Abstract
This paper reviews economic developments in Iceland during 1990–96. It analyzes the origins of the current economic expansion associated with a swing in the current account and in emerging inflation pressure. Three driving forces are emphasized: the positive supply shock affecting the fisheries; the expansion of the power intensive industry; and brisk increases in real wages over the past two years (1995–96). The paper highlights that the main sources of upside risks comprise the likely construction of a new aluminum smelter.
II. POTENTIAL OUTPUT AND THE OUTPUT GAP14
A. Introduction
32. This chapter presents tentative estimates of potential output and the output gap for Iceland. For the short term, measures of the cyclical position of the economy help to assess inflationary pressures and the current stance of macroeconomic policies. For the medium term, measures of potential output generally provide a useful guide to aggregate supply capacities and help to identify the scope for sustainable non-inflationary growth and sources of structural imbalances.15
33. The definition and estimation of the trend and cyclical position admittedly raise a number of unresolved theoretical and empirical questions, which reflect the still controversial debate on the origins of economic fluctuations (see, for instance, Canova, 1993). These difficulties are, however, significantly accentuated in the case of Iceland for several obvious reasons. First, Iceland is a small open economy subject to frequent and large supply shocks originating in the fishing sector. Second, Iceland’s history of deep-rooted high inflation, which prevailed until the late 1980s and reflected excessively accommodative macroeconomic policies, suggests that cyclical demand pressures have not been a primary source of inflation in the past. Third, the Icelandic economy has undergone major structural changes in recent years—including profound financial liberalization reforms, increased competition and openness, and further diversification of its narrow production base—and the recent move towards stable macroeconomic policies brought a remarkable break with the past inflationary environment, with presumably large consequences on the formation of private agents’ expectations and decisions. Altogether, these factors render any measure of trend and cyclical output very delicate, and some would even question their relevance for assessing potential inflationary pressures in Iceland.
34. The analysis of the paper provides the following main conclusions. First, although the paper highlights the large uncertainty surrounding estimates of potential output and the output gap, and their inherent dependence on the underlying assessment of the medium-term outlook, it also shows that different reasonable approaches provide very similar results and underscore, in particular, the fairly modest expansion of potential output in recent years. This could reflect a notable rise in the non accelerating rate of unemployment and some other factors, including low capital accumulation and limited factor productivity growth in the non-fish business sector, together with recent adverse supply developments in the fishing sector. Second, the paper presents some tentative results consistent with the existence of “speed-limit” effects in the formation of inflation, suggesting that fluctuations in activity and unemployment, rather than their level per se, would affect the inflation outcome in Iceland. Third and foremost, the results show that the present economic upturn, with the sharp acceleration of activity in 1996, has brought the economy close to, and likely somewhat above, its potential capacity. Though consumer price inflation remains fairly modest at present, this underscores the risk that, together with a further deterioration of the current account, inflationary pressures could start to develop should activity fail to slow down as expected in 1997.
35. The chapter is organized as follows. Section B reviews several detrending techniques and presents the results they generate for the Icelandic economy. Section C considers a more traditional approach, based on a production function for the non-fish business sector. Section D examines the past empirical links between measures of output gaps and inflation and Section E summarizes the main findings.
B. Estimates Based on Detrending Techniques
Approaches to estimating trend output
36. In this section, we briefly discuss three detrending methods which avoid some of the major shortcomings of more traditional approaches, such as the standard split time-trend method: the Hodrik-Prescott filter, the non-parametric approach developed by Coe and McDermott (1996) and the multivariate Beveridge-Nelson decomposition (Evans and Reichlin, 1994).
37. The Hodrick-Prescott filter (HP) is a smoothing procedure which has become increasingly popular because of its flexibility in tracking the characteristics of the fluctuations in trend output and because the method is intuitive and easy to apply. Trend output (denoted by y*) derived using the HP filter is obtained by minimizing a combination of the gap between actual output (y) and trend output and the rate of change in trend output for the whole sample of observations (t):
where λ determines the degree of smoothness of the trend.
38. Properties and shortcomings of the HP filter have been well documented (Danthine and Girardin, 1989, Harvey and Jaeger, 1993). A major drawback comes from the inability of the filter to properly extract the trend and cyclical components for a vast set of standard statistical processes, and the difficulty in identifying the appropriate detrending parameter, λ which is often overlooked by using arbitrary values popularized by the real business cycle literature. In particular, mechanical detrending based on the HP filter can lead to spurious cyclical behavior. A second important flaw of the HP filter arises from its high edge-sample biases which reflect the symmetric trending objective of the method across the whole sample and the different resulting constraints that are imposed within the sample and at its edges. The use of medium-term growth projections can, however, help to reduce this end-point bias.
39. The detrending method proposed by Coe and McDermott (1996) also identifies trend output based on a nonparametric regression estimation, whose functional form need not be specified, for it. The approach basically assumes that the trend has an adequate number of derivatives so that it is smooth enough compared to the cyclical component, and thus allows quite flexible functional forms to be considered. The originality of the method is that the degree of smoothing, based on the size of the data window used for each observation, is determined by the minimization of a fairly general global error criterion. Contrary to the HP filter, there is therefore no need to specify an arbitrary smoothing parameter or data window size. Whether the kernel and the general error criterion used are most appropriate for extracting trend output from actual GDP in each country case remains, however, unclear.
