Abstract
This Selected Issues paper and Statistical Appendix assesses Switzerland’s recent real GDP performance in terms of underlying movements in potential output and the cyclical output gap. The paper highlights that Swiss real GDP has been stagnant since 1990, after expanding at an average rate of some 1¾ percent during 1977–90. The evidence presented indicates that potential output growth during 1991–95 was significantly below historical average. This paper also tries to assess the possible effects of stage 3 of European Monetary Union on Switzerland.
I. Estimates of Potential Output Growth and the Cyclical Output Gap1
This chapter assesses Switzerland’s recent real GDP performance in terms of underlying movements in potential output and the cyclical output gap. The level of Swiss real GDP has stagnated since 1990, after expanding at an average growth rate of 1¾ percent during the period 1977-90. The protracted slump in real GDP raises three questions. One, to what extent does the stagnation of real GDP since 1990 reflect a slowdown in potential output growth? Two, what is the current amount of economic slack (or output gap) in the economy? And three, given the assessment of recent potential output growth and the current output gap, what are Switzerland’s medium-term growth prospects?
Although any decomposition of observed output movements into trend (or potential) and cyclical (or gap) components is fraught with considerable uncertainty, even approximate estimates are useful in evaluating the stance of macroeconomic policies. The Swiss authorities’ monetary framework targets an expansion of the monetary base consistent with potential output growth and a medium-term inflation target of about 1 percent. This framework necessitates the estimation and projection of potential output growth, which is currently estimated at about 2 percent by the Swiss National Bank (SNB) (see Lüscher and Ruoss (1996)). As indicated further below, assuming potential output growth during 1991-95 of some 2 percent would yield a large output gap of about 6½ percent of potential GDP in 1996. As regards fiscal policy, the authorities assess the size of the structural deficit at the federal level, with a view to gauge the additional fiscal consolidation efforts needed to balance the federal budget by 2001. The Federal Finance Administration (FFA) has estimated that almost the entire federal fiscal deficit of about 1½ percent of GDP in 1996 was structural, reflecting an estimate of a cyclical output gap close to zero in 1996. As is shown later, these estimates reflect the range of output gaps associated with different methodological approaches and detrending choices.
The balance of evidence presented in this chapter suggests that potential output growth in Switzerland has slowed significantly since 1990, amounting to about 1 percent in the period 1991-95 compared with an average potential growth of 1¾ percent during 1977-90. The recent slowdown in potential output growth reflects a marked slowdown in the growth rates of potential labor input and potential total factor productivity. The size of the output gap in 1996 is estimated at about 3½ percent of potential GDP. This estimate is broadly supported by evidence from other macroeconomic indicators including inflation, capacity utilization in manufacturing, and private consumption. However, output gap estimates are particularly sensitive to assumptions about the recent trend behavior of total factor productivity. In particular, assuming that potential total factor productivity growth during 1991-96 had remained unchanged from its average growth during 1977-90 would raise the output gap in 1996 to 5¼ percent of potential GDP. Regarding prospects for medium-term growth, potential output growth is projected to increase slightly to 1¼-1½ percent. In that case, actual real GDP growth would need to average about 2 percent per annum during 1997-2002 to close the estimated output gap.
The remainder of this chapter is structured as follows: Section A contains a descriptive analysis of Switzerland’s GDP growth experience since 1948. Section B provides univariate decompositions of real GDP movements into trend and cyclical components based on an unobserved components (UC) model and compares the results with estimates obtained from deterministic detrending and Hodrick-Prescott filtering. Section C presents estimates of potential output and the output gap drawing on trend-cycle decompositions of aggregate production function inputs (labor, capital, and total factor productivity). And Section D evaluates information on the degree of economic slack provided by the time series behavior of private consumption and inflation.
