Discussion Paper

No. 2008-24 | June 17, 2008
Inflation Targeting Is a Success, So Far: 100 Years of Evidence from Swedish Wage Contracts


Inflation targeting was adopted by several countries, including Sweden, in the 1990s. We evaluate the Swedish inflation targeting regime since 1995 using a novel approach based on a unique data set on the characteristics of collective wage agreements between 1908 and 2008. First, we establish that the length of wage contracts decreases in response to an increase in “macroeconomic uncertainty” across policy regimes. Second, using contract length as the assessment criteria for regime performance, we find that the inflation targeting regime of 1995–2008 stands out as an exceptionally stable policy regime as judged by the willingness of wage contract-makers to repeatedly commit to three-year non-indexed wage agreements. In addition, inflation targeting gained instant credibility in the sense that the labor market organizations entered long-term wage agreements at the same time as this new regime was announced.
Inflation targeting has thus reduced macroeconomic uncertainty compared to previous regimes adopted in Sweden during the 20th century. Our approach to evaluate inflation targeting is different from the traditional one commonly based on cross-section samples comparing inflation outcomes. Instead we focus on the actual decisions of private-sector wage setters under different monetary regimes. Judging from their behavior across a century of observations, inflation targeting in Sweden is a success – at least so far.

Submitted as Policy Paper

JEL Classification

E30 E42 E65

Cite As

Klas Fregert and Lars Jonung (2008). Inflation Targeting Is a Success, So Far: 100 Years of Evidence from Swedish Wage Contracts. Economics Discussion Papers, No 2008-24, Kiel Institute for the World Economy. http://www.economics-ejournal.org/economics/discussionpapers/2008-24


Comments and Questions

Anonymous - Associate Editor Report
June 23, 2008 - 15:51

see attached file

Klas Fregert - Author comment on editor report
July 09, 2008 - 11:39

Comments on ”Report on ‘Inflation targeting is a success, so far: 100 years of evidence from Swedish wage contracts’ by Klas Fregert and Lars Jonung, May 26, 2008” by anonymous editor.

Klas Fregert and Lars Jonung

The use of a linear regression with 7 observations:

... />Since the regression of length against wage inflation variability fails to convince the reviewer beyond the graph, due to the low number of observations, we certainly consider deleting the regression. On our part, we felt that the use of degree-of-freedoms adjustment implied by the t-distribution made the regression admissible, but appreciate that this is a cause for concern. Regarding the claim that the regression is based on an incomplete model, we are not sure if the reviewer questions the derivation of wage variability as an ideal proxy given the assumptions of the Gray (1978) model or if he considers the Gray model incomplete.

Real uncertainty may lead to longer contracts

We dismissed the Danziger hypothesis model of contract length as a positive function of real uncertainty on two grounds: the consistency of the long-run data with the Gray model and the lack of support for it in the latest previous study (see footnote 10). In addition, we have failed to find evidence in the historical material on this effect, while there are several instances when the parties refer to the unclear business cycle situation when entering into short contracts. Finally and most importantly, we have dismissed the hypothesis as at odds with the historical evidence: long contracts are associated with stable output growth and stable inflation, e.g. during the last period of the “Great Moderation”, well documented on a global scale. The reviewer’s point is well taken and we will be more explicit about our reasons for dismissal of the real uncertainty effect.

The definition of a regime

We will clarify that our regime division is one of policy regimes rather than monetary regimes to take account of the reviewer’s point that a floating exchange rate regime as such or a “full employment regime” are not proper monetary regimes. What we label the “full employment regime” is one without a long-run nominal anchor. We will check our language to clarify our division into periods, which in itself we believe is conventional and used in many cross-regime studies.

The centralization effect on length

We included centralization in our regression as we relied on an extension of the Gray model by Groth and Johansson (2004). Instead of the expected positive effect, we found a negative effect on contract length. The reviewer’s interpretation of centralization causing shorter contracts during high inflation is interesting. We will consider it briefly if we retain the regression.

What is the best evidence for the success inflation targeting?

In the last paragraph the reviewer mentions previous evidence on the success of inflation targeting: the quick convergence to low inflation and the decrease in the dispersion across forecasters of expected inflation around the target. The reviewer’s comment alerts us to that there are several measures of the success of inflation targeting in addition to the mean and standard deviation of the inflation rate stressed in most studies. In addition to the long sample across many regimes, we stress that our measure, contract length, is an uncertainty measure which is closely related to inflation uncertainty. We will make some further reference to the literature on inflation uncertainty for comparison with our measure, including the dispersion among forecasters. All the suggested measures of inflation uncertainty are open to interpretation as analyzed i.a. by Giordani and Soderlind (EER, 2003). We have found only one study on the effect of inflation targeting on the dispersion of inflation expectations (negative as expected), Working paper 2007-11, Banco de México.

Anonymous - Referee Report
July 24, 2008 - 11:54

see attached file

Klas Fregert - Author response
July 26, 2008 - 18:32

Author comment

The referee lists six “substantive problems”, which we respond to point by point.

1. The referee suggest that the synchronization of contracts between LO and SAF caused the lengthening of wage agreements in 1992. Wage agreements have been synchronized since 1956 through central wage agreements between ...[more]

... LO and SAF in Sweden. Thus union contracts have been continuously synchronized from 1956 until now. This is mentioned in section 2.

