Back from the Dead: the GFC and the Resurrection of Long Term Unemployment Bruce Chapman* Crawford School of Economics and Government School Seminar Australian National University September 22 2009 * With many thanks to Kiatanantha Lounkaew
OUTLINE 1 Some comments on the dynamic nature of the labour market. 2 Where does LTU come from? 3 Why worry about LTU? 4 Projections of LTU: method 5 Using unemployment counter-factuals (2010/2011) 6 Big points: (i) LTU must increase now from the GFC by over 40,000 (2010) and 75,000 (2011) no matter what; and (ii) with bad luck/bad management the increase would be to 139,000 (2010) and 249,000 (2011) 7 What should be done now?
1 The dynamic nature of the labour market It s all about flows The duration dimension is critical for policy
2 Where does LTU come from: 1980-2009 LTU (thousands) Unemployment rate 400 LTU stock 28% 350 Unemployment rate 300 23% 250 18% 200 150 13% 100 8% 50 0 3% 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Year
3 Why worry about LTU? Labour market disadvantage Low education, Indigenous Older males By definition not accumulating labour market experience Macroeconomic efficiency issues The UV Curve The NAIRU
3 Why worry: the UV curve and the Working Nation debate II
4 Projections of LTU: method Estimate regression models explaining LTU using unemployment rate data Choose models on the basis of statistical stability Use different future possible unemployment rates to predict future LTU
LTU Econometric Modelling Conventional OLS regression Dependent variable: % change in LTU Independent variables: % change in unemployment; % change in unemployment rates Flexible functional forms (for projections) Present and Lags (0, 1 and 2) Interactions
The Simple Model Model: OLS, using observations 1983-2009 (T = 27) Dependent variable: % Change in LTU Coefficient Std. Error t-ratio p-value const -0.0135266 0.00985061-1.3732 0.18353 x t 0.271396 0.0641782 4.2288 0.00035 *** x t-1 1.24524 0.0960187 12.9687 <0.00001 *** x t-2 0.402057 0.064857 6.1991 <0.00001 *** x t-1 _x t-2 1.50102 0.468834 3.2016 0.00412 *** Mean dependent var 0.031088 S.D. dependent var 0.271628 Sum squared resid 0.052036 S.E. of regression 0.048634 R-squared 0.972874 Adjusted R-squared 0.967942 F(4, 22) 197.2595 P-value(F) 6.85e-17 Log-likelihood 46.08599 Akaike criterion -82.17197 Schwarz criterion -75.69279 Hannan-Quinn -80.24537 rho -0.258456 Durbin-Watson 2.427541
The (More) Flexible Model Model : OLS, using observations 1983-2009 (T = 27) Dependent variable: % Change in LTU Coefficient Std. Error t-ratio p-value const -0.0238118 0.0368991-0.6453 0.53678 x t 0.0697512 0.219854 0.3173 0.75917 x t-1 0.953134 0.177994 5.3549 0.00068 *** x t-2 0.384377 0.194073 1.9806 0.08297 * x t _sq 1.03653 0.647544 1.6007 0.14811 x t-1 _sq 0.631678 0.416364 1.5171 0.16771 x t-2 _sq -0.046267 0.506899-0.0913 0.92952 x t _x t-1-0.526204 0.95556-0.5507 0.59689 x t _x t-2 0.76426 1.7792 0.4296 0.67886 x t-1 _x t-2 1.55661 0.606367 2.5671 0.03328 ** inv_ x t -0.000137312 0.000663506-0.2069 0.84122 inv_ x t-1 0.00052079 0.000808058 0.6445 0.53729 inv_ x t-2 0.00045886 0.000594689 0.7716 0.46253 inv_ x t _x t-1 1.67994e-05 1.48764e-05 1.1293 0.29151 inv_ x t _x t-2-8.1084e-06 2.42199e-05-0.3348 0.74640 inv_ x t-1 _x t-2-2.30149e-05 1.62721e-05-1.4144 0.19497 inv_ x t _sq 5.94008e-07 1.23854e-05 0.0480 0.96292 inv_ x t-1 _sq -1.4678e-05 1.39688e-05-1.0508 0.32407 inv_ x t-2 _sq -8.04198e-06 1.28788e-05-0.6244 0.54973 Mean dependent var 0.031088 S.D. dependent var 0.271628 Sum squared resid 0.015010 S.E. of regression 0.043315 R-squared 0.992176 Adjusted R-squared 0.974571 F(18, 8) 56.35868 P-value(F) 1.76e-06 Log-likelihood 62.86985 Akaike criterion -87.73969 Schwarz criterion -63.11879 Hannan-Quinn -80.41861 rho -0.201321 Durbin-Watson 2.385303