A. Real GDP Facts
Real GDP growth in Switzerland has slowed markedly since 1975, both compared with Switzerland’s growth experience before 1975 and with real GDP growth in most other industrial countries. Chart I-1 shows the behavior of Swiss real GDP from 1948 to 1995 against the backdrop of real GDP movements in the United States and Germany (data for 1990-95 refer to west Germany).2 Until 1975, Swiss real GDP moved in close tandem with real GDP in the two reference countries; in fact, average real GDP growth rates and standard deviations of growth rates for the period 1961-74 were broadly similar. However, since 1975 Swiss output performance has deviated markedly from the two comparison countries. In particular, real GDP dropped sharply in 1975 (by 6.3 percent) and did not subsequently revert to its previous trend. Average real GDP growth during 1976-95 slowed significantly compared both with Switzerland’s previous growth experience and the growth experience of the U.S. and Germany (Table I-1).3
Switzerland: Real GDP, 1948–1995
Citation: IMF Staff Country Reports 1997, 018; 10.5089/9781451807141.002.A001
Sources: IMF, IFS database; Swiss Institute for Business Cycle Research, data tape.Switzerland: Real GDP, 1948–1995
Citation: IMF Staff Country Reports 1997, 018; 10.5089/9781451807141.002.A001
Sources: IMF, IFS database; Swiss Institute for Business Cycle Research, data tape.Switzerland: Real GDP, 1948–1995
Citation: IMF Staff Country Reports 1997, 018; 10.5089/9781451807141.002.A001
Sources: IMF, IFS database; Swiss Institute for Business Cycle Research, data tape.Switzerland: Real GDP Developments, 1960-95
1990-95 West Germany.
Long-run effect on the level of real GDP (in percent) of a 1 percentage point shock to real GDP, based on an estimated autoregressive model of order two for the period 1960-95.
Switzerland: Real GDP Developments, 1960-95
| Switzerland | Germany 1/ | United States | ||
|---|---|---|---|---|
| (In percent) | ||||
| Average real GDP growth | ||||
| 1961-95 | 2.3 | 2.9 | 2.9 | |
| 1961-74 | 4.2 | 4.0 | 3.6 | |
| 1976-95 | 1.5 | 2.4 | 2.7 | |
| Standard deviation of real GDP growth | ||||
| 1961-95 | 2.6 | 2.2 | 2.1 | |
| 1961-74 | 1.7 | 2.1 | 2.0 | |
| 1976-95 | 1.6 | 1.9 | 2.1 | |
| Persistence of real GDP shocks 2/ | 2.0 | 1.2 | 1.2 | |
| Memorandum items: | ||||
| Average per capita GDP growth | ||||
| 1961-95 | 1.5 | 2.4 | 1.9 | |
| 1961-74 | 2.8 | 3.2 | 2.4 | |
| 1976-95 | 1.0 | 2.0 | 1.7 | |
1990-95 West Germany.
Long-run effect on the level of real GDP (in percent) of a 1 percentage point shock to real GDP, based on an estimated autoregressive model of order two for the period 1960-95.
Switzerland: Real GDP Developments, 1960-95
| Switzerland | Germany 1/ | United States | ||
|---|---|---|---|---|
| (In percent) | ||||
| Average real GDP growth | ||||
| 1961-95 | 2.3 | 2.9 | 2.9 | |
| 1961-74 | 4.2 | 4.0 | 3.6 | |
| 1976-95 | 1.5 | 2.4 | 2.7 | |
| Standard deviation of real GDP growth | ||||
| 1961-95 | 2.6 | 2.2 | 2.1 | |
| 1961-74 | 1.7 | 2.1 | 2.0 | |
| 1976-95 | 1.6 | 1.9 | 2.1 | |
| Persistence of real GDP shocks 2/ | 2.0 | 1.2 | 1.2 | |
| Memorandum items: | ||||
| Average per capita GDP growth | ||||
| 1961-95 | 1.5 | 2.4 | 1.9 | |
| 1961-74 | 2.8 | 3.2 | 2.4 | |
| 1976-95 | 1.0 | 2.0 | 1.7 | |
1990-95 West Germany.
Long-run effect on the level of real GDP (in percent) of a 1 percentage point shock to real GDP, based on an estimated autoregressive model of order two for the period 1960-95.
The fluctuations of real GDP during the period 1960-95 appear to reflect the prevalence of highly persistent real GDP shocks, where persistence is defined as the long run response of the output level (in percentage points) to a 1 percentage point innovation in real GDP. Indeed, the persistence of real GDP shocks in Switzerland is estimated at almost double the persistence estimates for the United States and Germany (Table I-1). This is prima facie evidence that fluctuations of real GDP in Switzerland reflect at least some highly persistent real shocks. Persistent shocks would be reflected in movements in potential output, although high persistence estimates do not rule out the existence of significant cyclical output gaps.
Chart I-2 provides a decomposition of real GDP fluctuations during the period 1976-95 in terms of underlying aggregate production function inputs, i.e. labor input measured in hours, the capital stock, and total factor productivity (the sources and methodology for calculating the production function inputs are described in the data appendix).4 The series on the production function inputs appear to indicate that capital accumulation was the key driving force behind Swiss GDP growth since 1976, while labor input and total factor productivity growth were relatively sluggish during this time period.