40. The multivariate extension of the Beveridge-Nelson approach (Evans and Reichlin, 1994) provides a forecast-based decomposition of output into trend and cycle, using the information contained in a whole set of economic variables, when output is integrated of order 1. The method allows the various economic variables used to be non-stationary and assumes that they can be represented by a standard vector autoregressive (VAR) model. After proper identification of the long–run relationships and the short–term dynamics of the model, the method identifies the trend component of output as a function of the stochastic trends that have permanent effects on GDP and the cyclical component as the sum of all the future expected changes in output that cannot be explained by its deterministic trend. Several studies have underscored the advantages of this method which potentially relies on stronger economic foundations and is not subject to end-sample biases (Barell and Sefton, 1995). However, this rather sophisticated approach also suffers from major weaknesses. In particular, quantitative features of the estimates can depend significantly on the information set and the precise econometric specification used.
Empirical estimates
41. In order to reduce the end-sample bias, the HP filter was applied on a large sample of observations (1945-1996) complemented by staff medium-term projections, with the two standard detrending parameters λ = 25 and λ = 100. 16 Notwithstanding the dependence on these uncertain projections (see below), the results obtained are quite similar (Chart 1). They show how Iceland’s activity was subject to large fluctuations in the after-war period, although there is no clear evidence that domestic fluctuations have been higher than in other industrial countries since the early 1970s. As expected, the last economic downturn appears particularly sharp with an estimated deterioration of the cyclical position close to 10 percent of trend GDP from 1987 to 1993, but the cyclical trough would be in line with the preceding recessions. The results also suggest that trend growth has been fairly modest since the early 1990s and that the present upswing, with hefty growth in 1996, has brought actual output close to and possibly somewhat above potential (by about 1 percent, with λ = 25). Although part of last year’s GDP growth is explained by the expansion of the existing aluminum smelter which is expected ultimately to deliver additional productive capacity of 1 percent of GDP, this observation would be consistent with emerging pressures observed in the labor market and with the marked deterioration in the current account. Estimates for the present period are, however, significantly dependent on the medium-term growth projections that are used in the estimation sample. As a simple illustration, alternative estimates assuming (less realistically) a constant higher GDP growth of 3.5 percent per year over of 1997-2001, would suggest that actual output was respectively at or 1 percent below potential in 1996 with λ=25 and λ =100, while a more pessimistic assessment of the medium-term outlook, with growth at 1.5 percent per year, would tend to indicate that GDP was 1½ to 2 percent above potential last year.
ICELAND: TREND GDP AND OUTPUT GAPS
(estimated by Hodrick-Prescott filter)
Citation: IMF Staff Country Reports 1997, 015; 10.5089/9781451819205.002.A002
Sources: National Economic Institute; and staff calculations.1/ Based on staff medium-term projections, with real GDP growth at 3.0 and 2.4 percent in 1997 and 1998, and 2.3 percent for the later years.42. The results derived from the Coe and McDermott method share the main characteristics of those obtained with the HP filter, and they also suggest that actual output closed the gap and expanded beyond its trend level by about 1 percent in 1996 (Chart 2). As for the third set of estimates, based on the multivariate Beveridge-Nelson decomposition, it was obtained with a VAR model composed of real GDP, real consumption and real exports of goods and services (in logs) over 1945-1996, and no evidence of cointegration was found between these series–chosen for their presumed information content for short and long-term fluctuations in output. Though they also notably point to a resorption of the output gap in 1996, these last estimates appear quite unreliable since their interpretation is somewhat problematic and their shape very sensitive to the information and the econometric specification used (order of the VAR and estimation sample).
ICELAND: OUTPUT GAPS
(estimated by non-parametric method and multivariate Beveridge-Nelson decomposition)
Citation: IMF Staff Country Reports 1997, 015; 10.5089/9781451819205.002.A002
Sources: National Economic Institute; and staff calculations.1/ See Coe and Mc Dermott (1996). Computations are based on the program made available by the authors.2/ Based on a VAR model of order 4 composed of real GDP, private consumption and exports (in log).C. Estimates Based on a Production Function
43. In order to identify the sources of potential growth and imbalances, this section presents tentative estimates of potential output and the output gap derived from a standard structural production–based approach.
44. As noted in the introduction, a number of characteristics of the Icelandic economy render estimates of potential output particularly uncertain. In practice, these specificities also contribute to increasing notably the usual difficulties met in finding an appropriate stable production function and defining the “normal” degree of use of factor inputs. The structural transformation due to the recent financial liberalization, together with increased competition and openness, increased diversification and the modernization of the fishing sector, and the trend decline of investment in the public sector, make it difficult to assess the efficiency of the aggregate capital stock in the economy, and no direct indicator of capital utilization is available. As for labor input, trend progress in the educational background of the labor force and the development of part–time work should also be taken into account as for other countries. In addition, available indicators of the number of hours worked are generally believed to be a poor guide to actual labor input in Iceland, as they have fluctuated sharply in the last decades and are admittedly biased upward due to the system of remuneration. Assessing the “normal” degree of use of labor input is further problematic since participation rates have displayed large swings in the last decades, reflecting structural trends such as increased female participation, but also suggesting also a high responsiveness to economic conditions and changes in the taxation and welfare system, as illustrated by the importance of migration flows (Chart 5).
ICELAND: UNEMPLOYMENT and NAIRU
Citation: IMF Staff Country Reports 1997, 015; 10.5089/9781451819205.002.A002
Sources: National Economic Institute; and staff calculations.1/ Agriculture, fishing and public services other than hospitals excluded.ICELAND: POTENTIAL GDP AND OUTPUT GAPS
(estimated by structural method)
Citation: IMF Staff Country Reports 1997, 015; 10.5089/9781451819205.002.A002
Sources: National Economic Institute; and staff calculations.ICELAND: LABOR INPUT
(Structural method)
Citation: IMF Staff Country Reports 1997, 015; 10.5089/9781451819205.002.A002
Sources: National Economic Institute; and staff calculations.1/ For full-time employed workers.The NAIRU
45. Measuring potential output requires empirical estimates of the non-accelerating inflation rate of unemployment (NAIRU) which are known to be highly uncertain.17 As noted above, a number of factors complicate further any estimation of the NAIRU for Iceland, inasmuch as large supply shocks have presumably induced frequent shifts in the past. In this context, we will confine ourselves to using simple analysis.
