rough draft:

comments welcome.

do not quote, unless


Productivity Growth and the Recent "Fall in the NAIRU"


James G. Devine

Loyola Marymount University.

May 10, 2001

In an article titled "Do We Have a 'New' Macroeconomy" (, U.C.-Berkeley economist J. Bradford deLong runs into a contradiction that I've seen in the business press a lot recently. This is to assert that the recent (alleged) acceleration of labor productivity growth has caused the NAIRU to fall. However, this assertion contradicts the received (neoclassical) economic theory.

DeLong seems quite uneasy with the lack of acceptance of his assertion by orthodox economists, so it could be said that he is conscious of the contradiction. However, it’s useful to tease out the nature of the problem, i.e., the conflict between the casual empiricism of deLong (or Alan Greenspan) and the official neoclassical party line on this subject. If we address the contradiction, we have to acknowledge that the perceived improvements in the NAIRU are intimately linked with the shift in the distribution of income away from labor and toward capital.

In the below, I want to simply focus on the contradiction. I'll take the acceleration of labor productivity growth for granted (and won't quibble about its causes, e.g., to note that it may be partly because people are working more intensively). More importantly, I'll also initially take as fact the existence of a unique NAIRU (U*) that's independent of the actual unemployment rate; thus I ignore a major issue that deLong avoids until Part II. In tune with the reigning orthodoxy, I also initially treat U* as given by structural conditions in labor markets, i.e., the degree of structural or mismatch unemployment, the amount of frictional or turnover unemployment, and the amount of impediments to labor-market adjustment due to rigidity of real wages (classical unemployment). I also present a clear model, rather than leaving things implicit or referring to the assumed consensus of orthodox economists, the way deLong does. (For people who don't like math, skip to part II.)

The first section deals with the contradiction between deLong’s assertion and the dominant theory, while the second section considers what this all means in terms of policy. Though the first might be interpreted as a criticism of deLong, the second praises his assertion.

part I: Productivity Growth and the NAIRU.

In order to understand the contradiction, first state a short-run "Phillips curve" for wages:

(1) w = -b(U - U*) + w* + pex

All lower-case variables refer to percentage rates of change: w is the rate of wage inflation, pex is the expected rate of inflation, and w* is the trend rate of growth of nominal wages. The parameter b is a positive constant, while U is the (official) unemployment rate and U* is the notional NAIRU. Note that I call the last variable "notional." As seen below, the effective NAIRU – i.e., the NAIRU relevant to economic policy-makers – may differ from U*.

Equation (1) as three parts. The first represents the cyclical or conjunctural component of the bargaining power of labor: as unemployment falls, workers are more able to raise money wages, all else constant. The second part (w*) represents one part of the structural-institutional or organizational part of workers' bargaining power. (I'll complete the definition below.) If workers have stronger organizational bargaining power vis-à-vis their employers (due to increased unionization and the like), w* is higher. Finally, pex represents the efforts of workers to defend real wages from expected inflation (since they only have say over nominal wages in most cases). Note that I assume that a rise in pex is directly translated into actual wage inflation (as its coefficient equals unity). This is quite an orthodox assumption (and I disagree with it), but I make it for argument's sake.

Orthodox economics discusses the components of (1) totally in terms of markets (supply and demand), ignoring institutions. I totally disagree. Instead of presenting an argument against that interpretation at this point, I will simply note that the argument developed below indicates that a version of the Phillips curve that incorporates institutional insights helps avoid the contradiction between the productivity growth/NAIRU nexus that deLong posits and the inability of orthodox economics to accept that nexus.

Next, given money wage growth, employers pass on the increased costs to consumers:

(2) p = w - q* + ss

I assume that businesses can "mark up" unit labor costs, so that if money wage growth (w) is higher than the trend growth of labor productivity (q*), prices will rise following w - q*. This keeps the mark-up on unit labor costs constant. Though this assumption seems less valid nowadays than in the 1960s (and I reject it below), I'll make it initially in order to draw out the key contradiction. Finally, ss refers to supply shocks. For simplicity, I'll assume that ss = 0, though a lot of the recent low-inflation experience in the U.S. has been due to beneficial supply shocks (falling real oil prices until 1999 or so, a highly-valued dollar). Further, I ignore the possible direct effect of inflationary expectations (pex) on actual inflation.