Switzerland: Real GDP, Employment (Hours), Capital, and Total Factor Productivity, 1976–1995
Citation: IMF Staff Country Reports 1997, 018; 10.5089/9781451807141.002.A001
Sources: Swiss Institute for Business Cycle Research, data tape, and staff estimates.Switzerland: Real GDP, Employment (Hours), Capital, and Total Factor Productivity, 1976–1995
Citation: IMF Staff Country Reports 1997, 018; 10.5089/9781451807141.002.A001
Sources: Swiss Institute for Business Cycle Research, data tape, and staff estimates.Switzerland: Real GDP, Employment (Hours), Capital, and Total Factor Productivity, 1976–1995
Citation: IMF Staff Country Reports 1997, 018; 10.5089/9781451807141.002.A001
Sources: Swiss Institute for Business Cycle Research, data tape, and staff estimates.B. Univariate GDP Decompositions
To derive estimates of potential output growth and the cyclical output gap, the logarithm of observed real GDP (Y(t)) is decomposed into two unobservable components:
where YPOT(t) is potential output and YGAP(t) is the cyclical output gap component. The two unobserved components are defined in terms of their statistical properties: movements in the potential output component are assumed to reflect the effects of those GDP shocks that lead to permanent (i.e. persistent) changes in the level of GDP, while movements in the output gap are assumed to reflect the effects of those GDP shocks that lead to transitory (i.e. non-persistent) changes in the level of GDP. Statistical identification of the two unobserved components can be based on univariate models of real GDP (univariate decomposition) or by exploiting postulated links between one or both of the unobserved components and other macroeconomic variables (multivariate decomposition) including the inputs of an aggregate production function, unemployment, private consumption, or the inflation rate.
As a first approach, the two unobservable components in (1) were estimated by postulating a statistical model for the trend component of real GDP—a linear deterministic time trend—and by employing a linear filter procedure—the Hodrick-Prescott filter—and then the output gap was derived as the residual component between actual and estimated trend output. However, these detrending approaches, while simple and easy to implement, amount to a rather mechanical assignment of output fluctuations to the potential and cyclical components of real GDP. In particular, these approaches do not exploit all the information provided by the observed GDP time series. A useful alternative framework for achieving decomposition (1) is provided by the class of unobserved components (UC) models. In the context of the trend-cycle decomposition exemplified by (1), UC models specify statistical models for both the trend and the cyclical component (see Harvey (1989) and Harvey and Jaeger (1993) for details). In particular, potential output is modeled by a flexible linear trend:
where η(t) and ζ(t) denote shocks to the level and drift (average growth) rate of potential GDP, respectively. The cyclical component is modeled as a stationary autoregressive moving-average process with two autoregressive and one moving-average term (ARMA(2,1)):
where Φ(1), Φ(2), and θ are the parameters of the ARMA process, and ϵ(t) denotes the shock to the cyclical component (which, by assumption, has only a transitory effect on the level of real GDP). A number of popular detrending methods are special cases of the UC model. For example, restricting the variances of the permanent shocks η(t) and ζ(t) to zero is equivalent to assuming that potential output follows a linear deterministic trend. When only the variance of ζ(t) is restricted to zero, but the variance of η(t) is allowed to be positive, potential output is restricted to follow a random walk with constant drift. Finally, the results of applying a Hodrick-Prescott filter would be replicated by an UC model if the assumption of a fixed numerical ratio between the variance of changes in the potential output growth rate and the variance of the cyclical output gap would obtain in the GDP series.5
Estimates of potential output and the output gap based on the UC model (2) to (4) are set out in Chart 3 and Table 2.6 Estimating the unrestricted unobserved components model resulted in an acceptable fit, but diagnostic test statistics indicated that the large one-time level shock in 1975 is not well captured by the model, constituting an outlier. As a consequence, the model was augmented by incorporating a one-time level shock in 1975, which improved both the fit (as indicated by the significant increase in the model’s likelihood value) and the diagnostic test statistics turn out satisfactory. The estimates of the unresticted unobserved components model with a trend break in 1975 suggest a substantial slowdown in potential output growth during 1991-95 compared with the period 1985-90; average annual growth in potential output during 1991-95 was only 1 percent compared with 1 ¾ percent during 1977-90. Reflecting the sharp slowdown in potential output growth, the size of the cyclical output gap in 1995 amounted to about 1½ percent, widening to 3 percent of potential GDP in 1996 if real GDP contracts by ½ percent in 1996 as currently projected. However, using mechanical detrending based on deterministic extrapolation indicates significantly larger output gaps of 4½ and 6½ percent in 1995 and 1996, respectively. Finally, applying a Hodrick-Prescott filter to the annual time series results in estimates of output gap that are smaller—2½ percent of potential GDP in 1995 and a projected gap of almost 4 percent in 1996.7
Switzerland: Univariate Real GDP Decompositions, 1948–1995
Citation: IMF Staff Country Reports 1997, 018; 10.5089/9781451807141.002.A001
Sources: Swiss Institute for Business Cycle Research, data tape, and staff estimates.Switzerland: Univariate Real GDP Decompositions, 1948–1995
Citation: IMF Staff Country Reports 1997, 018; 10.5089/9781451807141.002.A001
Sources: Swiss Institute for Business Cycle Research, data tape, and staff estimates.Switzerland: Univariate Real GDP Decompositions, 1948–1995
Citation: IMF Staff Country Reports 1997, 018; 10.5089/9781451807141.002.A001
Sources: Swiss Institute for Business Cycle Research, data tape, and staff estimates.Unobserved Components (UC) Models of Real GDP, 1948-95
In percent.