46. The Icelandic labor market was viewed to be very flexible until the late 1980s, as real wages seemed to be rather sensitive to changes in unemployment, while recent years tend to indicate a different picture (Gudmundsson, 1995). This very likely reflects a shift in the short-run Phillips curve and an increase in the NAIRU as suggested by the top left panel of Chart 3. Apparent shifts in the Beveridge curve and in the standard Okun relation also support this view. As changes in unemployment affect changes in inflation more clearly than does the level of unemployment, this increase in the NAIRU may reflect not only a shift of the underlying structural rate of unemployment but possibly also “speed limit effects” associated with rapid growth resulting from recent supply and demand shocks. Any further interpretation would seem premature at this stage, although the development of long-term unemployment (to about one third of the total) could reflect some asymmetric adverse hysteresis effects.
47. Tentative estimates of the NAWRU (non-accelerating wage rate of unemployment) were derived from the simple approach proposed by Elmeskov (1993). This method essentially assumes that changes in wage inflation are proportional to the gaps between actual unemployment and the NAWRU:
where W is an index of nominal wages.
48. With the additional assumption that the NAWRU does not significantly change from one year to another, the NAWRU can thus be simply approximated by:
and the resulting variable is then smoothed with the HP filter after outliers are eliminated by interpolation.18
49. As expected, the results display a marked increase in the estimated NAIRU and suggest that its present level would stand at 3.9 percent, a figure well within the range of 3.5 to 4.5 percent considered by the National Economic Institute (NEI). This result is also consistent with a trend Okun curve derived similarly, by substituting a measure of the output gap to the year–to–year change in wage inflation in equation (2). With a decline from 5.0 percent in 1995 to 4.1 percent of the labor force on average last year, actual unemployment would thus be close to the NAIRU, a feature consistent with emerging labor shortages in some sectors (such as computer experts) and recent high wage claims put ahead of the present general wage negotiations.
Approach to estimating potential output
50. The approach used draws from Giorno et al (1995) with additional recourse to simple economic assumptions when necessary. Since it is subject to well–known large supply shocks, the fish sector (fish primary and fish processing industry) is assumed to remain at its potential level of production, and we focus on the non-fish business sector, for which we postulate a standard two–factor Cobb–Douglas production function:
where:
Y = non-fish business sector value added
L = actual non-fish business sector labor input
K = actual non-fish sector capital stock (beginning of period)
tfp = trend total factor productivity (log index)
util = intensity of use of capital and employed labor (log index)
α = elasticity of output with respect to labor
b = elasticity of output with respect to capacity utilization
c = a constant
e = a random shock.
Total labor input is obtained as:
with:
Pwa = working age population
PR = participation rate
U = unemployment rate
EFG = actual employment in fish and government sectors
H = average working hours
51. In addition, as there is no reliable statistical indicator available for capacity utilization (util), we assume in what follows that it depends on the deviation of output from a “normal” level, and therefore (somewhat unconventionally) use as a proxy the deviation of actual output from trend (in logs) as measured by the HP filter in section B.
52. Total factor productivity (TFP) and parameter b (the elasticity of output with respect to capacity utilization) are jointly estimated by an iterative procedure. Assuming that parameter α is well approximated by labor’s share in value added, the respective contributions of labor and capital to output are computed and subtracted from the observed (log) GDP, the residual being denoted resd0. A first estimate of parameter b is then obtained by regressing this residual on a constant, a linear time trend and the capacity utilization variable:
53. The component of resd0 which is not explained by changes in capacity utilization provides the first approximation of the log index of total factor productivity:
54. This approximation is then smoothed with the HP filter and substituted for the constant and linear time trend component r + v t in equation (6). This model is then re-estimated, yielding a new estimate of b, and therefore a new estimate of trend total factor productivity tfp, which is again smoothed, etc. Final estimates for parameter b and for trend total factor productivity are thus obtained when the procedure finally converges.
Potential output is then computed as:
where util* stands for the “normal” degree of capital utilization (equal to zero with the log index used) and L* is calculated as:
with PR* and H* as estimates of the trend participation rate and of the average working hours, and NAWRU the estimate of the non-accelerating wage rate of unemployment.
Empirical estimates
55. The variables used for factor inputs are from the NEI. Chart 5 highlights in particular the sharpness of fluctuations in the participation rate and in the indicator for the number of hours, which makes estimates of actual and trend labor input very uncertain. Chart 6 also illustrates how the last downturn, with the accompanying large drop in investment, led to a sustained slowdown in business productive capital.
