A slump can end while some of its damage remains. Production begins to grow, the queue for work shortens, and the economy again moves forward; none of these changes establishes that it has regained the path it might have followed without the interruption. A factory never built will not appear merely because orders have stopped falling. A worker who has lost contact with a trade may need more than the return of a vacancy. The distinction is between recovery of growth and recovery of the level from which growth proceeds. An economy that resumes the same percentage growth from a lower base remains below its former projected path. The Cycle has so far asked how activity departs from a trend and how it might return. Its final question reaches into the trend itself: can the passage through recession alter the productive possibilities and institutions to which the economy returns?
Hysteresis names the possibility that an economy's resting position depends on the route by which it arrived. In their 1986 account of unemployment, Olivier Blanchard and Lawrence Summers challenged the clean separation between a slowly changing equilibrium rate and an actual rate that temporarily wanders around it. The experience of unemployment, in their models, can change the rate toward which adjustment proceeds. They distinguished strict dependence of the long-run equilibrium on history from the looser usage in which a disturbance affects equilibrium unemployment for a long time. That distinction matters. A shock that fades very slowly can impose immense losses without lasting forever; a permanently altered equilibrium is a stronger claim. Both require a mechanism carrying yesterday into today. Merely observing that unemployment is still high supplies the phenomenon to explain, while the explanation must identify what the intervening experience has changed.1
One mechanism runs through the people represented when wages are set. In Blanchard and Summers's pure insider model, incumbent workers choose a wage with their own expected employment in mind, and firms subsequently choose how many to employ. A contraction can remove some workers from the group whose interests shape the next bargain. The smaller surviving group need not choose a wage that would restore the employment of those excluded. Under the model's strongest membership assumption, the loss can persist even after the original surprise has passed. The qualification is as consequential as the result: outsiders may exert pressure, new firms may hire them, and fear of unemployment may alter the incumbents' demands. When those forces operate, adjustment toward a prior benchmark can resume. The model locates a possible dependence on history in bargaining and membership; it does not make every wage bargain, every union, or every observed recession obey that particular account.2
The duration of exclusion can create another connection between the slump and what follows it. A long spell without work may weaken skills, search, contacts or an employer's confidence in an applicant; the institutional and personal responses to unemployment can then affect the prospect of leaving it. Blanchard and Summers examine duration through the possibility that the long-term unemployed exert less influence on wage setting than people recently employed. Their discussion treats the explanations as hypotheses, and explicitly notes the limited evidence then available for skill deterioration. The distinction should survive the appealing image of a wound. A person still looking for work is not thereby shown to have become less capable; an employer's inference about a long absence is not itself proof of lost ability. The analytical task is to identify whether exclusion changes productivity, access, bargaining or information, and whether a change in those conditions can restore the opportunities that ordinary recovery leaves closed.3
A simple pair of calculations shows why persistence needs to be specified. Suppose a disturbance initially adds five percentage points to unemployment above an otherwise unchanged benchmark. If four-fifths of the excess survives each period, only about 1.64 points remain after five periods, and eventually the excess disappears. If all of it survives, the five-point displacement remains; if none survives, it is gone in the next period. Investment has a comparable distinction. Start with capital of two hundred, depreciation of one-tenth each period, and investment of twenty maintaining the stock. One period of investment at ten leaves capital at one hundred and ninety. Restoring investment to twenty raises it to one hundred and ninety-one in the next period and gradually back toward two hundred. Merely replacing depreciation on the reduced stock holds it at one hundred and ninety. The lasting difference in this second case depends on the subsequent investment rule. A temporary reduction in investment alone has not proved a permanent loss.4
Evidence of a scar therefore requires a defensible account of the path that would otherwise have been followed. Drawing a straight line through an earlier expansion is a starting comparison, and can be a misleading one if that expansion was already unsustainable. A later shortfall might reflect the consequences of recession, an independent change in productivity or population, a new disturbance, or errors in the earlier forecast. Valerie Cerra and Sweta Saxena make the distinction between renewed growth and return to the former path central to their study of economic recovery. Their theoretical discussion also shows why different models predict different outcomes: restoring a depleted capital stock can bring catch-up, while a change in the determinants of the steady state or in the level of productivity can leave a lasting difference. The observed path must be used to discriminate among such accounts. A statistical description of persistence cannot, on its own, identify which policy failure or economic mechanism caused it.5
The possibility of scarring makes the timing and composition of recovery matter beyond the date at which growth turns positive. Preventing avoidable exclusion can preserve something that would be costly to rebuild; restoring demand can help where orders remain the obstacle; training, access and renewed investment may matter where the interruption has changed the means of production or the routes into employment. Each response still needs a diagnosis. The existence of lasting damage does not imply that unlimited spending can recover every lost path, any more than a return to growth establishes that the damage has vanished. The volume ends with a less comfortable relation between the cycle and the trend than the wave first suggested. A downturn occurs within an economy that has a history, and can become part of that history's machinery. The hands left idle and the projects left unbuilt may change what the next expansion can accomplish. Growth then inherits the consequences of the cycle it seemed merely to rise above.