Economic Recoveries since WWII & forecasting magnitude of current Recovery Guy Lion January 2010
Introduction <ul><li>Here we focus on change in GDP and unemployment rate level to observe trends in economic recoveries since WWII. </li></ul><ul><li>In turn, we use this information to forecast the strength of the current recovery. </li></ul>
Definition of Recession <ul><li>I do not use NBER complex definition of recession that includes more than half dozen economic indicators that often give contradictory signals as to the timing of a recession. Instead, I just focus on periods of GDP contraction and slow growth (<0.1%) that can be interrupted by brief periods of growth (double dip recession). </li></ul>
Time horizon for unemployment <ul><li>Here I focus on three periods: </li></ul><ul><li>The first one is the unemployment level in the last quarter before the recession. I call it PreR; </li></ul><ul><li>The second one is the last four quarters of the recession (R-3, R-2, R-1, R-0). If the recession lasts only 3 quarters I add an R-3 period unemployment rate to be equal to the R-2 period; </li></ul><ul><li>The third one is the following eight quarters of the recovery or expansion that I call E1, E2,…E8. </li></ul>
Unemployment patterns <ul><li>I notice the expansions can readily be grouped into three categories: </li></ul><ul><li>The ones where the unemployment drops in the very first quarter of the expansion; </li></ul><ul><li>The ones where the unemployment drops in the second quarter of the expansion; </li></ul><ul><li>The ones where the unemployment keeps on rising for more than two quarters. </li></ul>
Unemployment rate drops in first quarter of expansion The two lines are very much parallel. But, there is a large near constant difference between the two.
Unemployment rate drops in second quarter of expansion In the 1973+ period, the unemployment rate remained much higher throughout the next two years of the expansion than during the last four quarters of the recession in the 1953+ and 1956+ periods.
Unemployment rate keeps on rising for a while… See how abruptly unemployment has risen in the most recent 2008 recession. And, it has kept on rising during the first two quarters of the current expansion.
GDP (quarterly annualized rates) The yellow line denotes the first quarter of the expansion (E1). The green cells indicate intermittent GDP growth embedded within recessionary periods. The red cells indicate when GDP growth briefly fell below 0% during the expansion period. The three indicators at the bottom indicate the recession length in # of quarters, the average change in GDP during the recessionary period and during the expansion.
Regression Model to forecast the current expansion The dependent variable is the average GDP growth rate during the eight quarters of the expansion. We ran a stepwise regression using many variables. The optimal regression model used just three independent variables and two dummy variables. They are: X1 the recession length in quarters, X2 GDP in first quarter of expansion (GDP E1), X3 average unemployment in first two quarters of expansion (Avg. U E1E2). The dummy variables are X4 recessions that occurred after 1970 and X5 recessions that occurred before 1950. The model forecasts an average GDP growth of 5.1% for the eight quarters of the current expansion. This is high for a post 1970 expansion.
Regression Statistics & Parameters See how the average unemployment in the first two quarters of the expansion (Avg. U E1E2) has a strong positive regression coefficient. That is probably a combination of the V shape recovery and the unemployment lag phenomena. This model fit was excellent with an R Square close to 1 and a standard error of 0.2%. We could see that on the previous slide.
Model’s limitations <ul><li>It uses only 9 historical data points (9 recessions after WWII and before the current one). Some of the data does not go back further in time (BLS and BEA series extracted through Moody’s Economy.com); </li></ul><ul><li>Given its excessively small sample size, the model does not facilitate the use of a hold out period. We reran the model using a hold out period of one single recession period. And, it did not have a material effect. But, XLStat was unable to calculate many of the model’s goodness of fit statistics; </li></ul><ul><li>Predicting avg. GDP growth over next two years is difficult. The economic profession is so far not a reliable forecaster. This simple regression will most probably be no better or inferior to traditional economic forecasting. </li></ul><ul><li>See the last slide (19) for a narrative suggesting why the recovery may be much weaker than the model forecast. </li></ul>
Appendix: Data on the relevant recession and expansion periods with analytical comments reflecting the economic performance of the time
1948 an Outlier with a spectacular recovery Green zone: unemployment rate in the last quarter before the recession started. Blue zone: unemployment rate during the recession. Yellow zone: unemployment rate during the first eight quarters of the recovery or expansion. Red zone: recession period showing quarterly GDP on annualized basis. Note the recession period includes quarters of weak growth (not just negative) and it includes quarters of strong growth intertwined between quarters of negative growth (double dip situation). We are showing change in annual CPI as a matter of curiosity as we did not use this variable in our models. CPI measures had very low correlation with GDP growth.
1953 & 1956 a true double dip… The 1953+ ensuing expansion lasted only six quarters before it dovetail into the 1956 recession.
1960 & 1969 early jobless recoveries In both cases, unemployment remained moderate during the recession and rather high during the expansion vs what it was during the recession.
1973 & 1980 Stagflation periods Same observation regarding unemployment as on the previous slide. Those two periods experienced stagflation related to two major oil shocks when OPEC rose oil prices by a high multiple. See the very high inflation rates during the recessions.
1990 & 2000 The U shaped recoveries The recessions were short and mild. But, the ensuing recoveries were also mild meaning weak.
2008 will it be a V, U, or L recovery? The data reflects what was available at the time of this analysis. The recession was certainly severe enough to justify a V shape recovery with strong growth. And, the regression model suggests that too. However, if unemployment remains high and if the CPI remains flat for a few more quarters we may have an L situation (severe recession with a weak recovery).