The referee takes up the possibility that price uncertainty may have decreased for other reasons than inflation targeting, such as a decrease in oil price volatility. We discuss in general terms in section 3.2 the identification problem of separating out other reasons than inflation targeting as a cause of the increase in contract length after the introduction of inflation targeting. In panel studies, which include countries with and without inflation targeting, it is in principle easier to control for other sources of a decrease in volatility, such as oil shocks. Still the two panel studies we have found, see footnote 30, which try to control for other sources of lower and more stable inflation come to different conclusions of the link between inflation targeting and reduced inflation variability.

We view our one-country study of many regimes as a complement to the panel studies, which use before and after inflation targeting samples. We argue that the introduction of inflation targeting is linked in our historical sample to a decrease in perceived uncertainty (as opposed to actual used in the panel studies). We stress that the increase in length comes directly after the introduction of inflation targeting. This argument is strengthened by the observation that previous regime changes have changed length slowly after a regime change, such as the introduction of the Bretton Woods regime. This is discussed more fully in section 5. In addition, there was a major inflation shock 1989-1990 before the steep downturn of the Swedish economy 1992-94, which makes the introduction of 3-year contracts in 1995, not seen since 1969, even more remarkable. This also suggests a role for inflation targeting, since clear signs of any actual reduction in inflation variability had yet to materialize in 1995. We will look over our writing to emphasize this point.

2. According to the referee: “What specific features of inflation targeting contribute to a longer time horizon of wage contracts? This issue needs to be addressed in the paper. In particular, the forward-looking and the forecast-based nature of inflation targeting are likely to generate such result.” This is a point we will try to clarify. Our cross-regime test of the Gray length model is based on the assumption of nominal wages and prices following random walks. We show that the forecast uncertainty of the future level wages increases over the forecast horizon in direct proportion to the one-period forecast uncertainty, which in turn we show can be proxied by the actual standard deviation of one-year wage inflation. In this sense we treat all regimes alike. This is a natural assumption for all the regimes we study, except perhaps the gold standard. The price level is a random walk in the inflation targeting regime, but with slower increase in forecast uncertainty than the other regimes due to how the policy is implemented (forward-looking) in addition to possible effects of its institutional anchoring. The gold standard regime appears in other studies (Benjamin Klein, 1976) to have exhibited less long-run price-level uncertainty than short-run uncertainty, due to a larger mean-reverting component in the price level. In our sample, however, the gold-standard period is a short one, and one which believe should be seen as special (see further point 4 and technical point 3 below).

3. According to the referee: “The authors’ imply that inflation targeting has reduced ‘macroeconomic uncertainty’ by containing wage indexation.” This must be a misunderstanding by the referee. We only take up indexation as a response of the contract makers to increased perceived uncertainty. It is a treated as an imperfect substitute for length reductions and therefore as an indicator of the perceived macroeconomic uncertainty in addition to length. We do not mean to discuss the causal link from wage indexation to macroeconomic uncertainty.

4. According to the referee “It seems that during the gold standard period and the later stage of Bretton Woods fixed exchange rate regime the wage contract length exceeded the three-year period prevalent
during the inflation targeting policy. Therefore, it needs to be proven that inflation targeting is still superior to these two alternative regimes.” As we mention, the long contracts during the gold standard and the Bretton Woods period only lasted for a short period in contrast to the inflation targeting regime. We believe it is an important aspect of the inflation targeting regime that it has lasted a long period in a secular perspective, with five successive long three-year wage contracts. In this respect it deserves to be called a success relative to the gold standard and the Bretton Woods periods. This finding complements the panel study of the length of the inflation targeting regimes by Rose (see footnote 30). In addition, we argue that the gold standard period in our sample is special due to the general strike in 1909 and that it was the starting period of collective agreements, which we discuss in section 5.

5. According to the referee “The variance analysis based on Eq.(3) and reported in Table 2 seems rather irrelevant for this paper considering that the variance of wage contracts during the inflation targeting period is zero.” We are unsure of the meaning of this point. Table 2 gives the regression of average new length per regime as a function of the standard deviation of the wage inflation variance as specified in equation 3 and derived from Gray (1978). The variance of length does not figure in the regression.

6. The referee asks: “What was the role of fiscal discipline in expanding the length of wage contracts?” We will add to the discussion of the role of fiscal discipline in the paper that fiscal discipline was strong during the gold standard and the during the Bretton Woods period, without leading to sustained periods of long contracts. In addition, even though strong signals had been given beginning in 1994 about the need for fiscal discipline, the budget deficits were still large and the new budget law was not introduced until 1996, that is after the introduction of inflation targeting and three-year contracts.

The referee also takes up “three technical imperfections”.

1. We will shorten the abstract
2. We will make a cross-reference to footnote 30, which cites four recent empirical references.

3. The referee notes that “the coefficient of variation is considerably
lower for both the gold standard and the Brettom Woods regime periods. This deserves further explanation.” Regarding the gold standard, we only use the 1908-1914 period, when prices were rising at a faster pace than during the whole classical gold standard period. Regarding the Bretton Woods period, real wage growth was higher than all the other periods, and hence also nominal wage growth for a given inflation rate. This explains that the coefficient of variation is lower in those periods than during the inflation targeting regime. The Gray theory is predicated on the contract makers minimizing the dead-weight loss, which is proportional to the standard deviation of nominal wage growth, not the coefficient of variation.

Anonymous - Associate Editor Decision
July 31, 2008 - 12:44

see attached file