Tactical Asset Allocation session 5 Andrei Simonov Tactical Asset Allocation 1 3/21/2016 Agenda What is tactical asset allocation? Mean-variance perspective on TAA and SAA Predictability – – – – – January dummy Business cycle variables Explaining risk premia: US, World, Sweden. Currency risk premia Caveats: data snooping, statistical issues. Tactical Asset Allocation 2 3/21/2016 What is TAA? Exists since early-to-mid- 80-ies. By now $100-200 bln are under management by TAA managers A TAA managers’s investment objective is to obtain better-than-expected return with (possibly) lower-thanbenchmark volatility by forecasting the returns of two or more asset classes and varying asset class exposure in systematic manner (Phillips, Rogers & Capaldi, 1996) Can TAA funds be interpreted as stand-alone asset class? Tactical Asset Allocation 3 3/21/2016 Conditioning Information and Portfolio Analysis Er Add conditioning information and weights change through time. Frontier shifts. Vol Tactical Asset Allocation 4 3/21/2016 Optimal portfolio for risk-averse investor max w E (R ) T 2 w T Vw s. t. w T 1 1 11 .. 1N T T Here w ( w1 , w2 ,...),1 (1,1,1,...,1), V .. .. .. .. NN N1 min L w T E (R ) w T Vw 1 w T 1 2 E (R ) Vw 1 0 V 1 E (R ) 1 w . Summing up : T 1 w 1 0 1T V 1 E (R ) T 1 T 1 1 V 1 1 V 1 V 11 V 1 1T V 1 E (R ) * E (R ) 1 T 1 w T 1 1 V 1 1 V 1 Global min var portfolio Tactical Asset Allocation 5 3/21/2016 Equilibrium and TAA Let us assume that there exists long-term expected returns vector e. However, due to predictability of asset returns, eE(R) V 11 V 1 e1T 1eT V 1 V 1 1T 1T V 1 w T 1 1 1 T 1 T 1 1 V 1 1 V 1 1 V 1 * Global min var portfolio StrategicBet TacticalBet 0 E rn en ( E r1 e1 ) 1T 1T E r1 e1 ( E rj e j ) 0 E r1 e1 ( E rn en ) 0 Tactical Asset Allocation 6 3/21/2016 How to do it? We need a model that explains the connection between today’s variables and tomorrow returns. Candidates: economic business cycle variables and Jan. Effect. Tactical Asset Allocation 7 3/21/2016 Example: Incredible January Effect Excess returns associated with small firms w.r.t. Large-cap stocks Ritter: Tax effect. Is it so? Incredibly Shrinking January Effect (William J. Bernstein ). Tactical Asset Allocation 8 3/21/2016 Example: dividend yield Fama-French (1988). 1927-1986 Holding Coeff. period M 0.21 Q 1.07 1 2.47 2 7.38 3 9.94 4 12.86 t(coeff) 1.40 2.10 1.27 2.04 2.21 2.43 R2 0.00 0.01 0.01 0.09 0.13 0.19 • May not be sustained out of sample Tactical Asset Allocation 9 3/21/2016 Risk and return over the business cycle m,t Et Rm rt m vart Rm ???? G-7 output, 1973Q2 to 1996Q2 output level potential line Average returns Return volatility end. recess beg. expan end. expan beg. recess 15.23% 10.36% 6.96% 2.86% Tactical Asset Allocation 12.59% 10.63% 16.85% 26.98% 10 3/21/2016 Evaluation of Recent Recession In July 2000, the Yield Curve inverted forecasting recession to begin in June 2001. Official NBER Peak is March 2001 (Yield Curve within one quarter accurate). In March 2001, the Yield Curve returned to normal forecasting the end of the recession in November 2001. On July 17, 2003 the NBER announced the official end of the recession was November 2001. Tactical Asset Allocation 11 3/21/2016 Exhibit 1 