LTU Level Predictions (within sample) Number of LTU (thousands) 400 350 Actual Predicted 300 250 200 150 100 50 0 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Year
5 Using unemployment counter- Four scenarios: factuals (2010/2011) (i) (ii 2008 and continuing unemployment rate of 4.3 % (no crisis) Unemployment rate of 8.5 % by 2011(Treasury forecast) (iii) Unemployment rate of 9.5 % by 2011 (very pessimistic) (iv) Unemployment rate of 6.5 % by 2011 (somewhat optimistic)
Unemployment rate 12% LTU Forecasting Scenarios No crisis=4.3% Optimistic=6.5% 10% Treasury = 8.5% Very pessimistic=9.5% 7.7% 8% 9.5% 8.5% 6% 5.8% 7.2% 6.2% 6.5% 4% 4.3% 4.3% 4.3% 4.3% 2% 0% 2008 2009 2010 2011 Year
6 LTU Projections 2010-2011 Number of LTU (thousands) 275 250 225 249 220 200 175 150 169 125 100 75 92 91.6 92.0 50 25 0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Year
LTU Projections(2010/2011): Big points (i) LTU if business-as-usual, 2010-2011 = 91,600; 92,000 (ii) LTU with the GFC and good management/good luck = 133,000;169,000 (iii) Net good GFC effect (2010/2011) = (ii) (i) = 41,400; 75,000 (iv) Implicit Treasury Predicted LTU = 136,000; 220,000 (v) Implicit Treasury Projected Increase (2010/2011) = (iv) (i) = 44,400;128,000 (vi) LTU with the GFC and bad management/bad luck = 139,000; 249,000. (vii) Net bad GFC effect (2010/2011) = (vi) (i) = 47,400; 157,000
7 What should be done now? Estimation with a more sophisticated model is important: C, J and K quarterly data, robust estimates for 1990s The experience of, and wash-up from, Working Nation Evaluation data unclear; now in a better position; convening experts a solid idea The role of targeted employment programs generally; targeting on the basis of expected unemployment duration; - Piggott and Chapman issues: deadweight, substitution and effectiveness - likely outcomes not better than 1/3 of those targeted. Resumption of economic growth is critical
Simple Model (RHS: % Change of UR) Model : OLS, using observations 1982-2009 (T = 28) Dependent variable: % Change of LTU Coefficient Std. Error t-ratio p-value const 0.0210264 0.00977921 2.1501 0.04229 ** x t 0.274113 0.0618941 4.4287 0.00019 *** x t-1 1.26769 0.0867776 14.6085 <0.00001 *** x t-2 0.39719 0.0653948 6.0737 <0.00001 *** x t-1 _x t-2 1.49963 0.464753 3.2267 0.00373 *** Mean dependent var 0.030413 S.D. dependent var 0.266575 Sum squared resid 0.054977 S.E. of regression 0.048891 R-squared 0.971346 Adjusted R-squared 0.966363 F(4, 23) 194.9222 P-value(F) 2.20e-17 Log-likelihood 47.53231 Akaike criterion -85.06463 Schwarz criterion -78.40360 Hannan-Quinn -83.02828 rho -0.157920 Durbin-Watson 2.256701
Simple Model Projection (RHS: % Change of UR) Long-term unemployment (thousands) 275 250 225 258 230 200 175 176 150 125 100 75 92 94 96 50 25 0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Year
Projections Model Year Unemployment Rate 4.3% 6.5% 8.5% 9.5% Simple (% U) Flexible (% U) Simple (% UR) 2010 95.8 137.7 141.5 143.3 2011 99.9 178.9 229.5 253.3 2010 91.6 132.5 135.9 138.9 2011 92.0 168.9 220.0 249.3 2010 94.1 136.5 141.0 143.2 2011 96.1 175.5 230.0 257.8