ICELAND: CAPITAL INPUT AND TOTAL FACTOR PRODUCTIVITY
(Structural method)
Citation: IMF Staff Country Reports 1997, 015; 10.5089/9781451819205.002.A002
Sources: National Economic Institute; IMF, International Financial Statistics; and staff calculations.56. Implemented over 1973-1996 with an assumed constant labor share in income α = 0.66, the sequential procedure quickly converges to a fixed value for parameter b equal to 0.61.19 Trend labor input L* for the non-fishing business sector is then obtained through equation (9) using estimates of the trend participation rate and the number of hours worked, estimates of the NAWRU, and data on actual employment in the fish and government sectors (Chart 5).20
57. This approach delivers the following main results. First, the results appear very uncertain since unexplained residuals are unusually high (see bottom part of Chart 6), underscoring the poor fit of our simple production function for the business non-fish sector. Estimates of trend labor input are also very uncertain. Second, the measures obtained for potential output and the output gap are very similar to those obtained with the HP filter and the Coe and McDermott nonparametric method, suggesting that potential growth was remarkably poor since the late 1980s and that the output gap was closed in 1996 (Table 1 and Chart 5), with actual output exceeding its potential by about 1 percent.21 Third, the low potential growth estimated for the recent period would reflect a combination of different factors: a notable rise of the NAIRU as documented above—possibly supplemented by a modest trend decrease in the number of hours worked; weak capital accumulation by the business non-fish sector as a consequence of the sharp contraction in productive investment during the last downturn; and apparently also, very limited total factor productivity growth, estimated under 1 percent per year from the mid-eighties until the latest years (Table 2). This seems low both historically and in comparison to other industrialized countries over the period (Giorno et al., 1995) and would reflect a slowdown in the remarkable post–war growth catch–up process.
Iceland: Alternative estimates of potential growth and output gaps
In percent of estimated trend GDP.
Estimates from the Hodrick-Prescott filter based on a detrending parameter equal to 25 and staff medium-term projections.
Estimates based on the nonparametric method of Coe and McDermott (1996) with staff medium-term projections.
Estimates from structural approach (see section C)
Staff projections.
Iceland: Alternative estimates of potential growth and output gaps
| Annual growth rate | Output gaps 1/ | ||||||
|---|---|---|---|---|---|---|---|
| Actual GDP | Trend GDP by HP filter 2/ |
Trend GDP by Nonparam. 3/ |
Potential GDP 4/ |
HP filter 2/ | Trend GDP by Nonparam. 3/ |
Potential GDP 4/ |
|
| 1987 | 8.6 | 2.8 | 3.8 | 3.6 | 5.4 | 4.3 | 4.9 |
| 1988 | -0.1 | 2.2 | 2.9 | 3.8 | 3.0 | 1.3 | 1.0 |
| 1989 | 0.3 | 1.7 | 1.7 | 1.1 | 1.5 | -0.1 | 0.2 |
| 1990 | 1.1 | 1.3 | 0.4 | 1.4 | 1.3 | 0.5 | -0.1 |
| 1991 | 1.3 | 1.1 | 0.1 | 0.3 | 1.5 | 1.7 | 0.9 |
| 1992 | -3.4 | 1.1 | 0.4 | 1.6 | -3.0 | -2.1 | -4.0 |
| 1993 | 0.9 | 1.4 | 1.0 | 1.6 | -3.5 | -2.2 | -4.7 |
| 1994 | 3.6 | 1.8 | 1.9 | 1.5 | -1.9 | -0.6 | -2.7 |
| 1995 | 2.1 | 2.2 | 2.9 | 1.1 | -2.0 | -1.3 | -1.7 |
| 1996 | 5.5 | 2.5 | 3.3 | 2.6 | 0.8 | 0.8 | 1.0 |
| 1997 5/ | 3.0 | 2.7 | 3.1 | … | 1.1 | 0.7 | … |
In percent of estimated trend GDP.
Estimates from the Hodrick-Prescott filter based on a detrending parameter equal to 25 and staff medium-term projections.
Estimates based on the nonparametric method of Coe and McDermott (1996) with staff medium-term projections.
Estimates from structural approach (see section C)
Staff projections.
Iceland: Alternative estimates of potential growth and output gaps
| Annual growth rate | Output gaps 1/ | ||||||
|---|---|---|---|---|---|---|---|
| Actual GDP | Trend GDP by HP filter 2/ |
Trend GDP by Nonparam. 3/ |
Potential GDP 4/ |
HP filter 2/ | Trend GDP by Nonparam. 3/ |
Potential GDP 4/ |
|
| 1987 | 8.6 | 2.8 | 3.8 | 3.6 | 5.4 | 4.3 | 4.9 |
| 1988 | -0.1 | 2.2 | 2.9 | 3.8 | 3.0 | 1.3 | 1.0 |
| 1989 | 0.3 | 1.7 | 1.7 | 1.1 | 1.5 | -0.1 | 0.2 |
| 1990 | 1.1 | 1.3 | 0.4 | 1.4 | 1.3 | 0.5 | -0.1 |
| 1991 | 1.3 | 1.1 | 0.1 | 0.3 | 1.5 | 1.7 | 0.9 |
| 1992 | -3.4 | 1.1 | 0.4 | 1.6 | -3.0 | -2.1 | -4.0 |
| 1993 | 0.9 | 1.4 | 1.0 | 1.6 | -3.5 | -2.2 | -4.7 |
| 1994 | 3.6 | 1.8 | 1.9 | 1.5 | -1.9 | -0.6 | -2.7 |
| 1995 | 2.1 | 2.2 | 2.9 | 1.1 | -2.0 | -1.3 | -1.7 |
| 1996 | 5.5 | 2.5 | 3.3 | 2.6 | 0.8 | 0.8 | 1.0 |
| 1997 5/ | 3.0 | 2.7 | 3.1 | … | 1.1 | 0.7 | … |
In percent of estimated trend GDP.
Estimates from the Hodrick-Prescott filter based on a detrending parameter equal to 25 and staff medium-term projections.
Estimates based on the nonparametric method of Coe and McDermott (1996) with staff medium-term projections.