In percent of potential GDP.
Based on staff projections assuming a contraction in real GDP of ½ percent in 1996.
Unobserved Components (UC) Models of Real GDP, 1948-95
| Unrestricted UC model | Unrestricted UC model with trend break in 1975 | Deterministic trend model with trend break in 1975 | Hodrick- Prescott filter | |
|---|---|---|---|---|
| Likelihood value | 140.4 | 144.0 | … | … |
| Average potential output growth: | ||||
| 1985-90 1/ | 2.2 | 2.1 | 1.8 | 1.9 |
| 1991-95 1/ | 0.6 | 0.8 | 1.8 | 1.2 |
| Standard deviation of output gap in 1948-95 1/ | 2.1 | 1.9 | 2.4 | 2.7 |
| Estimated output gaps implied: | ||||
| 1995 2/ | -0.9 | -1.5 | -4.3 | -2.4 |
| 1996 2/3/ | -2.5 | -3.0 | -6.5 | -3.9 |
In percent.
In percent of potential GDP.
Based on staff projections assuming a contraction in real GDP of ½ percent in 1996.
Unobserved Components (UC) Models of Real GDP, 1948-95
| Unrestricted UC model | Unrestricted UC model with trend break in 1975 | Deterministic trend model with trend break in 1975 | Hodrick- Prescott filter | |
|---|---|---|---|---|
| Likelihood value | 140.4 | 144.0 | … | … |
| Average potential output growth: | ||||
| 1985-90 1/ | 2.2 | 2.1 | 1.8 | 1.9 |
| 1991-95 1/ | 0.6 | 0.8 | 1.8 | 1.2 |
| Standard deviation of output gap in 1948-95 1/ | 2.1 | 1.9 | 2.4 | 2.7 |
| Estimated output gaps implied: | ||||
| 1995 2/ | -0.9 | -1.5 | -4.3 | -2.4 |
| 1996 2/3/ | -2.5 | -3.0 | -6.5 | -3.9 |
In percent.
In percent of potential GDP.
Based on staff projections assuming a contraction in real GDP of ½ percent in 1996.
Potential GDP Growth and Output Gap Estimates Based on Production Function, 1976-2002
Staff projections.
Potential GDP Growth and Output Gap Estimates Based on Production Function, 1976-2002
| A. Potential Output Growth | ||||
|---|---|---|---|---|
| Contributions to potential GDP growth by: | ||||
| Period | Potential GDP growth | Labor | Capital | Total factor productivity |
| 1977-90 | 1.8 | 0.4 | 0.7 | 0.7 |
| 1991-95 | 1.0 | 0.1 | 0.6 | 0.3 |
| 1996 1/ | 1.3 | 0.2 | 0.6 | 0.5 |
| 1997-02 1/ | 1.3 | 0.2 | 0.6 | 0.5 |
| B. GDP Gap | ||||
| Contributions to GDP gap by: | ||||
| Period | GDP gap | Labor | Capital | Total factor productivity |
| 1990 | 3.1 | 1.8 | 1.7 | -0.4 |
| 1991 | 1.1 | 1.5 | -- | -0.4 |
| 1992 | -0.2 | -0.1 | -0.8 | 0.7 |
| 1993 | -1.4 | -0.8 | -1.0 | 0.4 |
| 1994 | -0.8 | -0.9 | -- | 0.1 |
| 1995 | -1.6 | -1.3 | 0.5 | -0.8 |
| 1996 1/ | -3.5 | … | … | … |
Staff projections.
