the presence of a unit root. Second, we can show that the Campbell and Diebold (2005) results do not hold when their real GDP expectation term is expressed in changes, as we do in Table 3. Third, Campbell and Diebold use the original Lettau and Ludvigson (2001) consumption-to-wealth ratio, rather than a version that is calculated in real time and appropriately lagged to avoid look-ahead bias. Finally, the positive Campbell and Diebold results are not observed using either the quarterly SPF or the monthly Blue Chip expectations databases. 11 While Croushore (2006) finds the median SPF inflation forecasts to be unbiased, Rudebusch and Williams (2007) show that the average SPF forecaster places too little weight on financial-market information—especially the Treasury yield curve—when projecting real GDP growth. 12 We cannot reject the null hypothesis that changes in the median SPF forecast do not Granger-cause excess stock returns, but we can reject at the 1% significance level the null hypothesis that excess stock returns do not Granger-cause changes in the median SPF forecast. 6 > Vanguard Investment Counseling & Research An alternative measure of macro expectations: Financial variables For these and other reasons, many researchers use financial indicators as a proxy for expected macroeconomic conditions. One could certainly argue that the financial markets’ expectations for future U.S. economic performance are more “representative” than any one survey, as the number of participants in the global financial markets dwarfs the 50 or so forecasters canvassed by the SPF in any quarter. Since they are available in real time, financial indicatorbased expectations should be at least as up-to-date as the quarterly SPF surveys. Figure 2. The stock market tends to anticipate “surprises” to survey-based macro expectations Correlation of stock returns with outcomes not predicted by the median SPF forecast 30% 20 Correlation 10 0 –10 –20 Stock returns lead surprises Stock returns lag surprises –30 –6 –5 –4 –3 –2 –1 0 1 2 3 4 5 6 Quarters before (–) / after (+) economic announcement Correlation of stock returns with real GDP growth surprises Correlation of stock returns with inflation surprises Notes: Quarterly sample for 1969Q1–2008Q1. “Surprises” are defined as the difference between the actual, as-reported values for real GDP growth and inflation and the consensus expectation for those two values from the SPF survey conducted in the prior quarter. A positive (negative) surprise would mean that real GDP growth or inflation data were higher (lower) than expected. Correlations are asymptotically consistent approximations. Source: Author’s calculations. Academic studies contend that a select number of financial market indicators significantly predict future stock returns precisely because they represent realtime expectations for (and uncertainty about) future macroeconomic conditions.13 According to Campbell and Diebold (2005) and Goyal and Welch (2006), the four most common and significant indicators include: • The shape of the Treasury yield curve (TERMt ), defined as the difference between the 10-year Treasury yield and the 3-month Treasury bill yield. • The default spread (DEFt ), defined as the difference between yields of high-grade AAA-rated and riskier BBB-rated corporate bonds. • The dividend yield of the S&P 500 Index (DIVt ). • The consumption-to-wealth ratio (CAYt -1 ), calculated in real time and appropriately lagged.14,15 13 See, among many others, Campbell and Shiller (1988), Fama and French (1989), and Goyal and Welch (2006). 14 Lettau and Ludvigson (2001) show that the consumption-wealth ratio (CAYt )—the error term from the cointegration relation among consumption, net worth, and labor income—forecasts stock market returns out of sample because it summarizes expected returns on aggregate wealth. However, Avramov (2002) shows that when the consumption-wealth ratio is calculated in real time to avoid look-ahead bias, the resultant variable (CAYt-1) does not significantly predict future stock returns. 15 When we run a contemporaneous regression of the SPF data (both one and four quarters ahead) on the values for the four financial indicators above, we find that the survey-based and financial indicator-based measures of U.S. macroeconomic expectations are indeed positively correlated. Specifically, the regression reveals that the SPF inflation expectations series are highly and positively correlated with the dividend yield and negatively correlated with the term spread, producing an adjusted R 2 exceeding 80%. This nearly one-to-one conditional relationship between dividend yields and inflation expectations would seem to accord with the dividend discount model, which predicts that the dividend yield should rise