Estimates from structural approach (see section C)
Staff projections.
Iceland: Contributions to potential GDP growth and output gaps in structural approach
See section C for a presentation of the method.
Estimated trend total factor productivity.
Iceland: Contributions to potential GDP growth and output gaps in structural approach
| Potential growth | Ouput gap | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Total economy |
Non-fish business sector |
Contribution 1/ | Total economy |
Non-fish business sector |
Contribution 1/ | |||||
| Labor | Capital | Trend 2/ | Labor | Capital | Residual | |||||
| 1987 | 3.6 | 0.8 | -0.6 | 0.8 | 0.6 | 4.9 | 7.2 | 6.1 | 3.2 | -2.4 |
| 1988 | 3.8 | 4.1 | 2.1 | 1.3 | 0.6 | 1.0 | 1.4 | 0.1 | 1.8 | -0.6 |
| 1989 | 1.1 | 1.8 | 0.0 | 1.1 | 0.7 | 0.2 | 0.3 | -0.8 | 0.9 | 0.1 |
| 1990 | 1.4 | 0.5 | -1.1 | 0.8 | 0.7 | -0.1 | -0.2 | -1.6 | 0.8 | 0.6 |
| 1991 | 0.3 | 1.4 | -0.4 | 1.0 | 0.8 | 0.9 | 1.3 | -0.4 | 0.9 | 0.7 |
| 1992 | 1.6 | 2.2 | 0.3 | 1.1 | 0.9 | -4.0 | -5.8 | -2.9 | -1.9 | -1.2 |
| 1993 | 1.6 | 0.5 | -0.8 | 0.4 | 0.9 | -4.7 | -6.7 | -2.9 | -2.2 | -1.9 |
| 1994 | 1.5 | 1.1 | -0.2 | 0.3 | 1.0 | -2.7 | -3.9 | -2.4 | -1.1 | -0.5 |
| 1995 | 1.1 | 1.1 | -0.2 | 0.2 | 1.1 | -1.7 | -2.5 | -1.1 | -1.2 | -0.3 |
| 1996 | 2.6 | 2.5 | 0.9 | 0.3 | 1.2 | 1.0 | 1.5 | 0.5 | 0.5 | 0.4 |
See section C for a presentation of the method.
Estimated trend total factor productivity.
Iceland: Contributions to potential GDP growth and output gaps in structural approach
| Potential growth | Ouput gap | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Total economy |
Non-fish business sector |
Contribution 1/ | Total economy |
Non-fish business sector |
Contribution 1/ | |||||
| Labor | Capital | Trend 2/ | Labor | Capital | Residual | |||||
| 1987 | 3.6 | 0.8 | -0.6 | 0.8 | 0.6 | 4.9 | 7.2 | 6.1 | 3.2 | -2.4 |
| 1988 | 3.8 | 4.1 | 2.1 | 1.3 | 0.6 | 1.0 | 1.4 | 0.1 | 1.8 | -0.6 |
| 1989 | 1.1 | 1.8 | 0.0 | 1.1 | 0.7 | 0.2 | 0.3 | -0.8 | 0.9 | 0.1 |
| 1990 | 1.4 | 0.5 | -1.1 | 0.8 | 0.7 | -0.1 | -0.2 | -1.6 | 0.8 | 0.6 |
| 1991 | 0.3 | 1.4 | -0.4 | 1.0 | 0.8 | 0.9 | 1.3 | -0.4 | 0.9 | 0.7 |
| 1992 | 1.6 | 2.2 | 0.3 | 1.1 | 0.9 | -4.0 | -5.8 | -2.9 | -1.9 | -1.2 |
| 1993 | 1.6 | 0.5 | -0.8 | 0.4 | 0.9 | -4.7 | -6.7 | -2.9 | -2.2 | -1.9 |
| 1994 | 1.5 | 1.1 | -0.2 | 0.3 | 1.0 | -2.7 | -3.9 | -2.4 | -1.1 | -0.5 |
| 1995 | 1.1 | 1.1 | -0.2 | 0.2 | 1.1 | -1.7 | -2.5 | -1.1 | -1.2 | -0.3 |
| 1996 | 2.6 | 2.5 | 0.9 | 0.3 | 1.2 | 1.0 | 1.5 | 0.5 | 0.5 | 0.4 |
See section C for a presentation of the method.
Estimated trend total factor productivity.
58. As for the present economic outlook, our estimates also suggest that the (negative) 1 percent output put gap estimated for 1996 would partly reflect higher–than–normal utilization of both capital and labor. This result should however be taken with great caution, given the size of the unexplained residual and the other large sources of uncertainty mentioned above. In addition, the strong GDP growth posted in 1996 reflected in part a surge in investment (24 percent) particularly due to the expansion of the existing aluminum smelter, which contributed to increasing Iceland’s production capacity significantly for the years ahead.
D. Output Gap and Inflation
59. For most industrialized countries, the cyclical position of activity is believed to be a good guide to identifying inflationary pressures, although there is ample evidence that the link between measures of the output gap and inflation is unstable, as a consequence of the large uncertainty surrounding these estimates and the variable lag structure of the relationship. Besides, for small open economies whose domestic prices are also largely determined by world market prices, departure from internal balance is likely to be reflected at least partly in changes in business profits and the external position rather than in actual inflation.
60. As underscored in previous sections, such factors are particularly pronounced in Iceland due to its large exposure to supply shocks and recent profound transformations of the economy. Moreover, the high inflation history was the product of excessively accommodative policies, together with a structural adaptation stemming from pervasive financial and wage indexation, and cyclical fluctuations were clearly not a major determinant of inflation in recent decades (Chart 7). For these reasons, past observations are not expected to be necessarily informative on the relationship between cyclical output fluctuations and inflation which may prevail at the present juncture.