Next couple of slides are due to Cam Harvey Lead Lag Analysis in Months Business Cycle 5-Year Yield Spread NBER NBER Length Length of Peak Trough of Cycle Inversion Lead Normal Lead Inversion Dec-69 Nov-70 11 Oct-68 14 Feb-70 9 16 Nov-73 Mar-75 16 Jun-73 5 Jan-75 2 19 Jan-80 Jul-80 6 Nov-78 14 May-80 2 18 Jul-81 Nov-82 16 Oct-80 9 Oct-81 13 12 Jul-90 Mar-91 8 May-89 14 Feb-90 13 9 Average last four 11 11 7 15 Recent Recession Mar-01 Nov-01 Tactical Asset Allocation 8 Jul-00 8 Mar-01 8 8 12 3/21/2016 Exhibit 2 Forecast evaluation Term Structure Inversion Date Jul-2000 Average Forecast Actual Lead to Beginning Recession Recession of Recession Begins 11 Term Structure Average Normal Date Lead Mar-2001 Tactical Asset Allocation 8 Jun-2001 Mar-2001 Forecast End of Recession Error 3 Actual End Error Nov-2001 Nov-2001 0 13 3/21/2016 Yield Curve Inverts Before Last Six Recessions (5-year Treasury note minus 3-month Treasury bill yield-secondary) Annual GDP growth or Yield Curve % Source: Campbell R. Harvey. % Real annual GDP growth 8 6 4 2 Yield curve 0 -2 Recession Correct Recession -4 Correct 2 Recessions Correct Recession Correct Recent flattening Yield curve accurate in recent recession Data though April 11, 2006 19 68 19 70 19 72 19 74 19 76 19 78 19 80 19 82 19 84 19 86 19 88 19 90 19 92 19 94 19 96 19 98 20 00 20 02 20 04 20 06 -6 Tactical Asset Allocation 14 3/21/2016 Yield Curve Inverts Before Last Six Recessions (5-year Treasury note minus 3-month Treasury bill yield – constant maturity) Annual GDP growth or Yield Curve % 8 % Real annual GDP growth Source: Campbell R. Harvey. 6 4 2 Yield curve 0 -2 Recession Correct Recession -4 Correct 2 Recessions Correct Recession Correct Recent flattening Yield curve accurate in recent recession Data though April 11, 2006 19 68 19 70 19 72 19 74 19 76 19 78 19 80 19 82 19 84 19 86 19 88 19 90 19 92 19 94 19 96 19 98 20 00 20 02 20 04 20 06 -6 Tactical Asset Allocation 15 3/21/2016 Recent Annualized One-Quarter GDP Growth (10-year and 5-year Yield Curves-secondary market) Annualized 1-quarter GDP growth 8 10-year Yield curve 4 % Real annualized one-quarter GDP growth 6 3 4 2 2 1 0 0 -2 5-year Both curves invert 2000Q3 -1 Data though April 11, 2006 -2 M ar -9 Se 5 p9 M 5 ar -9 Se 6 p9 M 6 ar -9 Se 7 p9 M 7 ar -9 Se 8 p9 M 8 ar -9 Se 9 p9 M 9 ar -0 Se 0 p0 M 0 ar -0 Se 1 p0 M 1 ar -0 Se 2 p0 M 2 ar -0 Se 3 p0 M 3 ar -0 Se 4 p0 M 4 ar -0 Se 5 p0 M 5 ar -0 6 -4 Tactical Asset Allocation 16 3/21/2016 Recent Annualized One-Quarter GDP Growth (10-year and 5-year Yield Curves-constant maturity) Annualized 1-quarter GDP growth 8 10-year Yield curve 4 % Real annualized one-quarter GDP growth 6 3 4 2 2 1 0 0 -2 5-year Both curves invert 2000Q3 -1 Data though April 2006 -2 M ar -9 Se 5 p9 M 5 ar -9 Se 6 p9 M 6 ar -9 Se 7 p9 M 7 ar -9 Se 8 p9 M 8 ar -9 Se 9 p9 M 9 ar -0 Se 0 p0 M 0 ar -0 Se 1 p0 M 1 ar -0 Se 2 p0 M 2 ar -0 Se 3 p0 M 3 ar -0 Se 4 p0 M 4 ar -0 Se 5 p0 M 5 ar -0 6 -4 Tactical Asset Allocation 17 3/21/2016 What shall we expect now? US yield curves, 2006 5.4 5.2 5 4.8 4.6 4.4 1/3/2006 4/3/2006 7/3/2006 8/29/2006 4.2 4 1 mo 3 mo 6 mo Tactical Asset Allocation 1 yr 2 yr 3 yr 5 yr 7 yr 10 yr 20 yr 30 yr 18 3/21/2016 May 2007: Practically flat 