A Note on Pre-2008 Unemployment-Rate Mean-Reversion: Wednesday Focus for October 8, 2014

Back before 2008, we neoclassical new Keynesian-new monetarist types were highly confident that the U.S. macroeconomy as then constituted had very powerful stabilizing forces built into it: if the unemployment rate rose above the so-called natural rate of unemployment, the NAIRU, it would within a very few years return to normal.

This is why we were confident:

1948-2014: Share of Deviation of Unemployment from Trend Erased After…

...1 Yr    ...2 Yrs    ...3 Yrs    With NAIRU trend...

33.7%      67.4%       88.3%       Cubic
32.4%      63.6%       84.1%       Quadratic
31.8%      61.5%       80.5%       Linear
27.8%      52.5%       69.2%       No Trend

Standard errors? Bootstrapping the three-year forecast cubic-trend model with 10,000 replications produced a mean estimated coefficient of 84.3% with a standard error of the estimate of 17.4%-points–half again as large as the OLS estimated standard error.

We were wrong.

If you have R, and wish to play with the Rmarkdown files, take a look:

• RMarkdown File: https://www.dropbox.com/s/egv37ytaz4lyc2h/2014-09-29%20Unemployment%20Rate%20Mean%20Reversion.Rmd?dl=0
• Output File: https://www.dropbox.com/s/q11gbj64u2f31m9/2014-09-29_Unemployment_Rate_Mean_Reversion.html?dl=0
• RMarkdown File: https://www.dropbox.com/s/uffbtork0dre3ku/2014-09-29_Unemployment_Rate_Mean_Reversion_Simple_Bootstrap.Rmd?dl=0
• Output File: https://www.dropbox.com/s/ztw7kjve1ia98gt/2014-09-29_Unemployment_Rate_Mean_Reversion_Simple_Bootstrap.html?dl=0