Combining equations (1) and (2) give us a general short-run price-inflation Phillips curve:

(3) p = -b(U - U*) + w* - q* + pex

According to the NAIRU theory, in the long run, we're not dead but in equilibrium. In this state, p = pex (people expect the inflation that occurs and so show no inclination to change their expectations). In equilibrium, the unemployment rate equals Ue:

(4) Ue = U* + (1/b)(w* - q*)

This must be the kind of equation that deLong must be thinking about when he asserts that rising labor productivity growth (rising q*) leads to allows the economy to attain lower unemployment rates without suffering from accelerating inflation. If w* stays constant while q* rises, then Ue falls. Ue represents the threshold below which inflation tends to accelerate. It is the effective NAIRU, given the institutional situation. It is akin to what economists at the O.E.C.D. [2000] call the "short-term NAIRU."

More generally, what equation (4) says is that labor-market institutions matter in determining the effective NAIRU (and by "institutions" I mean more than transgressions against market perfection). In other words, we should reject the neoclassical conception that a world without non-market institutions is a good approximation for the real world, as Blanchard and Katz [1997] do in practice.

I would fully define "the structural-institutional or organizational part of workers' bargaining power" as workers' trend ability to win nominal wage increases in step with the rise in their productivity (w* - q*). If this bargaining power declines – as in the last quarter-century or so – then (w* - q*) also falls. Thus, Ue falls, as deLong suggests.

Further, b represents how exposed normal wage bargains are to market forces. If b is low, there is not much of an impact of unemployment on worker's cyclical ability to raise money wages. (The short-run Phillips curve (3) is flat in a graph with p and U on the axes.) On the other hand, if b is high, i.e., if workers' can't insulate their lives from the vicissitudes of the market, then the short-run Phillips curve is steep. Crucially, if b high, the importance of the second term in equation (4) shrinks, so that (all else constant) Ue approaches U*.

I would guess that the apparent fall in the NAIRU in the 1990s was largely due to (1) the falling ability of workers to raise money wages in step with labor productivity (falling w* - q*). At the same time, there has been (2) an increased exposure of workers to the fickle finger of the Invisible Hand (rising b). Both of these are suggested by Alan Greenspan’s musings that increased worker insecurity might be behind the fact that any amount of unemployment seemed to have a greater ability to deter inflation: both of these cause Ue to fall. We could also point to the way in which changes in workers’ institutional situation also affects the attainment of the assumed equilibrium (by changing the process of adjustment of pex), but that would lead us off the point.

Where's the contradiction, though? The problem is the introduction of labor productivity growth into the story goes against standard neoclassical presumptions. I have seen a lot of orthodox versions of (3) in which both w* and q* are dropped, so that in the absence of supply shocks:

(3N) p = -b(U - U*) + pex

That means that the trend growth of money wages must be equal that of labor productivity, so that unit labor costs (wage costs/labor productivity) and the organizational part of workers' bargaining power stay constant. Most importantly, this equation also indicates that labor productivity growth plays no role at all, going against deLong’s assertion. This is quite clear when we note that the long-run equilibrium unemployment rate equals:

(4N) Ue = U*

So the effective NAIRU equals its notional value. Crucially, we can’t use recent productivity growth as an explanation of the falling NAIRU: the U* is assumed to be determined by such issues as demographics, which deLong rejects as an explanation of the falling NAIRU in the 1990s.

On the other hand, if w* - q* is allowed to fall (i.e. if workers’ institutional bargaining power is weak), we can use labor productivity as an explanation of the low NAIRU in the 1990s. But then we have to address the distributional issues, something that neoclassicals would like to leave to references to technologically-determined parameters (such as the marginal product of labor ground out by making the silly assumption of the existence of an aggregate production function).

The fact is that the institutional changes behind the fall in w* - q* (the decline in worker’s institutional bargaining power, as seen in declines in unionization and the like) also lead to a stagnation of real wages in the 1990s – until the temporary demand-side boom of the late 1990s (i.e., 4 percent unemployment) implied greater cyclical bargaining power for workers. That is, though workers benefit from low unemployment as Ue falls, they pay for it in terms of their ability to raise wages in step with productivity and the trend shift in the distribution of income toward capital.