5.1 5 4.9 4.8 4.7 5-5.1 4.6 4.9-5 4.5 4.8-4.9 4.4 4.7-4.8 4.3 4.6-4.7 4.5-4.6 4.2 5/1/2007 5/3/2007 4.4-4.5 4.3-4.4 5/7/2007 5/9/2007 4.2-4.3 5/11/2007 5/15/2007 5/17/2007 5/21/2007 1mo Tactical Asset Allocation 3mo 6mo 1yr 2yr 3yr 5yr 7yr 10yr 20yr 30yr 19 3/21/2016 August 2007 5-6 4-5 6 3-4 2-3 5 1-2 0-1 4 3 2 1 0 Tactical Asset Allocation 20 3/21/2016 Current Situation: Economic growth •The economy expanded at an annual pace of 4.1%, the most in more than a year, according to the median estimate of 81 economists surveyed by Bloomberg News. The Commerce Department last month calculated the growth rate at 3.4%. • But the outlook for the second half of 2007 has soured in recent weeks as the subprime mortgage crisis has restricted access to credit. The Federal Reserve this month said risks to growth had ``increased appreciably'' and economists at JPMorgan and Lehman are among those that have reduced forecasts. •There are growing signs of a housing slowdown; new home sales down, housing prices down, and homeowners with ARMs facing much higher interest rates. Tactical Asset Allocation 21 3/21/2016 Current Situation Inflation perceptions. The long-term rate is a combination of expected inflation, expected real interest rates and an inflation risk factor. Long-term inflation expectations have decreased mainly due to the glut of cheap labor resulting from globalization. Tactical Asset Allocation 22 3/21/2016 Current Situation Strong buying of long-term bonds by foreigners. For the past few years, strong buying by Asian central banks have pushed up the Treasury bond prices. However, there is a debate as to whether this has had a large impact on bond prices. In addition, this buying has flattened out recently. A recent Fed study estimated that the foreign buying pushed yields down by 150bp. Subprime crisis does not end buying of T-debt by foreigners. Demand for 5yr TB last week was very high. Tactical Asset Allocation 23 3/21/2016 Current Situation Hedge funds. There has been a recent increase in demand for U.S. bonds from the Caribbean area indicating hedge fund activity. With long-rates above short rates, many managers do “carry trades” (borrow short-term and buy long-term bonds hoping the relation between rates remains stable). As the term structure flattens, many of these managers increase their leverage which means more buying pressure on the long-term bonds. Tactical Asset Allocation 24 3/21/2016 Current Situation Demographic forces. As the population ages, more money is allocated into fixed income and long-term bond yields may decrease. Inflation risk. The long-rate rates contain expected inflation, expected real rates and an inflation risk factor. It is widely perceived that inflation risk (an unexpected episode of inflation turbulence) has decreased. Tactical Asset Allocation 25 3/21/2016 Annual Real Economic Growth After Yield Curve Inversions 4.50% 4.00% 3.50% 3.00% 2.50% 2.00% 1.50% 1.00% 0.50% 0.00% Up to one year after inversions Tactical Asset Allocation Other quarters 26 3/21/2016 Stock Returns and U.S. Yield Curve Average Monthly Returns in % 3 2.5 2 1.5 1 0.5 0 Data through November 2000 Tactical Asset Allocation Inversion W O US UK CH SE ES SG NO NL JP IT HK DE FR DK