when necessary to offset higher expected inflation (Asness, 2003). The financial indicators are also positively correlated with SPF expectations for real GDP growth, although the relationship is weaker, with an adjusted R 2 of approximately 15%. Consistent with the above discussion, the weaker correlation can be explained by the fact that the financial indicators—especially the shape of the Treasury yield curve—are updated before the SPF real GDP growth forecast. Indeed, the R 2 on the real GDP regression more than doubles to over 30% when the financial indicators are regressed on the median SPF response for the following quarter, rather than on the contemporaneous forecast. Vanguard Investment Counseling & Research > 7 Table 4. Financial indicator-based macro expectations and future stock returns: 1969Q1–2008Q1 Financial proxies for macro expectations/risk Dependent variable and time period Constant DIVt DEFt TERMt CAYt-1 Equation Adjusted R2 Excess stock returns –4.03 –1.35 0.88 1.12 0.91 0.36 1.15* 2.40 –0.60 –0.79 1.6% 1st subsample: 1969Q1–1988Q4 –13.67* –2.72 2.21* 2.27 2.61 0.90 1.63* 2.77 –0.07 –0.05 9.6% 2nd subsample: 1989Q1–2008Q1 –3.76 –0.63 3.53 1.67 –3.99 –0.83 0.39 0.52 –1.44 –1.28 0.7% Full sample: 1969Q1–2008Q1 Notes: The regression’s dependent variable is the quarter-ahead excess return of U.S. stocks over the 3-month T-bill return. Bolded coefficients with asterisks are statistically significant at the 10% level. Listed beneath the regression coefficients are the Newey-West-corrected t-statistics (in italics). Source: Author’s calculations. Table 5. Financial indicator-based macro expectations and future U.S. equity style returns: 1969Q1–2008Q1 Financial proxies for macro expectations/risk Dependent variable and time period Constant DIVt DEFt TERMt CAYt-1 Equation Adjusted R2 Value stocks minus growth stocks Full sample: 1969Q1–2008Q1 1.85 0.80 –0.62 –1.14 1.60 1.18 –0.39 –0.77 0.76 0.94 –1.0% 1st subsample: 1969Q1–1988Q4 5.06 1.48 –1.01 –1.47 1.08 0.67 –0.56 –1.21 0.21 0.28 –2.6% 2nd subsample: 1989Q1–2008Q1 7.94 1.44 –3.32* –1.93 0.86 0.40 –0.10 –0.11 2.00 1.47 0.9% –3.75* –2.36 0.48 0.74 1.72 1.10 0.60* 2.01 –0.51 –1.08 2.3% 1st subsample: 1969Q1–1988Q4 –10.74* –3.20 2.40* 2.63 0.56 0.29 0.43 0.83 –0.35 –0.42 10.1% 2nd subsample: 1989Q1–2008Q1 –1.79 –0.46 4.65 1.57 0.96* 2.97 0.23 0.36 5.5% Small-cap stocks minus large-cap stocks Full sample: 1969Q1–2008Q1 –1.37 –1.31 Notes: Bolded coefficients with asterisks are statistically significant at the 10% level. Listed beneath the regression coefficients are the Newey-West-corrected t-statistics (in italics). Value minus growth is measured as the difference between the MSCI US Investable Market Value Index and the MSCI US Investable Market Growth Index. Small-cap minus large-cap is measured as the difference between the MSCI US Small Cap 1750 Index and the MSCI US Large Cap 300 Index. Source: Author’s calculations. 8 > Vanguard Investment Counseling & Research We reran the predictability regressions for future excess stock returns, this time with macroeconomic expectations represented by the four financial proxies. The results are shown in Table 4. As was the case in Table 2, the financial markets’ current expectations for future U.S. macroeconomic conditions explain less than 2% of the future volatility in the next quarter’s stock returns over the full sample period. Consistent with various academic studies, Table 4 shows that dividend yields are poor indicators of future stock returns over short horizons. Only the term spread has been significantly related to future stock returns: The steeper the Treasury yield curve is today, the higher stock returns should be next quarter, all else equal. Table 4 also reveals that the full-sample relationships change through time. For instance, Fama and French (1989) document that the dividend yield, the default spread, and the term spread predict future excess stock returns. In our 1989–2008 subsample that spans the 20-year period since the publication of Fama and French’s research, we fail to uncover any significant relationship between these proxies for expected business conditions and next quarter’s stock returns.16 We observe that the statistical significance of the coefficients, including the term spread, is period-dependent. Robustness check: Style and size returns A critic of the disappointing results in Table 4 could argue that consensus macro expectations have meaningful—albeit offsetting—correlations with the four primary style and size categories of the U.S. equity market: value and growth styles, and large- and small-cap size indexes. That is, even if consensus expectations do not meaningfully predict next quarter’s absolute stock returns, they might predict future relative returns. As an