The rise in the coefficient b (the greater exposure of workers to market forces) implied a steep rise of wages during the late 1990s, but it suggests that wages will fall behind inflation just as steeply in the near future (perhaps in 2000). This raises the possibilities of deflation, something that can be quite disastrous in a country like the U.S., where much of the population is deep in debt. At one point in his ponderings, Paul Krugman rejected the possibility of deflationary depression because wages are sticky downward. The institutional changes in labor markets makes such optimism less apt.

The distributional issues are crucial and encourage us to break with the assumption made above that firms aim to keep a constant mark-up. If they are successful, this assumption implies that the share of profits in national income is constant, something that has been shown to be incorrect in recent years. Instead, it might be argued that businesses aim for a higher mark-up when their profitability is low and are willing to accept a lower mark-up when profitability is high, while they need not always achieve their goals.

Under this alternative conception, when worker organizational bargaining power is low (as in recent decades), real wages lag behind productivity growth, so that profitability rises in the trend. This reduces the need to raise mark-ups and prices, reducing the inflationary potential of the economy. Alternatively, with workers having less organizational bargaining power, less unemployment (a smaller "reserve army of labor") is needed to protect profits. In general, this suggests a negative relationship between profit rates and stagflation (as measured by the sum of unemployment and inflation rates).

This relationship is seen in the graph below, copied from another paper of mine. The profit rate is corrected for changes in capacity utilization, i.e., cyclically corrected, since falling utilization rates discourage inflation. The sum of unemployment and inflation rates is termed "the stagflation potential factor" or SPF. The consumer price index for urban consumers is used to measure inflation. The relationship stands up pretty well when different measures of profitability and inflation are used. In sum, there is a strong relationship between distributional issues and the seeming fall in the NAIRU during the 1990s. 

Part II: the NAIRU, Policy, and Reality.

Though I think that there’s a little bit of truth in the NAIRU theory (i.e., if unemployment equaled 2 percent or lower for a few years, inflation would begin to get worse and worse), the whole theory has severe problems. As seen above, even the presumption of that a unique NAIRU that exists independent of the actual unemployment rate cannot avoid the political issues of distribution. But there is more.

The key issue can be stated as follows. In the late 1990s, was it that the actual notional NAIRU was falling or was it rather that there was a rising willingness to let the actual unemployment rate to fall, which allowed the realization of a lower NAIRU?

The problem is that there are major errors in the estimation of the NAIRU (Staiger, Stock, and Watson, 1997). The O.E.C.D. (2000) tells us that the U.S. NAIRU equaled 5.2 percent in 1999, as do the Office of Management and Budget and the Congressional Budget Office. Unlike most authors, the O.E.C.D. economists state this number plus or minus a margin of error (a standard error) of 1.2 percentage points. That means that the NAIRU could be anywhere between 4 and 6.4 percent of the labor force. This in turn implies that we may see a "fat" NAIRU, as suggested by authors such as Abba Lerner [1951] and Robert Eisner [1997]: the NAIRU (presuming that it exists) represents a range, from "high full employment" to "low full employment" (in Lerner’s terms). Given the quality of the empirical results, we really can’t assert anything else except by imposing an arbitrary assumption.

In this view, the Federal Reserve isn’t simply intuiting the structure of the economy to decide what specific unemployment rate (the NAIRU) should prevail in order to stabilize the inflation rate. Rather, it is choosing among a number of possible NAIRUs (assuming that they really have enough control over demand to do so). Since the desired or target unemployment rate isn’t given by the parameters of the economy, the Fed’s leadership’s political biases can easily creep in.

The Federal Reserve, like most Central Banks, is dominated by the collective interests of bankers and financiers. It pushes for and cultivates its independence from democratically-elected politicians (while the latter are often willing to let it go independent, since they can avoid responsibility when things go wrong). But it is dominated by inflation hawks, whose always emphasize the "worst case" scenarios while emphasizing the false precision of their estimates. (For those who like historical analogies, their attitudes are akin to that of the Committee on the Present Danger and "Team B" toward the late Soviet Union.)

