The Role of Shocks and Institutions in the Rise of European Unemployment: The Aggregate Evidence

Questions & answers

Q1. What are the two “key facts” about European unemployment the paper sets out to explain, and why does each class of single-cause explanation fail on its own?

The paper states the two facts to be explained are “the rise in unemployment since the 1960s, and the heterogeneity of individual country experiences” (Abstract). Explanations based solely on shocks “run… into a major empirical problem”: shocks “can potentially explain the general increase in unemployment over time,” but “they do not differ enough across countries to explain the cross-country variation” (p. 3). Explanations based solely on institutions run into the opposite problem: “many of these institutions were already present when unemployment was low… and, while many became less employment-friendly in the 1970s, the movement since then has been mostly in the opposite direction,” so that institutions alone “do not appear able to explain the general evolution of unemployment over time” (p. 3).

Q2. What three shocks does the paper identify as most important, and why is the slowdown in TFP growth judged an incomplete explanation of today’s high unemployment on its own?

The paper identifies three shocks: the decline in total factor productivity (TFP) growth starting in the early 1970s (from close to 5 percent in the 1960s to about 2 percent since the late 1970s), movements in the real interest rate (sharply negative in Europe in the second half of the 1970s, then strongly positive through the 1980s-90s), and shifts in labor demand (Sec. 1, pp. 5-10). On TFP growth specifically, the authors argue that “theory suggests that the answer, to a first approximation, is no” to whether its effects on unemployment can be permanent, since once wage expectations adjust to the lower growth rate the effect should mostly dissipate – “after more than twenty years, it is hard to believe that its effects are not largely gone. So, what accounts for today’s high unemployment? There is much less agreement here” (p. 7).

Q3. Through what two channels does the paper argue labor market institutions can shape how shocks translate into persistent unemployment?

The paper distinguishes institutions that affect the impact of a shock on unemployment (for example, more coordinated collective bargaining may allow real wages to adjust faster to a productivity slowdown) from institutions that affect the persistence of unemployment once a shock has raised it (Sec. 3, pp. 17-18). On the persistence channel, the central mechanism is duration-dependence: “if some of the unemployed remain unemployed for a long time, they may either stop searching or lose skills” and become “irrelevant to wage formation,” so that “the pressure of unemployment on wages decreases, and unemployment becomes more persistent” – a channel the authors attribute to Layard and Nickell (1987) and link to their own earlier “hysteresis” work (Blanchard and Summers 1986) (p. 18).

Q4. What empirical specification does the paper use when treating shocks as unobservable but common across countries, and what is the central result?

The specification regresses the unemployment rate on country fixed effects, period (time) fixed effects, and the interaction of the time effects with eight time-invariant measures of labor market institutions (following Nickell 1997): the benefit replacement rate, benefit duration, active labor market policy spending, employment protection, the tax wedge, and three measures of collective bargaining (coverage, union density, coordination) (Sec. 4, eq. 1, pp. 20-21). All eight institutional coefficients have the predicted sign, and all but the union-coverage variable are statistically significant; the authors describe the results as “surprisingly strong (relative to our priors),” and report that “interactions between common shocks and different institutions can account for much of the actual difference in the evolution of unemployment rates across countries” (pp. 21-23), with the model’s R² of 0.863 close to the 0.903 obtained from a fully unrestricted model with separate time-country interaction terms.

Q5. How large does the estimated effect of institutions turn out to be, in the paper’s own numerical example?

Using the estimated coefficients, the paper illustrates that a shock which would raise unemployment by 1 percentage point in a country with mean employment protection would raise it by only 0.58 percentage point in a country with the lowest sample value of employment protection, but by 1.42 percentage points in a country with the highest value – a roughly two-and-a-half-fold difference in the impact of the identical shock depending solely on this one institution (Table 1, Sec. 4, p. 22). The paper notes that, across all eight institutions examined this way, “the range of the effects of institutions on the impact of a given shock on unemployment is roughly similar across institutions” (p. 22).

Q6. How does the second empirical approach, using country-specific observable shock series, relate to the first, and what caveats does the paper attach to both?

The second specification (Sec. 5) replaces the unobserved common time effects with constructed, country-specific series for the shocks identified in Section 1 (TFP growth, the real interest rate, and labor demand shifts), again allowing their effects to be scaled by the same institutional measures; the paper reports that “the second constructs series for the macro shocks, and again finds evidence that the same size shock has differential effects on unemployment when labor market institutions differ” (Abstract). The authors are explicit about the preliminary nature of both exercises: “the results are typically weaker when we allow for time-varying rather than time-invariant measures for institutions,” and “the fact that the specifications fit the data does not prove that the underlying theories are right; just that they are not obviously inconsistent with the aggregate data” (p. 4).

Q7. How does the paper interpret the economic mechanism behind the shocks-institutions interaction?

The paper’s interpretive summary is that “institutions determine the relevance of the unemployed to wage-setting, thereby determining the evolution of equilibrium unemployment rates following a shock” (Abstract) – that is, institutions do not necessarily cause unemployment to rise on their own, but they govern whether, after an adverse shock has raised unemployment, the resulting pool of unemployed workers continues to exert downward pressure on wage growth (allowing unemployment to fall back) or becomes increasingly disconnected from wage bargaining (allowing unemployment to persist at the new, higher level).

Q8. What outlook does the paper offer for the future of European unemployment, conditional on its account being correct?

The conclusion states that “if our account is correct, one can be mildly optimistic about the future of European unemployment,” reasoning that “the effects of some of the adverse shocks should go away” – the real interest rate was expected to be lower than in the recent past, the dynamic effects of earlier adverse labor-demand shifts were expected eventually to turn favorable, and labor market institutions were “slowly becoming employment-friendly,” so that “the more favorable macroeconomic environment and the improvement in institutions should lead to a substantial decline in unemployment” (Sec. 6, p. 33). The authors nonetheless describe their results overall as “preliminary,” citing a crude dynamic specification, missing time series for several institutions, and concern about the endogeneity of labor market institutions themselves (p. 33).