ICELAND: OUTPUT GAP AND INFLATION
Citation: IMF Staff Country Reports 1997, 015; 10.5089/9781451819205.002.A002
Sources: National Economic Institute; Central Bank of Iceland; and staff calculations.1/ Based on the main 11 industrial trading partners with the official basket weighting scheme updated in September 1996.61. The following empirical evidence however suggests that cyclical variations might have had, if not a predominant, at least a significant influence in the formation of inflation in past decades. Three simple models relating consumer prices to the output gap, together with developments in the exchange rate and broad money—taken in turn individually or jointly—were estimated, along the lines of Coe and McDermott (1996), as encompassed by:
where cpi, gap, exch and mon are, respectively: the index of consumer price, the deviation of actual output from its trend, as estimated from the HP filter (λ = 25), an index of the nominal effective exchange rate and an index from broad money M3 (all in log).22 23
62. The estimations were carried out on annual data for two different periods, 1965-1996 and 1970-1996, in order to identify possible parameter instability due to the earlier years.24 The results, obtained after dropping statistically insignificant lagged terms for the explanatory variables, are presented in Table 3. The three different specifications used however raise potential problems of interpretation that can also lead to serious econometric biases, as they are not firmly based on well–founded theoretical models. Model (i) (with the exchange rate alone) would identify domestic and foreign sources of inflation as resulting from cyclical variations on one hand, and monetary developments purely through the exchange rate on the other hand. In model (ii), inflation is assumed to result from a combination of cyclical demand tensions and monetary factors purely as reflected in the path of the broad monetary aggregate. Model (iii) finally postulates that, in addition to the output gap, developments in the exchange rate and broad money are both relevant to assess inflationary pressures in the economy.
Iceland. Simple Regression Estimates for Inflation
(Dependent variable: Δcpi, 1/)
See section D for the notations. The figures in parenthesis represent the standard Student coefficients.
Regressions with instrumental variables when indicated, and based on OLS otherwise.
Durbin-Watson statistics for residual autocorrelation.
Iceland. Simple Regression Estimates for Inflation
(Dependent variable: Δcpi, 1/)
| model | Δcpi(-1) | gap(-1) | gap(-2) | Δexch | ΔM3 | Constant | Instruments 2/ | R2 | DW3/ |
|---|---|---|---|---|---|---|---|---|---|
| (Period of estimation: 1965-1996) | |||||||||
| (i) | 0.47 | 0.80 | -0.78 | 0.58 | … | 0.013 | no | 0.87 | 2.09 |
| (5.3) | (2.2) | (2.0) | (6.7) | (0.7) | |||||
| (ii) | 0.14 | -0.18 | 0.84 | … | 0.81 | -0.023 | no | 0.84 | 1.88 |
| (0.9) | (0.4) | (2.0) | (5.7) | (1.0) | |||||
| (iii) | 0.20 | 0.27 | -0.07 | 0.40 | 0.45 | -0.012 | no | 0.91 | 1.75 |
| (1.7) | (0.8) | (0.2) | (4.2) | (3.3) | (0.7) | ||||
| (i) | 0.70 | 0.73 | -0.22 | 0.19 | … | 0.026 | Δexch(-1),Δexch(-2) | 2.13 | |
| (1.4) | (15) | (0.2) | (0.2) | (0.7) | |||||
| (ii) | 0.18 | -0.13 | 0.79 | … | 0.76 | -0.019 | ΔM3(-1),ΔM3(-2) | 1.90 | |
| (0.7) | (0.3) | (1.6) | (2.8) | (0.7) | |||||
| (iii) | 0.40 | 0.54 | -0.32 | 0.40 | 0.21 | 0.005 | Δexch(-1), Δexch(-2) | 1.95 | |
| (1.6) | (0.9) | (0.4) | (1.7) | (0.5) | (0.2) | ΔM3(-1),ΔM3(-2) | |||
| (Period of estimation: 1970-1996) | |||||||||
| (i) | 0.47 | 1.12 | -0.68 | 0.54 | … | 0.028 | no | 0.88 | 2.07 |
| (4.1) | (2.0) | (1.3) | (4.9) | (1.2) | |||||
| (ii) | 0.17 | 0.71 | 0.52 | … | 0.73 | -0.003 | no | 0.89 | 1.62 |
| (1.2) | (1.3) | (0.9) | (5.5) | (0.1) | |||||
| (iii) | 0.15 | 0.50 | 0.20 | 0.33 | 0.50 | 0.004 | no | 0.93 | 1.57 |
| (1.2) | (1.1) | (0.4) | (3.3) | (3.9) | (0.2) | ||||
| (i) | 0.69 | 1.76 | -0.78 | 0.22 | … | 0.032 | Δexch(-1),Δexch(-2) | 2.22 | |
| (1.7) | (1.3) | (1.2) | (0.4) | (1.1) | |||||
| (ii) | 0.57 | 1.59 | -0.29 | … | 0.30 | 0.020 | ΔM3(-1),ΔM3(-2) | 2.08 | |
| (1.7) | (1.7) | (0.3) | (0.9) | (0.6) | |||||
| (iii) | 0.56 | 1.47 | -0.55 | 0.25 | 0.12 | 0.026 | Δexch(-1),Δexch(-2) | 2.09 | |
| (1.9) | (1.8) | (0.6) | (0.9) | (0.3) | (0.8) | ΔM3(-1),ΔM3(-2) | |||
See section D for the notations. The figures in parenthesis represent the standard Student coefficients.