This encourages the Fed to define the NAIRU as on the high end of the possible range while actively fighting inflation even when it isn’t there. (Central Bankers proudly proclaim that their job is to "take away the punch-bowl before the party gets wild," ignoring the impact of their policies on Main Street. Some people benefit only when the party is so wild that the party-goers become generous.) That’s why, for example, the Fed kept on raising interest rates during 1999 and most of 2000, even though inflation wasn’t to be seen. At least for awhile, workers could thank their lucky stars that it didn’t have enough power over the economy to squelch the boom right away, since this is when their cyclical bargaining power rose, counteracting the decrease in their institutional power.

That is, it is quite extraordinary and unexpected that the Fed allowed the unemployment rate to fall to 4 percent of the labor force. What had happened was that events in 1997 and 1998 – the East Asian crisis, the Russian crisis, the collapse of such U.S. speculative organizations as Long-Term Capital Management, etc. – pushed the Fed lower interest rates. Greenspan commented that the Fed was steering on a knife-edge between worsening inflation and deflation – and was pushed to risk the former. This meant that the economy was pushed into a region of low unemployment rates as the Fed dropped the "worst case" inflation scenarios for awhile (until 1999).

Then, with lower unemployment, we see the realization of what seemed to be a lower NAIRU. Not only was there no serious inflation, but low unemployment tends to feed on itself, encouraging the NAIRU to fall (the hysteresis effect). (See Hargereaves-Heap, 1980.) In the late 1990s, Business Week reported that many employers were hiring people they normally wouldn’t hire (because of inadequate skills, wrong location, or poor attitudes) and then training them for the jobs available. This meant that the problem of structural or mismatch unemployment was being attacked. If there’s any truth to the NAIRU theory, then lowering structural unemployment should lower the NAIRU by lowering U* (the notional NAIRU).

Further, abundant demand encourages investment, learning-by-doing in production, the transfer of labor to high-productivity sectors, technological change, greater specialization, and thus labor productivity growth (Verdoorn’s Law). (It should be stressed that this complements rather than replacing the "new economy" story that deLong discusses.) Given the general stagnation of worker bargaining power, the surge in labor productivity growth encouraged the fall in the effective NAIRU.

Finally, note that deLong’s view that a surge of labor productivity growth lowers the NAIRU can help contribute to this process. If he can convince the Fed and other members of the policy-making elite that the "new economy" allows a lower NAIRU, it might allow the unleashing of hysteresis forces and Verdoorn’s Law, producing a virtuous circle. But it doesn’t fit well with neoclassical theory, which should be rejected. Nor does it deal with the distributional impact of the institutional changes that have allowed the effective NAIRU to fall.



Blanchard, Olivier and Lawrence Katz. 1997. What We Know and Do Not Know about the Natural Rate of Unemployment. Journal of Economic Perspectives. 11(1) Winter: 51-72.

deLong, J. Bradford. 2001. "Do We Have a 'New' Macroeconomy" (

Devine, James G. 2000. "Rising Profits and Falling Inflation: An Empirical Study," Review of Radical Political Economics, 32(3), 2000: 398-407.

Eisner, Robert. 1997. A New View of the NAIRU. In Paul Davidson and Jan A. Kregel, eds. Improvinq the Global Economy. Cheltenham, UK: Edgar Elgar, 1997.

Hargreaves-Heap, Shawn P. 1980. Choosing the Wrong 'Natural' Rate: Accelerating Inflation or Decelerating Employment and Growth? Economic Journal, Sept.

Lerner, Abba. 1951. Economics of Employment, New York: McGraw-Hill.

O.E.C.D. 2000. "Revised O.E.C.D. Measures of Structural Unemployment," December 2000 O.E.C.D. Economic Outlook , December, vol. 2000, no. 2.

Staiger, Douglas, James H. Stock, and Mark W. Watson. 1997. The NAIRU, Unemployment and Monetary Policy. Journal of Economic Perspectives. 11(1) Winter: 33-49.

James G. Devine

Professor of Economics

University Hall (Rm. 4227)

Loyola Marymount University

One LMU Drive, Suite 4200

Los Angeles, CA 90045-2659 USA

office phone: 310/338-2948; FAX: 310/338-1950