Adjusting the Taylor Rule for the Unemployment Rate Bias

November 3rd, 2013 at 12:17 pm

In a post the other day about the Federal Reserve and inflation, I noted that a commonly used rule to gauge where the Fed should set the Federal Funds Rate (FFR) was pointing at around -2% right now.  However, I noted that:

…if you tweak that calculation in a way that I think you should, i.e., by increasing the measured unemployment rate to reflect its downward bias due to people leaving the labor force because of weak labor demand, then you find that the FFR should be more like -4%.  No wonder we’re stuck in such a slog.  (Later, I’ll post something showing these calculations.)

It’s now later.

There are various different rules and techniques for estimating the optimal FFR, but John Taylor’s rule is probably the most common.* There are different versions of this rule and the one I favor is partly driven by the gap between the full employment unemployment rate and the actual rate.  The larger that gap, the lower the FFR should be.

But one problem with applying that formula right now is that, as noted, the unemployment rate is biased down which in turn biases up the FFR that pops out of the Taylor rule.  Economists have ginned up many estimates of that bias, but the great labor economist Heidi Shierholz of EPI does so each month and handily posts the data.  (Her estimate uses the counterfactual derived by the BLS: their participation rate consistent with full-employment.)

Here’s Heidi’s graph of actual vs. simulated unemployment, the latter being the jobless rate if all the “missing workers” were unsuccessfully seeking work.  The difference between the two rates is three percentage points, or about 4.5 million jobs given today’s (depressed) labor force.  As Heidi herself has noted, this is an upper bound on the jobless rate since presumably some of the missing workers would find jobs were they looking and others would still be out of the job market.


Source: Heidi Shierholz, EPI.

So in plugging this adjusted rate into a Taylor rule, I arbitrarily cut the difference in half, which actually gets you closer to other estimates of the current downward bias in the jobless rate, which range from 1-2 percentage points.**  Plotting both Taylor-rule-generated FFRs in the figure below, you see that the actual rate returns an FFR of about -2%, while adjusting for the downward bias in the jobless rate return an FFR of about -4%.


Source: My calculations, see text.

All’s I’m sayin’ here is that we still have a lot of slack in the job market, highly elevated unemployment (any way you want to measure it), and a strong rationale for aggressive monetary stimulus, not to mention fiscal stimulus, but that is blocked by government dysfunction.  The message here is thus: good for the politically independent Fed for keeping the monetary pedal to the metal.  That’s certainly the direction in which the data are pointing.


*Janet Yellen’s recent “optimal control” work is also gaining prominence in this space.  By simulating a path toward a goal like full employment, optimal control techniques return both a path for the FFR and inflation (other variables could easily be included but these are the key ones).

See the figures on pages 28-30 from this Yellen paper.  Relative to various Taylor rules, the technique allows inflation to temporarily exceed the target rate (2%) as the unemployment rate falls (I should note that optimal control assumes inflation expectations remain well-anchored, an assumption that will produce agita for some price hawks).

**I used this version: 2+p+(0.5(p-2))+y where p is year-over-year percent change in the PCE inflation index and y is the output gap: 2*(5-unemp) where 2 is the Okun coefficient and 5 is my NAIRU (the first “2” in the formula is the neutral FFR, though that too may be biased up these days).  I used a similar Taylor formula here, though no less than Taylor himself was not loving that earlier post.

Print Friendly, PDF & Email

8 comments in reply to "Adjusting the Taylor Rule for the Unemployment Rate Bias"

  1. Fred Donaldson says:

    Part of the lack of respect for government comes from the use of statistics and the sorting of facts to promote the agenda of one group or another. It extends from FOX saying most employers will stop their health insurance plans because of Obama to an unemployment statistic that doesn’t include all the unemployed, only the ones who are not totally desperate and have given up all hope of getting a job.

    Sometimes, watching politicians and their loyal pundits explain this “stopped looking for employment” situation, I wonder if they think these job-seeking weary folks are now living off their trust funds, their savings, or getting enough ($30 a week) food stamps to survive in modest comfort. That attitude is Americans can’t really be poor with our safety net. How untrue. How sad. How cruel.

  2. urban legend says:

    It’s virtually impossible to exaggerate how weak the U.S. labor market is. Look at the employment rates in OECD countries for the prime working age, 25-54: all but one other country had a higher employment-to-population ratio in 2012 than it had in 2000; the U.S. rate dropped 5.8 percentage points. The U.S, had one of the highest rates in 2000 at 81.5%. Now, while the rates in other countries are mostly well north of 80% or at worst very close, the U.S. rate is an abysmal 75.7%.

    Any talk of an improving market sufficient to even hint at making it appropriate to cut back on stimulating the economy with whatever tools are available is ridiculous. So is the idea that that there is a “NAIRU” even remotely close to the unemployment rate now. Saying the unemployment rate “vastly understates” the weakness of the labor market is itself an understatement of how terrible it is. “A lot of slack” is even further off the mark.

    Remember, these are prime working age adults. If you can’t employ people in that age group, you don’t even qualify as an advanced country.

  3. Bill Gatliff says:

    Mr. Bernstein:

    Could you PLEASE comment on where all the workers go when they “give up” on their searches for employment?

    I could see one member of a two-income household taking a year or two off to go back to school, and so on. But I just don’t see how decisions like that could absorb the large numbers at issue here, nor how they could go unnoticed in the economy as a whole given that pre-recession the numbers all suggested that those same households were just barely making ends meet as it was. They’d be completely upside-down now, to say the least.

    Can your research tell us where they are going? My pet theory is that they are turning to cash-only employment e.g. casual housekeeping, maintenance, and so on. But again, I can’t believe that the economy could absorb so many workers that way without it going noticed elsewhere…



    • Jared Bernstein says:

      Good question…you really need longitudinal data to answer this question, meaning data that follows people over time and we don’t have that yet on these folks. But I’ll try to look into it.

  4. William Dyess says:

    Very interesting, I noticed you used 5% as the NAIRU. The Fed may be using a different rate however.

    “We find that if the currently estimated shift in the BC is permanent and the economy returns to its long-run JCC, then the long-run natural rate of unemployment has increased from its 5% level in 2007 to somewhere between 5.6% and 6.9% as of November 2012.” – slide 17 (

    “We regard 6% as a plausible estimate of the current long-run natural rate of unemployment.” -slide 17 (

    • Jared Bernstein says:

      Yeah, well…they’re lucky I didn’t use 4%, which I can and will justify in later posts.

      • William Dyess says:

        Great, I look forward to reading it sir!

      • urban legend says:

        Given the terrible employer-to-population ratio, especially in the prime working age group (25-54), how can we imagine the black letter unemployment rate is the appropriate measure for the kind of analysis the NAIRU was designed for? Do we seriously think a 5% unemployment rate with an overall employee-to-population ratio of, say, 61 or 62%, and a 25-54 rate around a still-weak 78%, with continuing high levels of discouraged and involuntary part-time workers, will identify a situation that remotely can be called full employment? That at that point the labor market will be poised to push us towards persistent, wage-driven inflation?