Regressions with instrumental variables when indicated, and based on OLS otherwise.
Durbin-Watson statistics for residual autocorrelation.
Iceland. Simple Regression Estimates for Inflation
(Dependent variable: Δcpi, 1/)
| model | Δcpi(-1) | gap(-1) | gap(-2) | Δexch | ΔM3 | Constant | Instruments 2/ | R2 | DW3/ |
|---|---|---|---|---|---|---|---|---|---|
| (Period of estimation: 1965-1996) | |||||||||
| (i) | 0.47 | 0.80 | -0.78 | 0.58 | … | 0.013 | no | 0.87 | 2.09 |
| (5.3) | (2.2) | (2.0) | (6.7) | (0.7) | |||||
| (ii) | 0.14 | -0.18 | 0.84 | … | 0.81 | -0.023 | no | 0.84 | 1.88 |
| (0.9) | (0.4) | (2.0) | (5.7) | (1.0) | |||||
| (iii) | 0.20 | 0.27 | -0.07 | 0.40 | 0.45 | -0.012 | no | 0.91 | 1.75 |
| (1.7) | (0.8) | (0.2) | (4.2) | (3.3) | (0.7) | ||||
| (i) | 0.70 | 0.73 | -0.22 | 0.19 | … | 0.026 | Δexch(-1),Δexch(-2) | 2.13 | |
| (1.4) | (15) | (0.2) | (0.2) | (0.7) | |||||
| (ii) | 0.18 | -0.13 | 0.79 | … | 0.76 | -0.019 | ΔM3(-1),ΔM3(-2) | 1.90 | |
| (0.7) | (0.3) | (1.6) | (2.8) | (0.7) | |||||
| (iii) | 0.40 | 0.54 | -0.32 | 0.40 | 0.21 | 0.005 | Δexch(-1), Δexch(-2) | 1.95 | |
| (1.6) | (0.9) | (0.4) | (1.7) | (0.5) | (0.2) | ΔM3(-1),ΔM3(-2) | |||
| (Period of estimation: 1970-1996) | |||||||||
| (i) | 0.47 | 1.12 | -0.68 | 0.54 | … | 0.028 | no | 0.88 | 2.07 |
| (4.1) | (2.0) | (1.3) | (4.9) | (1.2) | |||||
| (ii) | 0.17 | 0.71 | 0.52 | … | 0.73 | -0.003 | no | 0.89 | 1.62 |
| (1.2) | (1.3) | (0.9) | (5.5) | (0.1) | |||||
| (iii) | 0.15 | 0.50 | 0.20 | 0.33 | 0.50 | 0.004 | no | 0.93 | 1.57 |
| (1.2) | (1.1) | (0.4) | (3.3) | (3.9) | (0.2) | ||||
| (i) | 0.69 | 1.76 | -0.78 | 0.22 | … | 0.032 | Δexch(-1),Δexch(-2) | 2.22 | |
| (1.7) | (1.3) | (1.2) | (0.4) | (1.1) | |||||
| (ii) | 0.57 | 1.59 | -0.29 | … | 0.30 | 0.020 | ΔM3(-1),ΔM3(-2) | 2.08 | |
| (1.7) | (1.7) | (0.3) | (0.9) | (0.6) | |||||
| (iii) | 0.56 | 1.47 | -0.55 | 0.25 | 0.12 | 0.026 | Δexch(-1),Δexch(-2) | 2.09 | |
| (1.9) | (1.8) | (0.6) | (0.9) | (0.3) | (0.8) | ΔM3(-1),ΔM3(-2) | |||
See section D for the notations. The figures in parenthesis represent the standard Student coefficients.
Regressions with instrumental variables when indicated, and based on OLS otherwise.
Durbin-Watson statistics for residual autocorrelation.
63. The econometric estimation of these models is thus complicated by potential biases due to simultaneity and exogeneity factors. In particular, the exchange rate and broad money can both reasonably be expected to be determined simultaneously with consumer prices, as a simple consequence of relative purchasing power parity for the former and through standard money demand for the latter. Furthermore, estimates of model (ii) and model (iii) could even further mistake actual features of money demand and exchange rate equations, as model (iii) contains in particular some of the key ingredients of the basic monetary model with flexible prices, for the assumed determination of consumer prices. In order to take into account possible simultaneity, the three models were both estimated by standard OLS, but also with instrumental variables for the current changes in the exchange rate and in broad money. In practice, while the former set of econometric estimates are potentially biased, the accuracy of the second set is limited by the difficulty of finding satisfactory instruments for these variables.
64. Based on their estimation over 1970-1996 with instrumental variables, models (ii) and (iii) suggest a significant effect of the level of the output gap on future inflation, as (other things being equal) a 1 percent point increase in the output gap would translate into a 1½ percentage point increase in inflation in the following year (see bottom part of Table 3). Several features, however, tend to strongly question the relevance of these estimates. First, while estimates from OLS appear very different, the instruments used for these estimations appear quite poor and the fact that no statistically significant effect of current changes in the exchange rate or in broad money is found, somewhat at odds with expectations. Second, as noted below, exogeneity problems are potentially large for these two models and the interpretation is further complicated by apparent instability in the estimates.