example, large-capitalization stocks are believed to outperform small-cap stocks when economic growth is expected to be weak and/or when inflation is expected to be high. Kao and Shumaker (1999) claim that that it can be extremely profitable to exploit changes in expected business conditions by tactically shifting the allocation of growth stocks and value stocks in a portfolio based on the shape of the yield curve (TERMt ). Table 5 presents the results of regressing the future return spreads for (1) value minus growth and (2) large-cap minus small-cap on the four financial indicators. Overall, the regressions are as weak for style and size premiums as they are for future stock market returns. Consensus expectations explain nothing of the time-series volatility of the value premium over our sample period. For the size premium, we do find some evidence of predictability. However, the size, magnitude, and statistical significance of the financial proxies for expected business conditions change between the two subsamples. 16 See also Tokat and Stockton (2006) for a discussion of the sensitivity of tactical signals in projecting future stock returns. Vanguard Investment Counseling & Research > 9 Figure 3. The consensus macro view is uncorrelated with future near-term stock market returns Excess stock returns: Actual versus predicted 30% 20 10 0 –10 –20 –30 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 Actual returns Predicted returns: Survey-based expectations Predicted returns: Financial indicator-based expectations These results may be surprising to some, given the considerable attention paid to such expectations by the investment community. However, our results are consistent with the view of a stock market that is reasonably efficient (it tends to “price in” the mainstream macro outlook), that is forward-looking (as a leading indicator, it tends to anticipate economic shifts rather than lag them), and that can be quite volatile in the short run for a number of reasons, including changes in investor sentiment (Fisher and Statman, 2000). In addition, consensus expectations for the economy can prove inaccurate.17 An economic game plan for investors: Four key macro considerations, plus one caveat Source: Author’s calculations from the regression models presented in Tables 2 and 4. How, then, should long-term investors think about the economy when constructing portfolios? For sophisticated The importance of focusing strategically investors, a strategic asset allocation will incorporate on macroeconomic trends expectations regarding long-term returns, volatility, and correlations among asset classes. Forming Figure 3 combines our findings and underscores such expectations requires an understanding of the that the consensus view of future macroeconomic components that make up long-term returns. For conditions—whether derived from financial indicators stocks, the primary building blocks include growth or professional forecasters—has provided little insight in corporate earnings per share and dividends per regarding the near-term direction of stock returns over share, as well as changes in market valuations, the past 40 years. We have also shown this to be true such as the price/earnings (P/E) ratio. Recalling for the return spreads of (1) value versus growth style Figure 1, the trends in these building blocks are indexes and (2) small-cap versus large-cap indexes. highly correlated with fundamental trends for The results in this paper supplement the research of productivity and inflation. Goyal and Welch (2006), who find little robust evidence of near-term stock return predictability using a wide set of financial variables. In short, foreseeing the stock market’s short-term fluctuations is an extremely difficult endeavor. 17 See, for instance, Brooks and Gray (2004). 10 > Vanguard Investment Counseling & Research As part of this assessment, investors should focus on a big-picture view of the global economic landscape. To do so requires asking several important and related questions. First, what will be the dominant themes in the economy and financial markets in the years ahead? Second, are these macro trends and concerns already priced in? This step is critical because, as we have shown, the consensus view has had little correlation with near-term stock returns. If one’s own assessment differs sufficiently from the mainstream view, then one may have identified the potential macro “surprises” that may later move the stock market. Naturally, the future success of such a macro-driven portfolio strategy will be heavily dependent on one’s accuracy in forecasting future economic trends. Figure 4. Growth in U.S. stock market fundamentals and economic growth, by decade Dividends and earnings growth per share 12 1940s 2000s 