CA BE AT AU -0.5 Normal 27 3/21/2016 Average Monthly Stock Returns After Yield Curve Inversions 1.40 1.20 1.00 0.80 Equally weighted 0.60 Value weighted 0.40 0.20 0.00 After first month of inversion Normal Based on 19 countries. Tactical Asset Allocation 28 3/21/2016 Trader’s calendar (from thestreet.com) Time Indicator (EST) (click for definition) Source (click for press release) Forecast Previous (revised) Previous (original) n.a. +0.8% +0.8% n.a. +2.2% +2.5%* n.a. -- 675.5 -- n.a. -- 432.3 -- n.a. -- -7 -- +305,000 -- +293,000 --- n.a. +0.9% --- +306,000 +3.7% -- n.a. -- +1.5% Census Bureau -- .860M -- .858M Bureau of the Public Debt The Treasury announces the size of its next monthly two-year note auction, next Tuesday. National Association of Realtors -- 6.10M -- 6.12M Economic Cycle Research Institute -- n.a. -- +6.1% Actual Monday, May 21 No releases. Tuesday, May 22 9 a.m. ICSC-UBS Weekly Chain Store Sales Snapshot for International Council of Shopping Centers -1.5% the week ended May 19 and UBS 9 a.m. Johnson Redbook Retail Sales Index for the week Redbook Research +2.0% ended May 19, vs. April Wednesday, May 23 9 a.m. Mortgage Applications Survey for the week ended -May 18 -- Market Composite Index Mortgage Bankers Association Purchase Index 9 a.m. Consumer Comfort Index for the week ended May 20 Thursday, May 24 8:30 Initial Jobless Claims for the week ended May 19 a.m. 8:30 a.m. Four-week average Durable goods orders for April Ex-transportation 10 a.m. New home sales for April 2:30 p.m. Treasury auction announcement Friday, May 25 10 a.m. Existing Home Sales for April 10:30 Weekly Leading Index for the week ended May 18 a.m. Tactical Asset Allocation ABC News and Washington Post Labor Department Census Bureau 29 3/21/2016 What variables matter? Methodology: 1. Exploratory: regressing returns at t on informational variables at t-1 2. ”Correct one”: first finding economic risk premia (a la APT) and then regressing it on informational variables at t-1 Tactical Asset Allocation 30 3/21/2016 Do informational variables have predictive ability? Info variables: – January dummy – Past excess return on Equally weighted CRSP index – Spread between 1 and 3 mo Tbills – Dividend yield – Spread between Baa and Aaa corporate bonds – 1-mo T-bill rate Tactical Asset Allocation 31 3/21/2016 Tactical Asset Allocation Here how it looks like... 32 3/21/2016 Performance & Business Cycle Average Annual Returns During U.S. Business Cycle Phases 30 20 10 0 -10 -20 A us tr A alia u Be stri lg a Ca ium D nad en a m Fi ark nl Fr and G an H erm ce on a g ny K Ire ong la n Ita d N J ly e N the apa ew rl n Ze and a s N l and o Po rwa rtu y g Sp al S Sw we ain it z de er n la nd U K U W or W S ld or ex ld -U EA S FE -30 Expansion geometric mean Data through June 2002 Tactical Asset Allocation Recession geometric mean 33 3/21/2016 us tr A alia u Be stri lg a Ca ium D nad en a m Fi ark nl Fr and G an H erm ce on a g ny K Ire ong la n Ita d N J ly e N the apa ew rl n Ze and a s N l and or Po wa rtu y g Sp al Sw Swe ain it z de er n la nd U K U W or W S ld or ex ld -U EA S FE A Performance & Business Cycle (2) Average Annual Volatility During U.S. Business Cycle Phases 60 50 40 30 20 10 0 Expansion