Potential GDP Growth and Output Gap Estimates Based on Production Function, 1976-2002
| A. Potential Output Growth | ||||
|---|---|---|---|---|
| Contributions to potential GDP growth by: | ||||
| Period | Potential GDP growth | Labor | Capital | Total factor productivity |
| 1977-90 | 1.8 | 0.4 | 0.7 | 0.7 |
| 1991-95 | 1.0 | 0.1 | 0.6 | 0.3 |
| 1996 1/ | 1.3 | 0.2 | 0.6 | 0.5 |
| 1997-02 1/ | 1.3 | 0.2 | 0.6 | 0.5 |
| B. GDP Gap | ||||
| Contributions to GDP gap by: | ||||
| Period | GDP gap | Labor | Capital | Total factor productivity |
| 1990 | 3.1 | 1.8 | 1.7 | -0.4 |
| 1991 | 1.1 | 1.5 | -- | -0.4 |
| 1992 | -0.2 | -0.1 | -0.8 | 0.7 |
| 1993 | -1.4 | -0.8 | -1.0 | 0.4 |
| 1994 | -0.8 | -0.9 | -- | 0.1 |
| 1995 | -1.6 | -1.3 | 0.5 | -0.8 |
| 1996 1/ | -3.5 | … | … | … |
Staff projections.
C. Production Function-Based Estimates of Output Gap
The production function-based approach for disentangling potential and cyclical output movements draws on estimated trend-cycle decompositions of underlying factor input. Following recent work of Lüscher and Ruoss (1996) and Marty (1996) on potential output estimates for Switzerland, a Cobb-Douglas production function is assumed, and thus real GDP can be written as:
where L(t) is effective labor input, K(t) is the effective stock of capital, and TFP(t) denotes total factor productivity, and a and β are the parameters of the Cobb-Douglas production function. The production function-based approach requires the calculation of decompositions of the three underlying inputs into potential and cyclical components. Accordingly, combining (1) and (5), potential output can be written as a linear combination of the potential components of the three production function inputs:
Similarly, the cyclical output gap is given by a linear combination of the cyclical gaps of the three production function inputs:
It is noteworthy that the production function-based approach effectively transforms the univariate detrending problem of Section B into a trivariate detrending problem, raising difficult measurement problems for the underlying factor input variables. On the other hand, tracing potential and cyclical output movements to movements in the underlying factor inputs adds economic content to the assessment of the cyclical position of the economy.
The determination of the labor input gap (LGAP(t)) is taken up first. By definition actual annual labor input in hours is given by:
where POPW(t) measures the resident working population aged 20-64, LFP(t) is the labor force participation rate, U(t) denotes the unemployment rate, and H(t) is a measure of average annual work hours per employed person. To estimate potential labor input, cyclical variations in the labor force participation and the unemployment rate need to be removed. Thus, the size of the resident working population and average annual work hours per employed person are assumed to be noncyclical. While the data series on average annual work hours appear to be consistent with this assumption, the data series on the resident working population (which includes foreign workers) appears to have been sensitive to the business cycle, although the data are not characterized by a clear cyclical pattern owing perhaps to changes over time in permit policies for foreign workers. On this, several observers, e.g. OECD (1996), have noted that the cyclical sensitivity of the labor force has declined since the mid-1970s.
As regards the unemployment rate, measurement of cyclical slack is made difficult by recent structural changes in the composition of the labor force (increased share of foreign residents with permanent work permits) and the reforms of the unemployment insurance system (increased coverage and, until more recently, a trend to more generous unemployment benefits). The Swiss unemployment rate—defined as the number of registered unemployed persons as a percent of the labor force—has risen drastically at the beginning of the 1990s (Chart I-4). For the purpose of this study, the rate of unemployment that abstracts from cyclical influences, the natural rate of unemployment, is taken from OECD (1996). The implied cyclical unemployment rate in 1995 is about 1 percent.