8 4 1920s 1970s 1960s 1900s 1980s 1990s 1950s 1910s 1890s 0 1880s 1870s –4 –8 1930s –2 0 2 4 6 8 10 Nominal GDP growth per capita Below we list four key macroeconomic concepts that can help investors build reasonable expectations for long-term stock returns. Macroeconomic trend 1: Productivity growth Since the 1870s, the growth in U.S. corporate earnings and dividends per share has averaged slightly over 4% per year. Figure 4 demonstrates that this growth has correlated with the average growth rate of economic productivity. In the past, productivity booms—such as those witnessed in the 1920s and 1990s—have coincided with robust growth in corporate earnings and/or dividends. Productivity booms are often driven by the rapid diffusion of general-purpose technologies that raise (at least for a time) the economy’s natural “speed limit” for noninflationary growth. Note: Data for the 1870s begin with 1872; data for the 2000s end with 2007. Source: Author’s calculations from the NBER Macro History Database and the data sources cited in Figure 1. What is the outlook for the U.S. economy’s future productivity growth? Presently, consensus expectations match the historical average: Nominal productivity is expected to grow at approximately 4.5% annually over the next ten years. That figure combines a real productivity growth rate of 2% and an average expected inflation rate of 2.5%. A key consideration for macro-minded investors is how innovation and technological progress will influence actual corporate productivity growth in the years ahead. Some key terms The F-statistic reported in the regression output is from a test of the hypothesis that all of the slope coefficients (excluding the constant, or intercept) in a regression are zero. The p-value of the F-statistic is the marginal significance level of the F-test. Predictability regression is a fitted relationship of the future values of the dependent variable regressed on (or fitted to) past values of the independent variable(s). R-squared represents the fraction of the variance of the dependent variable explained by the independent variable(s). Vanguard Investment Counseling & Research > 11 average inflation rate over the previous decade. The productivity booms of the 1920s, 1950s, 1980s, and 1990s were all associated with above-average excess stock returns in part because the productivity booms helped to engender disinflation. Figure 5. The equity risk premium and changes in realized inflation from previous decade 8 1910s 1940s Change in average inflation rate versus prior decade 6 4 1970s 1880s 1900s 2 2000s 1960s 0 1890s –2 1930s 1950s 1980s 1990s –4 1920s –6 –8 –4 0 4 8 12 16 20 Realized excess return of U.S. stocks Note: Data for the 1870s begin with 1872; data for the 2000s end with 2007. Source: Author’s calculations from the NBER Macro History Database and the data sources cited in Figure 1. Macroeconomic trend 2: Inflation Although Figure 4 shows a tight correlation between productivity growth and stock market fundamentals, this relationship has not always translated into high equity returns. A primary reason has been unexpected accelerations in the rate of trend inflation. In the 1970s, for instance, nominal output and corporate earnings appeared high by historical standards, but they were inflated by the oil-price shocks of the period. The war-related boosts in productivity during the 1910s and 1940s were also associated with a run-up in inflation. In the short and medium run, unexpected changes in the average inflation rate can have a negative effect on stock returns until monetary policy is able to return inflation expectations to a more stable and desired level.18 Figure 5 shows a negative correlation between the average realized excess return of U.S. stocks (over cash) in a decade and the change in the What is the secular outlook for U.S. inflation? Presently, the consensus long-run forecast is for an average inflation rate of 2%–3% per year, a rate consistent with the Federal Reserve Board’s objective of price stability. However, some experts challenge this baseline forecast, noting the growing global pressures on commodity prices and other inflationary forces.19 Certainly the rate of global inflation in the decades ahead will be a critical risk factor for equity investors. Macroeconomic trend 3: Productivity booms and stock market valuations Although periods of above-average economic productivity have often helped to push down inflation, they have also, on occasion, been accompanied by a rise in stock market valuation. Most notably, the information technology-related productivity boom in the late 1990s witnessed a sharp rise in P/E ratios, especially for technologyrelated companies, as investors grappled with the possible long-term benefits of rapidly