std.dev. Data through June 2002 Tactical Asset Allocation Recession std.dev. 34 3/21/2016 Performance & Business Cycle (3) Correlations During U.S. Business Cycle Phases 1 0.8 0.6 0.4 0.2 0 A us tr A alia u Be stri lg a Ca ium D nad en a m Fi ark nl Fr and G an H erm ce on a g ny K Ire ong la n Ita d N J ly e N the apa ew rl n Ze and a s N l and or Po wa rtu y g Sp al Sw Swe ain it z de er n la nd U K U W or W S ld or ex ld -U EA S FE -0.2 Expansion correlation with US Data through June 2002 Tactical Asset Allocation Recession correlation with US 35 3/21/2016 3. Performance & Business Cycle (4) Covariances During U.S. Business Cycle Phases 45 40 35 30 25 20 15 10 5 A us tr A alia u Be stri lg a Ca ium D na en da m Fi ark nl Fr and G an H erm ce on a g ny K Ire ong la n Ita d N J ly e N the apa ew rl n Ze and a s N l and or Po wa rtu y g Sp al Sw Swe ain it z den er la nd U K U W or W S ld or ex ld -U EA S FE 0 Expansion covariance with US Data through June 2002 Tactical Asset Allocation Recession covariance with US 36 3/21/2016 How important are global factors? Based on Ferson-Harvey RFS95 Question here is: what is more important, local or global factors for predictability of asset returns. Global Informational variables: : ”old friends”: 1 mo t-bill, div. Yield on MSCI World index, spread between 10yr and 3 mo Tbills, Eurodollar/US treasury spread, lagged market return, January dummy. Local informational variables: Country x div. Yield, 30-day tbill rate, term spread, lagged MSCI country x market return. E Rit Z t 1 0 Z t 1 ij Z t 1 j Z t 1 0l Z t 1,l K L j 1 l 1 L L ijl Z t 1,l jm Z t 1,m j 1 l 1 m1 K Tactical Asset Allocation 37 3/21/2016 So, what matters? ”Global only” model is already good enough Adding local factors increases explanatory power of the model Tactical Asset Allocation 38 3/21/2016 Changes in vs changes in risk premium Var E ' Z E ( ' )Var E Z E ( ) E ( ' )Var E Z E ( ) Only 2-4% of variation is due to beta’s. Tactical Asset Allocation 39 3/21/2016 Sweden (Robertsson, 2000): Sweden Market Index Small Stocks Bond Index bond –2.05 -1.09 –2.91 -1.27 –1.18 -0.32 Tactical Asset Allocation bill –1.02 -0.64 –0.75 -0.5 0.37 -0.18 mat –0.42 -0.62 1.19 -0.65 0.02 -0.13 def 3.19 -2.27 0.65 -2.17 0.7 -0.56 fx 1.26 -0.5 –0.05 -0.58 0.14 -0.14 irs 1.24 -0.58 –0.02 -0.61 0.32 -0.14 ey 0.35 -0.29 –0.67 -0.34 0.1 -0.08 mb 0.01 -0.03 0.08 -0.05 –0.01 -0.01 irw –6.50 -11.5 –13.6 -10 4.19 -2.41 R2 6.1 [0.00] 16.8 [0.00] 13.1 [0.00] 40 3/21/2016 What about currency risk premium? Currency specificiyy: zero-sum game Dumas-Solnik: currency risk premia exists. It is time-varying and predictable Tactical Asset Allocation 41 3/21/2016 Caveats: Data snooping – Foster, Smith and Whaley (98): by choosing to max R2 via choice of instruments one can get significance when there is none. – Not clear how to use as list of instruments already exists... In-sample vs. Out-of-sample validation Tactical Asset Allocation 42 3/21/2016 Caveats(2) Statistical biases: autocorrelation, heteroscedastisity (via Monte-Carlo simulations). Non-normality, excess skewness and kurtosis Tactical Asset Allocation 43 3/21/2016 How to deal with statistical issues? Bootstrap methodology: – – – – Form empirical distribution