The Unemployment Rate at Full Employment: How Low Can You Go?
Jared Bernstein and Dean Baker
A lot of people are talking about full employment these days. Lawrence Summers bemoaned its long-term absence at an important talk at the International Monetary Fund recently. Janet Yellen, nominated by President Obama to head the Federal Reserve, stressed that institution’s role in meeting the goal of full employment during her Senate hearing last week. Our own recent work documents the widespread benefits of full employment (and the costs of not being there) and prescribes a policy road map to get there.
All of which raises two questions. First, where is there? What is the unemployment rate consistent with full employment? And second, why does this matter now when the jobless rate is still highly elevated and only slowly coming down?
What follows is all about the first question, but let us briefly note why this matters so much, even — we’d say especially — now. To policy makers at the Federal Reserve and the many economists who watch and critique their every move, the unemployment rate associated with full employment is a key parameter. It’s a landing strip that determines their flight path in ways that have profound implications for the passengers on board, especially, as we stress below, those sitting in coach. A linchpin of our argument is that for much of recent history, this rate has been pegged too high, and the costs to working families have been steep.
For economists, the question we’re asking is this: What is the lowest unemployment rate consistent with stable inflation, otherwise known as the nonaccelerating inflation rate of unemployment, or Nairu?
We think we can do better. Our work suggests that 4 percent — the average unemployment rate for 2000, the last time we were at full employment — is a reasonable target, one worth shooting for.
As we’ll stress in a moment, despite all the point estimates out there, neither we nor anyone else knows the precise answer. But the costs of overestimating it — and economic history is replete with upwardly biased estimates — are a lot higher than those of underestimating it. Most important to us, those costs fall hardest on the working households that have faced wage and income stagnation for decades.
That said, let’s be clear. Compared with the current situation, we’d gladly welcome 5.5 percent unemployment, though frankly, not as much as the tens of millions of un- and underemployed people who depend on tight labor markets to give them a fair shot at claiming their fair share of the economy’s growth. But for the following four reasons, we do think the full-employment target should be lower than do most of our colleagues:
The Reliability of Estimates
Toward the latter 1990s there was a consensus within the economics profession that the Nairu was close to 6 percent. The 1994 Congressional Budget Office’s estimate for the end of the decade was 5.4 percent. Two years later it raised its estimate for the end of the decade to 5.8 percent. Most other estimates at this time were higher. In fact, this consensus led two Fed governors, Janet L. Yellen and Laurence H. Meyer, to argue with their chairman, Alan Greenspan, in 1996 that he should raise interest rates. With the unemployment rate falling below 5.5 percent by the spring, the contemporary estimates of the Nairu implied an acceleration of inflation. But based on his assessment that more rapid productivity growth would hold down inflation, Mr. Greenspan refused to go along with them, and the unemployment rate was allowed to fall to 5 percent by the middle of 1997, 4.5 percent by the middle of 1998 and 4 percent as a year-round average for 2000.
The Congressional Budget Office did eventually lower its estimate of the Nairu for the end of the decade, pushing its estimate for the year 2000 down to 5.2 percent, 5.0 percent in 2007, and 4.8 percent in 2008. (The table shows the C.B.O.’s estimates for the 2000 Nairu level given in the years preceding and following. The 1994 estimate is for 1999, not 2000, because of data availability.) The last estimate was a full percentage point below the 1996 estimate of the Nairu. These changes in the estimates of the rate over time from the same source using the same methodology demonstrate the enormous range of uncertainty around the measurement of the concept. A group of economists measured that uncertainty by calculating confidence intervals around a variety of Nairu estimates. Their 95 percent confidence interval included values of less than 4 percent and more than 8 percent.
In other words, no one should put too much faith in any particular estimate. While there are risks of higher inflation at low rates of unemployment, we simply cannot accurately pin down the level of unemployment that will lead to rising inflation.
The Experience With Low Unemployment in the 1990s Boom
As noted, there was considerable concern among economists, including those at the Fed, that the low unemployment rates in the late 1990s would lead to accelerating inflation. Yet as the unemployment rate fell below almost all the widely accepted measures of the Nairu, there is little evidence of any acceleration of inflation over this period.
The table below shows the unemployment rate from 1995 to 2001, along with various measures of inflation.
There was an uptick in the overall Consumer Price Index and the G.D.P. price index from 1999 to 2000, but this was probably more an energy story than a Nairu story. The price of oil fell from $20 a barrel in 1996 to $13 in 1998. This helped to dampen inflation in those years. It then began to rise in 1999 and 2000, peaking at more than $28 a barrel in 2000. The fact that the uptick in 2000 was due primarily to a reversal of the decline in oil prices can be seen by examining the core indexes. The core Consumer Price Index does show an uptick of 0.7 percentage points in 2000 from its 1999 level, but that placed it just 0.2 percentage points above the 1998 level, even though the unemployment rate had been below 4.5 percent for most of this three-year period. There was almost no further acceleration in the core C.P.I. into 2001, even though the unemployment rate remained below most estimates of Nairu for the whole year.
Interestingly, the inflation measure that least supports the Nairu story in these years is the Fed’s preferred measure: the core P.C.E. (personal consumption expenditures, excluding food and energy prices), which rose by just 0.2 percentage points in 2000 and even in 2001 was still lower than the 1997 rate. In other words, while one can find a bit of an acceleration in price growth over this period in certain series, that finding is far from robust. Unemployment was 4.5 percent or lower for three full years, from 1998 to 2000, yet the core rate of inflation was virtually the same at the end of this period as the beginning.
The Asymmetry Between the Gain from Lower Unemployment and Risks of Higher Inflation
So, collecting findings so far: There’s considerable uncertainty regarding the level of the Nairu, and estimates from the mid-1990s were demonstrably far too high.
In our view, this uncertainty provides a reason for shooting for a low target. Recent research from many sources, including the nonpartisan Congressional Budget Office, shows a distinct flattening of the Phillips Curve, which means that the cost of lower unemployment in terms of higher inflation is considerably less than was previously the case. In this research, the coefficient on the unemployment rate variable in a Phillips Curve is around 0.3, which means that if the unemployment rate is below the Nairu by a full percentage point for a full year, we would expect to see a 0.3 percentage point increase in the inflation rate.
Using Okun’s Law (a rule of thumb that maps reductions in the unemployment gap onto gross domestic product), the gains from being one percentage point below the Nairu for a full year would be two percentage points of higher output, or roughly $330 billion in the 2013 economy. According to our analysis, most of that additional output would flow to those at the middle and bottom of the income distribution. The impact on increasing employment, hours worked and hourly wages for the bottom 20 to 30 percent of the wage distribution would be especially large.
Against these benefits we face the risk that with the rate of unemployment pushed too low, the economy will now face a slightly higher rate of inflation. But the predicted rise is well short of hyperinflation. If the inflation rate had been 2 percent before we went below the Nairu, it would be 2.3 percent after the unemployment rate had been a percentage point too low for a year. That hardly seems like a major cause for concern.
If the inflation rate continued to rise because of a below-Nairu level of unemployment, the Fed would need to take steps to raise the unemployment rate before higher inflation actually became a problem. But the new Phillips Curve analysis suggests that there would be little urgency. The fact that inflation has grown less responsive to lower unemployment means the weighting of the risks associated with the unemployment-inflation trade-off has changed in favor of full employment. As Mr. Greenspan did in the 1990s, policy makers could aim for lower jobless rates — we’d recommend 4 percent. On behalf of those for whom economic growth has been a spectator sport for decades, we owe it to them to try.