The Paper That Made Economists Doubt the Long Run: Nelson and Plosser

Here is a question that sounds abstract and turns out to be one of the most consequential questions in economics.

When a recession hits, and output falls, does the economy eventually climb back to where it would have been?

Two answers, and the difference between them is the difference between two entire schools of macroeconomics.

Answer one: yes. The economy has a long-run growth path, set by technology and population and institutions. Recessions are temporary deviations below that path. Afterwards, the economy grows faster than usual and catches back up. The lost output is recovered. Recessions are painful but they are, in the long run, blips.

Answer two: no. There is no path to return to. A recession does not push the economy below its trend, it permanently relocates the trend downward. The output lost is lost forever. The economy simply resumes growing from a lower base.

In 1982, Charles Nelson and Charles Plosser took a hundred years of American macroeconomic data and asked which answer the numbers supported. The answer they got was answer two, and it detonated in the middle of the profession.

The problem: the two stories look almost identical

The frustrating thing is that both stories produce charts that look basically the same. Both show a series growing over time with wobbles. The difference is not visible to the eye. It is a difference about memory: does a shock today still matter in fifty years, or does it fade away?

Answer one says shocks fade. The series is "trend stationary": deviations from the trend are temporary and the trend is the anchor.

Answer two says shocks never fade. The series has a "unit root": each shock permanently shifts the level, and the series is a random walk with drift. There is no anchor.

Nelson and Plosser took this abstract distinction, applied the newly available Dickey-Fuller unit root tests, and ran them on fourteen long US macroeconomic series: real output, employment, prices, wages, interest rates, stock prices, and more, some of them stretching back into the nineteenth century.

The key idea via analogy: the injury that never heals

Imagine two people who each break a leg.

Person A goes to a good hospital, heals over six months, does physiotherapy, and eventually walks exactly as they did before. The injury was real and painful, and in the long run it left no trace on their walking speed. Their trajectory is unchanged. This is answer one.

Person B breaks their leg, it heals badly, and they walk with a permanent limp for the rest of their life. Their walking speed after recovery is permanently lower than it would have been. They can still improve over time, they can still get fitter, but they are forever walking on a path that sits below the one they were on. The injury did not deviate them from their path. It moved the path. This is answer two.

Nelson and Plosser looked at a century of the American economy and concluded that, for thirteen of their fourteen series, they could not reject the hypothesis that the economy is Person B. Only unemployment behaved like a series that returns to a fixed level, which makes intuitive sense, since unemployment cannot trend upward forever.

The implication that shook the profession

Why did this land so hard? Because of what it implies about the cause of recessions.

The dominant view at the time, broadly Keynesian, held that business cycles were driven by fluctuations in demand: a wave of pessimism, a monetary contraction, a spending slump. Those things are, by their nature, temporary. Demand recovers. And crucially, temporary causes should produce temporary effects. A demand-driven recession should eventually be undone.

But Nelson and Plosser found that recessions appeared to be permanent. And permanent effects require permanent causes. If output never recovers the lost ground, then something about the economy's actual productive capacity must have changed, not merely how much people happened to be buying.

That points away from demand and toward supply: technology, productivity, the real capacity of the economy to produce things. Their result gave enormous ammunition to the emerging real business cycle school, which argued that cycles are driven by real technology shocks rather than by demand failures, and which was about to dominate macroeconomics for a generation.

Nelson and Plosser did not merely make a statistical point. They made a statistical point with a very large policy shadow: if recessions are permanent supply shocks rather than temporary demand shortfalls, then the case for demand-side stabilisation policy is much weaker.

Why it mattered

• It made unit roots the central preoccupation of time-series econometrics for two decades. Almost everything covered in a modern time-series course, unit root testing, cointegration, structural break analysis, exists downstream of this debate. This paper created the demand for those tools.
• It reframed what a business cycle even is. Whether shocks are permanent or temporary is not a technical footnote. It determines what you think a recession is.
• Their dataset became a benchmark. The "Nelson-Plosser data" was used and re-used for the next twenty years as the standard test bed for every new unit root and structural break method. Perron used it. Zivot and Andrews used it. KPSS used it. It became the field's shared proving ground.
• The core question is alive in finance too. Is a drawdown in a strategy a temporary deviation from its true edge, or a permanent decay in that edge? Is a stock's price fall a deviation from fair value or a repricing of fair value? These are Nelson-Plosser questions, and answering them wrong is how careers end.

The honest limitations

The subsequent literature was, in effect, one long argument with this paper, and the arguments were good ones.

• Failing to reject is not proving. This is the deepest problem and it is fundamental. Nelson and Plosser did not prove that macro series have unit roots. They failed to reject the hypothesis that they do. Given that unit root tests are notoriously weak at distinguishing a random walk from a series that reverts slowly, that failure may say more about the test than about the economy. KPSS made exactly this argument a decade later, and found that for many of the same series, you cannot reject stationarity either. The honest reading is that the data was too weak to decide, which is not what a generation of macroeconomics built on top of it assumed.
• Perron's structural break objection is devastating and direct. Perron showed that a series that is perfectly stationary around a trend, but which experiences one permanent break (say, the 1929 crash), will be diagnosed as a random walk by exactly the tests Nelson and Plosser used. Rerun their analysis allowing for a break and most of the unit roots vanish. Was the Great Depression a permanent shift in a stable economy, or evidence that the economy has no anchor? The data cannot easily tell you, and your answer determines your conclusion.
• A century of data still is not much data. For distinguishing a random walk from a very slowly reverting process, what matters is not the number of observations but the span of time covered. A hundred years of annual data on output is a hundred observations. That is a small sample for a subtle question.
• The economic interpretation was a leap. Even granting the statistics, the jump from "output has a unit root" to "therefore supply shocks drive cycles" involves theoretical assumptions that not everyone accepted, and the debate over that inference was fierce.

The one-line takeaway

Nelson and Plosser found that a century of American economic data could not reject the possibility that recessions leave permanent scars rather than temporary dents, a conclusion that reoriented macroeconomics toward real supply-side theories and launched twenty years of econometric work, much of which was devoted to showing that their evidence was far less conclusive than everyone had treated it as being.