Labor market institutions and unemployment: can earlier findings be replicated?(Author abstract)

Introduction and Review

Replication in empirical economics is vital. Indeed, it has been argued that a result is not worth knowing if it cannot be reproduced (Mittelstaedt and Zorn, 1984). Because confirming econometric results serves a public interest, this journal gives priority to publishing replication studies. (1) Unfortunately, would-be replicators may be stymied by data problems (Feigenbaum and Levy, 1993). They may not have access to data used in original studies. Even if time series are available, the series often are revised, so that the numbers available to a replicator are not necessarily the same ones that were available to the original researcher. Moreover, it is not always clear what formulations were used to generate results, so findings cannot always be reproduced. For example, Wulwick (1996) described in detail the difficulties associated with trying to reproduce original Phillips Curve estimates.

There is a difference between attempts to duplicate results, so-called "econometric audits," and efforts to reproduce and extend findings, known as "replication with extension." (2) The latter efforts may try to keep all known conditions similar or they may involve variations in the original study. In either case, replication with extension tries to generalize the validity of empirical findings.

If empirical findings are robust--that is, they can be duplicated and extended--then we can have more confidence about policy recommendations based on those findings. Perhaps no policy issue receives more attention, and is more controversial, than labor market reforms. Despite chronically high unemployment in countries such as France or Germany, there is no consensus in those places about proposals to alter labor market institutions.

In a well-known book, Layard, Nickell, and Jackman (1994) tried to measure how unemployment is related to labor market institutions. Using a cross-section of 20 industrialized countries and data for 1983-1988, they found that changes in inflation and institutional variables can explain a large majority of unemployment differences across countries. To represent labor market institutions, they considered aspects of collective bargaining structure and labor market programs. Unemployment benefits are regarded as a passive type of program for the unemployed. Active help for the jobless includes training, placement services, direct job creation, and recruitment subsidies. In their analysis Layard, Nickell, and Jackman considered duration and amount of unemployment benefits. They also considered expenditures on active labor market programs.

Constructing measures to represent labor market institutions requires many judgments. For example, how does one classify a country's collective bargaining structure? What does one include in measuring expenditures on active labor market programs? Layard, Nickell, and Jackman discussed how they constructed their variables. Moreover, they provided the actual data used in their computations. Based on Layard, Nickell, and Jackman's data and discussion, we seek to duplicate their results. (3)

Since the Layard, Nickell, and Jackman book was published unemployment has been particularly acute in countries such as France, Germany, Italy, and Spain. Perhaps it is not surprising that research on the effectiveness of labor policies has concentrated on Europe. (See Kluve, 2006.) Analysts such as Belot and van Ours (2000) have pondered whether "cleverly designed" reforms can combat unemployment. Indeed, some countries have altered some of their labor market institutions. Therefore, it is particularly important to update the analysis of unemployment and labor market institutions.