Economic Indicators

Which economic and market variables are most effective in predicting U.S. stock market returns? In his October 2018 paper entitled “Forecasting US Stock Returns”, David McMillan tests 10-year rolling and recursive (inception-to-date) one-quarter-ahead forecasts of S&P 500 Index capital gains and total returns using 18 economic and market variables, as follows: dividend-price ratio; price-earnings ratio; cyclically adjusted price-earnings ratio; payout ratio; Fed model; size premium; value premium; momentum premium; quarterly change in GDP, consumption, investment and CPI; 10-year Treasury note yield minus 3-month Treasury bill yield (term structure); Tobin’s q-ratio; purchasing managers index (PMI); equity allocation; federal government consumption and investment; and, a short moving average. He tests individual variables, four multivariate combinations and and six equal-weighted combinations of individual variable forecasts. He employs both conventional linear statistics and non-linear economic measures of accuracy based on sign and magnitude of forecast errors. He uses the historical mean return as a forecast benchmark. Using quarterly S&P 500 Index returns and data for the above-listed variables during January 1960 through February 2017, he finds that: Keep Reading

Each month, the Census Bureau announces and the financial media report U.S. new home sales as a potential indicator of future U.S. stock market returns. Release date is about three weeks after the month being reported. Moreover, new releases may substantially revise recent past releases, so that the Census Bureau historical data set effectively has a longer lag. Does this economic indicator convey useful information about future returns for the broad U.S. stock market or for Real Estate Investment Trusts (REIT)? To investigate, we relate returns for the S&P 500 Index (SP500) and for the FTSE NAREIT All REITs total return index (REITs) to changes in new home sales at the monthly release frequency. Using monthly data for SP500 and for seasonally adjusted annualized new homes sales starting January 1963, and for REITs starting December 1971, all through September 2018, we find that: Keep Reading

Each month, the Census Bureau announces and the financial media report U.S. housing starts as a potential indicator of future U.S. stock market returns. Release date is about two weeks after the month being reported. New releases may substantially revise recent past releases, so that the Census Bureau historical data set effectively has a longer lag. Does this economic indicator convey useful information about future returns for the broad U.S. stock market or for Real Estate Investment Trusts (REIT)? To investigate, we relate returns for the S&P 500 Index (SP500) and for the FTSE NAREIT All REITs total return index (REITs) to changes in housing starts at the monthly release frequency. Using monthly data for SP500 and for seasonally adjusted annualized housing starts starting January 1959, and for REITs starting December 1971, all through September 2018, we find that:

Keep Reading

Which economic variables are most important for predicting stock returns? In their October 2018 paper entitled “Sparse Macro Factors”, David Rapach and Guofu Zhou apply machine learning to isolate via sparse principal component analysis (PCA) which of 120 economic variables from the FRED-MD database most influence stocks. These variables span output/income, labor market, housing, consumption, orders/inventories, money/credit, yields/exchange rates and inflation. As a preliminary step, they adjust raw economic variables by, where necessary: (1) transforming them to produce stationary series; (2) adjusting for reporting lags of one or two months. They next execute sparse PCA, which sets small component weights to zero, thereby facilitating interpretation of results without sacrificing much predictive power. For comparison, they also extract the first 10 conventional principal components from the same variables. Finally, they use 202 stock portfolios to estimate the influence of sparse and conventional principal components on the cross section of stock returns. Using monthly data for the 120 economic variables and 202 stock portfolios during February 1960 through June 2018, they find that:

Keep Reading

When the U.S. government runs substantial deficits, some experts proclaim the dollar’s inevitable inflationary debasement and bad times for stocks. Other experts say that deficits are no cause for alarm, because government spending stimulates the economy, and the country can bear more debt. Who is right? Using annual (end of fiscal year, FY) level of the U.S. public debt, interest expense on the debt, U.S. Gross Domestic Product (GDP), Dow Jones Industrial Average (DJIA) return and inflation rate as available during June 1929 through September 2018 (about 89 years), we find that: Keep Reading

A reader commented and asked: “A wide credit spread (the difference in yields between Treasury notes or Treasury bonds and investment grade or junk corporate bonds) indicates fear of bankruptcies or other bad events. A narrow credit spread indicates high expectations for the economy and corporate world. Does the credit spread anticipate stock market behavior?” To investigate, we define the U.S. credit spread as the difference in yields between Moody’s seasoned Baa corporate bonds and 10-year Treasury notes (T-note), which are average daily yields for these instruments by calendar month (a smoothed measurement). We use the S&P 500 Index (SP500) as a proxy for the U.S. stock market. We extend the investigation to bond market behavior via:

• Vanguard Long-Term Treasury Investors Fund (VUSTX)
• Vanguard Long-Term Investment-Grade Investors Fund (VWESX)
• Vanguard High-Yield Corporate Investors Fund (VWEHX)

Using monthly Baa bond yields, T-note yields and SP500 closes starting April 1953 and monthly dividend-adjusted closes of VUSTX, VWESX and VWEHX starting May 1986, January 1980 and January 1980, respectively, all through August 2018, we find that: Keep Reading

A subscriber asked (more than six years ago): “Everyone says I should not invest in bonds today because the interest rate is so low (and inflation is daunting). But real bond returns over the last 30 years are great, even while interest rates are low. Could you analyze why bonds do well after, but not before, 1981?” To investigate, we consider the U.S. long-run interest rate and the U.S. Consumer Price Index (CPI) series from Robert Shiller. The long-run interest rate is the yield on U.S. government bonds, specifically the constant maturity 10-year U.S. Treasury note after 1953. We use the term “T-note” loosely to refer to the entire series. We apply the formula used by Aswath Damodaran to the yield series to estimate the nominal T-note total returns. We use the CPI series to calculate inflation (12-month change in CPI). We subtract inflation from the T-note nominal total return to get the T-note real total return. Using annual Shiller interest rate and CPI data for 1871 through 2017, we find that: Keep Reading

The Mojena Market Timing strategy (Mojena), developed and maintained by professor Richard Mojena, is a method for timing the broad U.S. stock market based on a combination of many monetary, fundamental, technical and sentiment indicators to predict changes in intermediate-term and long-term market trends. He adjusts the model annually to incorporate new data. Professor Mojena offers a hypothetical backtest of the timing model since 1970 and a live investing test since 1990 based on the S&P 500 Index (with dividends). To test the robustness of the strategy’s performance, we consider a sample period commencing with inception of SPDR S&P 500 (SPY) as a liquid, low-cost proxy for the S&P 500 Index. As benchmarks, we consider both buying and holding SPY (Buy-and-Hold) and trading SPY with crash protection based on the 10-month simple moving average of the S&P 500 Index (SMA10). Using the trade dates from the Mojena Market Timing live test, daily dividend-adjusted closes for SPY and daily yields for 13-week Treasury bills (T-bills) from the end of January 1993 through August 2018 (over 25 years), we find that: Keep Reading

A subscriber requested testing of the relationship between U.S. M2 Money Stock and gold, offered in one form via “Why Gold May Be Looking Cheap”: “[O]ne measure I’ve found useful is the ratio of the price of gold to the U.S. money supply, measured by M2, which includes cash as well as things like money market funds, savings deposits and the like. The logic is that over the long term the price of gold should move with the change in the supply of money… That equilibrium level is also relevant for future price action. When the ratio is low, defined as 25% below equilibrium, the medium 12-month return has been over 12%. Conversely, when the ratio is high, defined as 25% above equilibrium, the 12-month median return has been -6%. …This measure can be refined further. [G]old tends to trade at a higher ratio to M2 when inflation is elevated.” Because it defines specific valuation thresholds, this approach is susceptible to data snooping bias in threshold selection. We consider an alternative setup that relates monthly change in M2 to monthly gold return. We also consider the effect of inflation on this relationship. Using monthly seasonally adjusted M2 and end-of-month London gold price fix during January 1976 (to ensure a free U.S. gold market) through June 2018 (510 months), we find that: Keep Reading

Each week the media report U.S. initial and continued unemployment claims (seasonally adjusted) as a potential indicator of future U.S. stock market returns. Do these indicators move the market? To investigate, we focus on weekly changes in unemployment claims during a period of “modern” information dissemination to release-day and next-week stock market returns. By modern period, we mean the history of S&P Depository Receipts (SPY), a proxy for the U.S. stock market. Using relevant news releases and archival data as available from the Department of Labor (DOL) and dividend-adjusted weekly and daily opening and closing levels for SPY during late January 1993 through mid-July 2018 (1,330 weeks), we find that:

Keep Reading