65. The results obtained with model (i) are perhaps more interesting. The estimates based on standard OLS would suggest a significant effect of the output gap on future inflation with speed limit effects”, as they indicate that changes in the output gap, as distinct from its level, affects future inflation (particularly for the longer period 1965-1996). Other things being equal, a strengthening of the cyclical position by 1 percentage of GDP would contribute to an increase of about ¾ to 1 percentage point in consumer price inflation for the following year. As this feature does not clearly emerge from the estimations using instrumental variables, there is a significant possibility that this reflects econometric biases due to simultaneity in the formation of consumer prices and the exchange rate. However, the instruments used here (lagged changes in the exchange rate) are too poor to bring this issue to a firm conclusion. All in all, while further investigation is needed, our tentative results suggest that “speed-limit” effects may have played a significant albeit limited role in the formation of inflation in recent decades.
E. Concluding Remarks
66. This paper has presented estimates of trend output and the output gap obtained with different detrending techniques and a more standard structural approach. Iceland is well known to be subject to large and frequent supply-shocks from the fish sector, and its high inflation history largely reflects overly accommodative macroeconomic policies. However, recent profound structural transformations in the economy and the move to a more stable macroeconomic environment suggest an increased role for cyclical fluctuations in the determination of inflation.
67. In this context, the main conclusions of the paper are threefold. First, the paper highlights the large uncertainty surrounding estimates of potential output and the output gap, with estimates for the more recent period inherently depending on our assessment of the medium-term outlook. But the paper nonetheless shows that different reasonable approaches provide very similar results and confirm, in particular, the fairly modest expansion of potential output in recent years. This could reflect a notable rise in the non–accelerating rate of unemployment and some other factors, including low capital accumulation and limited factor productivity growth in the non-fish business sector, together with recent adverse supply developments in the fishing sector. This underscores the importance of maintaining a stable macroeconomic environment and pursuing further structural reforms, in the financial sector in particular, in order to enable a sustainable rise in business investment and ensure that financial resources are appropriately channeled to the most efficient sectors and projects.
68. Second, we have found some tentative evidence consistent with “speed-limit” effects in the formation of inflation, suggesting that fluctuations in the cyclical position of the economy, as distinct from its level, could have had a significant albeit limited effect on inflation.
69. Third and finally, our results show that the present economic upturn, with the hefty growth of 1996, has brought the economy close to and likely somewhat above its potential capacity. While the present inflation conditions remain favorable, this raises the risk that inflationary pressures, together with a further deterioration of the current account, could start to develop should activity fail to slow down as expected.
REFERENCES
Barell, R., and J. Sefton, 1995, “Output Gaps. Some Evidence from the U.K., France and Germany”, NIESR Review, No. 151.
Beveridge, S., and C.R. Nelson, 1981, “A New Approach to Decomposition of Economic Time Series into Permanent and Transitory Components with Particular Attention to Measurement of the Business cycle”, Journal of Monetary Economics, Vol. 7.
Canova F., 1993, “Detrending and Business Cycle Facts”, C.E.P.R. Discussion Paper No. 782, (London: Center of Economic Policy Research).
Coe D.T, and C.J. McDermott, 1996, “Does the Gap Model Work in Asia”, IMF Working Paper No 96/69, (Washington: International Monetary Fund).
Danthine, J.P., and M. Girardin, 1989, “Business Cycle in Switzerland”, European Economic Review, Vol. 33.
Elmeskov J., 1993, “High and Persistent Unemployment: Assessment of the Problem and its Causes”, OECD Economics Department Working Paper No 132, (Paris: Organization for Economic Cooperation and Development).
Evans, G., and L. Reichlin, 1994, “Information, Forecasts, and Measurement of the Business Cycle”, Journal of Monetary Economics, Vol. 33.
Giorno, C., P. Richardson, D. Roseveare, and P van der Noord, 1995, “Estimating Potential Output, Output Gaps and Structural Budget Balances”, OECD Economic Department Working Paper No. 152, (Paris: Organization for Economic Cooperation and Development).
Gudmundsson B., 1995, “Note on Inflation and Unemployment in Iceland”, NEI mimeo, (Reykjavik: National Economic Institute).
Harvey, A.C., and A. Jaeger, 1993, “Detrending, Stylized Facts and the Business Cycle”, Journal of Applied Econometrics, Vol. 8.
Staiger D., Stock J.H., and M.W. Watson, 1996, “How Precise Are Estimates of the Natural Rate of Unemployment”, NBER Working Paper No 5477, (Cambridge: National Bureau of Economic Research).
Prepared by Antoine Magnier.
In this chapter, we use the terms “trend” and “potential” output without any distinction.
Abstracting from the (now likely) construction of a second aluminum smelter in 1997-1998, the staff projections envisage real GDP growth moderating to 3.0 percent and 2.4 percent in 1996 and 1997, and 2.3 in the later years, largely in line with the official NEI projections.
See, for instance, Staiger, Stock and Watson (1996) for a recent study on the precision of conventional and unconventional econometric estimates of the NAIRU in the United States.
The HP filter was used with a detrending parameter λ = 25 and staff medium-term projections for unemployment and wages.
National account estimates indicate large swings in the labor share of income. A standard value of 2/3 is however in line with recent observations of the ratio of the remuneration of employees to gross factor income for the total economy.
All trend variables are obtained from the HP filter, with a detrending parameter λ equal to 25 and staff medium-term projections.
Given the method’s extensive use of the HP filter and the proxy used for capital utilization, this similarity in the results should however not come as a full surprise.
The following results are not significantly affected by substituting estimates of the output gap from the nonparametric approach.
The effective exchange rate indicator is based on Iceland’s 11 main trading partners, with the weighting scheme of the official index basket revised in September 1996.
There are also indications that the official exchange rate of the krona was allowed to depart significantly from its underlying market value until the late sixties, affecting thus the relevance of the nominal effective exchange rate indicator used for this period.