diffusing technologies. Over horizons of five or ten years, P/E ratios can be powerful indicators of future stock returns (i.e., Campbell and Shiller, 1988), although Ibbotson and Chen (2003) demonstrate that P/E expansion accounted for only a small portion of the total return of U.S. stocks since 1926. Based on current valuation metrics, what is the outlook for U.S. stocks? Figure 6 reveals that forward-looking P/E ratios (as calculated by Vanguard’s Quantitative Equity Group using analyst corporate earnings forecasts) are near their historical average after having come down from the levels observed in the late 1990s. 18 See Campbell and Vuolteenaho (2004) and citations therein on the concept of “inflation illusion” and the record of stocks as a long-term inflation hedge. 19 For a discussion of evolving U.S. inflation dynamics, see Davis (2007). 12 > Vanguard Investment Counseling & Research Figure 6. Forward P/E ratio and subsequent 5-year average U.S. stock return, 1955–2007 30 25 20 15 10 5 0 –5 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 Average five-year geometric return of S&P 500 Index beginning in each year Forward P/E ratio of index Sources: Standard & Poor’s and Vanguard. Figure 7. U.S. corporations are increasingly tied to global economic conditions Percentage of U.S. corporate profits earned overseas: 1929–2008Q1 25% Macroeconomic trend 4: The global economy is increasingly relevant In the past, growth in real corporate earnings per share has been closely correlated with an economy’s own inflation-adjusted productivity growth rate. In the United States, this real rate of growth has been roughly 2% per year on average. This observation, combined with a current dividend yield of 2%, has led some practitioners to forecast a much narrower return premium for stocks over bonds and cash in the decade ahead (Arnott and Bernstein, 2002). That prognostication may prove too pessimistic, however, given that U.S. corporations are deriving a greater share of their earnings from overseas, as illustrated in Figure 7. Starting now, investors should be sure to examine trends in global growth, which has been as least twice as high as U.S. economic growth over the past several years. 20 15 10 5 0 1929 1935 1941 1947 1953 1959 1965 1971 1977 1983 1989 1995 2001 2007 Source: U.S. Bureau of Economic Analysis. Data for 2008 reflect first quarter only. Vanguard Investment Counseling & Research > 13 One caveat: Structural breaks, “black swans,” and long-run predictions It is important to remember that stock returns over the next decade are not necessarily more predictable than the returns for the next quarter. Kritzman (1994) points out that, although the annualized dispersion of stock returns moderates toward the expected mean, the dispersion of an investor’s potential terminal wealth increases as the investment horizon increases. To put this another way: Although the probability of losing money in stocks is lower over longer investment horizons than over shorter ones, the size of the potential loss increases. In surveying the debate over “time diversification,” Bennyhoff (2008) finds little empirical evidence to support the claim that time moderates the risks inherent in risk assets. Over the past 200 years, the American economy and its financial system have undergone fantastic changes and challenges. Amazingly, real stock returns have been remarkably similar across long periods, such as between the 19th and 20th centuries. Of course, going forward, an unforeseen structural break or “black swan” could fundamentally alter the U.S. economy and its role in the global financial system. Examples could include major armed conflicts, drastic changes in the tax code or the regulatory environment, technological change that fundamentally alters the nation’s comparative advantages, monetary policy mistakes, or even simply bad luck. For these reasons, global portfolio diversification remains critical, and it is important for strategic investors to periodically revisit their macro-based assumptions regarding expected long-term stock returns in the years ahead. References Andersen, Torben G., Tim Bollerslev, Francis X. Diebold, and Jin Wu, 2005. A Framework for Exploring the Macroeconomic Determinants of Systematic Risk. American Economic Review 95(2, May): 398–404. Arnott, Robert D., and Peter L. Bernstein, 2002. What Risk Premium Is “Normal?” Financial Analysts Journal 58(2): 64–85. Asness, Clifford, 2003. Fight the Fed Model: The Relationship Between Future Returns and Stock and Bond Market Yields. Journal of Portfolio Management 30(Fall): 11–24. Avramov, Doron, 2002. Stock Return Predictability and Model Uncertainty. Journal of Financial Economics 64: 423–58. Bennyhoff, Donald G., 2008. Time Diversification and