of returns Generate time series of returns (length T). Perform the regression of interest See how many times there exists significance on level a. Tactical Asset Allocation 44 3/21/2016 U.S. Risk Premium Survey Background Graham/Harvey: Survey CFOs every quarter Q2 2000 through Q4 2003 (15 quarters) Current survey attracts about 400 respondents Why CFOs? – We know from previous surveys and interviews that the CFOs use the risk premium for their capital budgeting – Hence, they have thought hard about risk premium – Should not be biased the way that analyst forecasts might be Tactical Asset Allocation 45 3/21/2016 U.S. Risk Premium One-Year Premium One-year risk premium variable. Currently, about 7% Tactical Asset Allocation 46 3/21/2016 U.S. Risk Premium Ten-Year Premium Ten-year risk premium is stable. Currently, about 3.7% Tactical Asset Allocation 47 3/21/2016 U.S. Risk Premium Momentum in Expectations for 1-year Premium 8 Mean one-year premium 7 6 5 4 y = 0.1912x + 3.8912 3 R2 = 0.5242 2 1 0 -15 -10 -5 0 5 10 15 Excess S&P 500 return in previous two months Tactical Asset Allocation 48 3/21/2016 U.S. Risk Premium Extreme Returns Cause Disagreement A. Disagreement over the one-year premium and past returns y = 0.0194x2 + 0.0247x + 3.3696 R2 = 0.5892 Disagreement over the one-year premium 6 5 4 3 y = -0.0614x + 3.9079 R2 = 0.1684 2 1 0 -15 Tactical Asset Allocation -10 -5 0 Past one-month excess S&P 500 return 5 10 49 3/21/2016 U.S. Risk Premium Positive Relation Between Disagreement and Expected 10-year Returns B. Ten-year premium and disagreement 8 Mean ten-year premium 7 6 5 4 3 2 y = 0.9777x + 1.5936 R2 = 0.3165 1 0 1.5 1.7 1.9 2.1 2.3 2.5 Disagreement of ten-year premium forecasts Tactical Asset Allocation 2.7 2.9 50 3/21/2016 U.S. Risk Premium Example Confidence Intervals: September 16, 2002 95% Confidence Interval Median Min Mean Standard deviation Over the next 10 years, I expect the average annual S&P 500 return will be: There is a 1-in10 chance it will be less than: 3.65 2.35 3.40 - 3.89 4 Over the next 10 years, I expect the average annual S&P 500 return will be: Expected return: 7.81 2.19 7.58 - 8.03 Over the next 10 years, I expect the average annual S&P 500 return will be: There is a 1-in10 chance it will be greater than: 11.5 3.33 Over the next year, I expect the average annual S&P 500 return will be: There is a 1-in10 chance it will be less than: -2.98 Over the next year, I expect the average annual S&P 500 return will be: Expected return: Over the next year, I expect the average annual S&P 500 return will be: There is a 1-in10 chance it will be greater than: Max Total -3 10 351 8 0 15 373 11.15 - 11.84 11 4 20 355 6.86 -3.7 - -2.26 0 -20 10 348 4.95 2.78 4.66 - 5.24 5 0 12 345 9.96 4.56 9.47 - 10.44 10 0 20 343 Notes: 10-year bond yield 3.9%; 1-year bill yield 1.6%. Confidence interval based on standard deviation of the mean. Tactical Asset Allocation 51 3/21/2016 Conclusion: TAA can be an important tool in asset allocation methodology. It is based on time variation of real economic risk premia. Selection of predictors is important. We are still in ”top-down” paradigm. Devil is in the details= implementation matters. Tactical Asset Allocation 52 3/21/2016