Horizon-Based Asset Allocations. Valley Forge, Pa.: Vanguard Investment Counseling & Research, The Vanguard Group. Brooks, Robert, and J. Brian Gray, 2004. History of the Forecasters. Journal of Portfolio Management 31(Fall): 113–17. Campbell, John Y., and Motohiro Yogo, 2003. Efficient Tests of Stock Return Predictability. NBER Working Paper No. 10026. Cambridge, Mass.: National Bureau of Economic Research. Campbell, John Y., and Robert J. Shiller, 1988. The Dividend-Price Ratio and Expectations of Future Dividends and Discount Factors. Review of Financial Studies 1: 195–227. Campbell, John Y., and Samuel B. Thompson, 2005. Predicting the Equity Premium Out of Sample: Can Anything Beat the Historical Average? NBER Working Paper No. 11468. Cambridge, Mass.: National Bureau of Economic Research. 14 > Vanguard Investment Counseling & Research Campbell, John Y., and Tuomo Vuolteenaho, 2004. Inflation Illusion and Stock Prices. NBER Working Paper No. 10263. Cambridge, Mass.: National Bureau of Economic Research. Inoue, Atsushi, and Lutz Kilian, 2004. In-Sample or Out-of-Sample Tests of Predictability: Which One Should We Use? Econometric Reviews 23: 371–402. Campbell, Sean D., and Francis X. Diebold, 2005. Stock Returns and Expected Business Conditions: Half a Century of Direct Evidence. NBER Working Paper No. 11736. Cambridge, Mass.: National Bureau of Economic Research. Kao, Duen-Li, and Robert D. Shumaker, 1999. Equity Style Timing. Financial Analysts Journal 55(1): 37–48. Chen, Nai-Fu, Richard Roll, and Stephen A. Ross, 1986. Economic Forces and the Stock Market. Journal of Business 59: 383–403. Croushore, Dean, 1993. Introducing: The Survey of Professional Forecasters. Business Review (Federal Reserve Bank of Philadelphia), November/ December: 3–13. Croushore, Dean, 2006. An Evaluation of Inflation Forecasts From Surveys Using Real-Time Data. Working paper. Philadelphia: Federal Reserve Bank of Philadelphia. Davis, Joseph H., 2007. Evolving U.S. Inflation Dynamics: Explanations and Investment Implications. Valley Forge, Pa.: Vanguard Investment Counseling & Research, The Vanguard Group. Fama, Eugene F., and Kenneth R. French, 1989. Business Conditions and Expected Returns on Stocks and Bonds. Journal of Financial Economics 25: 23–49. Fisher, Kenneth L., and Meir Statman, 2000. Investor Sentiment and Stock Returns. Financial Analysts Journal 56(2): 16–23. Goyal, Amit, and Ivo Welch, 2006. A Comprehensive Look at the Empirical Performance of Equity Premium Prediction. Review of Financial Studies, forthcoming. Kritzman, Mark, 1994. What Practitioners Need to Know. . . About Time Diversification. Financial Analysts Journal 50(1): 14–18. Lettau, Martin, and Sydney Ludvigson, 2001. Consumption, Aggregate Wealth, and Expected Stock Returns. Journal of Finance 56(3): 815–849. Mankiw, N. Gregory, Ricardo Reis, and Justin Wolfers, 2003. Disagreement About Inflation Expectations. NBER Macroeconomics Annual 2003. Resnick, Bruce G., and Gary L. Shoesmith, 2002. Using the Yield Curve to Time the Stock Market. Financial Analysts Journal 58(3): 82–90. Rudebusch, Glenn D., and John C. Williams, 2007. Forecasting Recessions: The Puzzle of the Enduring Power of the Yield Curve. Federal Reserve Bank of San Francisco Working Paper No. 2007-16. San Francisco: Federal Reserve Bank of San Francisco. Schwert, G. William, 1990. Stock Returns and Real Activity: A Century of Evidence. Journal of Finance 45(4): 1237–57. Stambaugh, Robert F., 1999. Predictive Regressions. Journal of Financial Economics 54(3): 375–421. Tokat, Yesim, and Kimberly A. Stockton, 2006. A Primer on Tactical Asset Allocation Strategy Evaluation. Valley Forge, Pa.: Vanguard Investment Counseling and Research, The Vanguard Group. Ibbotson, Roger G., and Peng Chen, 2003. LongRun Stock Returns: Participating in the Real Economy. Financial Analysts Journal 59(1): 88–98. Vanguard Investment Counseling & Research > 15 P.O. Box 2600 Valley Forge, PA 19482-2600 Connect with Vanguard® > www.vanguard.com Vanguard, Connect with Vanguard, and the ship logo are trademarks of The Vanguard Group, Inc. CFA ® is a trademark owned by CFA Institute. All other marks are the exclusive property of their respective owners. E-mail > [email protected] Contributing authors John Ameriks, Ph.D./Principal Joseph H. Davis, Ph.D./Principal Francis M. Kinniry Jr., CFA/Principal Roger Aliaga-Diaz, Ph.D. Donald G. Bennyhoff, CFA Maria A. Bruno, CFP ® Scott J. Donaldson, CFA, CFP ® Michael Hess Julian Jackson Colleen M. Jaconetti, CPA, CFP ® Karin Peterson LaBarge, Ph.D., CFP ® Christopher B. Philips, CFA Liqian Ren, Ph.D. Kimberly A. Stockton David J. Walker, CFA Investment products: Not FDIC-insured • No bank guarantee • May lose value © 2008 The Vanguard Group, Inc. All rights reserved. ICRMACR 092008