Switzerland: Cyclical Labor Market Indicators, 1976–1995
Citation: IMF Staff Country Reports 1997, 018; 10.5089/9781451807141.002.A001
Sources: Swiss Institute for Business Cycle Research, OECD Economic Survey 1994-95 for Switzerland, and staff estimates.Switzerland: Cyclical Labor Market Indicators, 1976–1995
Citation: IMF Staff Country Reports 1997, 018; 10.5089/9781451807141.002.A001
Sources: Swiss Institute for Business Cycle Research, OECD Economic Survey 1994-95 for Switzerland, and staff estimates.Switzerland: Cyclical Labor Market Indicators, 1976–1995
Citation: IMF Staff Country Reports 1997, 018; 10.5089/9781451807141.002.A001
Sources: Swiss Institute for Business Cycle Research, OECD Economic Survey 1994-95 for Switzerland, and staff estimates.Labor force participation rates in Switzerland also appear to exhibit cyclical fluctuations, reflected in variations of labor force participation rates of women and resident foreign workers. Moreover, variations in the number of seasonal and frontier workers—which are included in the labor force measure but not in the measure of the resident working population—arel also reflected in fluctuations of the participation rate. To account for the cyclical variation in the overall participation rate, a Hodrick-Prescott filter was applied to annual data for the period 1976-95 (the smoothing constant was set at 100). The resulting estimates suggest that the procyclial behavior of labor force participation rates add a considerable amount of cyclical slack to the labor market.8 Combining the estimates of the natural rate of unemployment and the HP-filter estimates of the “normal” labor force participation rate allows to estimate the level of potential labor input, and expressing the gap between actual and potential labor input as a percent of potential labor input yields the employment gap (Chart I-4).
Survey data on capacity utiliztion in manufacturing are used to calculate the level of potential capital. In particular, the average level of capacity utilization during 1976-95 (83.6 percent) is assumed to represent the normal level of capital stock utilization for the whole economy and the measured capital stock is adjusted for the variations of capacity utilization around the normal level to gauge the amount of effectively employed capital (Chart I-5). These assumptions imply that the “capital gap” is equivalent to the deviation of actual capacity utilization from its normal level.
Switzerland: Cyclical Fluctuations in Capital Utilization and Total Factor Productivity, 1976–1995
Citation: IMF Staff Country Reports 1997, 018; 10.5089/9781451807141.002.A001
Sources: Swiss Institute for Business Cycle Research, data tape, and staff estimates.Switzerland: Cyclical Fluctuations in Capital Utilization and Total Factor Productivity, 1976–1995
Citation: IMF Staff Country Reports 1997, 018; 10.5089/9781451807141.002.A001
Sources: Swiss Institute for Business Cycle Research, data tape, and staff estimates.Switzerland: Cyclical Fluctuations in Capital Utilization and Total Factor Productivity, 1976–1995
Citation: IMF Staff Country Reports 1997, 018; 10.5089/9781451807141.002.A001
Sources: Swiss Institute for Business Cycle Research, data tape, and staff estimates.Finally, the decomposition of total factor productivity growth into its trend and cyclical component is based on applying a Hodrick-Prescott filter to annual estimates of total factor productivity for the period 1976-95 (see the data appendix for the construction of the series for total factor productivity).
Combining the estimates of potential growth and cyclical gaps for the three production function inputs yields estimates of potential GDP and the cyclical output gap (Table I-3). Similar to the univariate decomposition evidence based on the estimated unobserved components model, the production function-based approach suggests that potential GDP growth slowed considerably since 1990, reflecting lower growth rates in the potential growth rates of all three production function inputs. Moreover, the estimates of the cyclical output gap corresponds closely with the univariate estimates derived from the unobserved components model.
The production-function based estimates are sensitive to the underlying trend assumptions for production function inputs. In particular, assuming that potential total factor productivity growth during 1991-95 did not slow from 0.7 percent (1977-90) to 0.3 percent—as suggested by the Hodrick-Prescott filter estimates—but remained at its previous level would increase the size of estimated cyclical output gaps in 1995 and 1996 by 2 percentage points of potential GDP and raise average potential output growth during 1991-95 to 1¾ percent.9
Regarding prospects for medium-term growth, potential output growth could accelerate in the second half of the 1990s to 1¼ percent, reflecting a partial recovery of total factor productivity growth to its previous trend growth rate. Potential labor input and capital stock growth during 1997-02 were assumed to remain in line with their respective trend growth rates during the first half of the 1990s. The resident population during the period 1997-02 is assumed to grow at an average rate of 0.6 percent, the rate projected in the “population continuity scenario” of the Federal Statistical Office (1992). The medium-term trend of the labor force participation rate is extrapolated to increase by ½ percentage point per annum during 1997-02. To close the estimated output gap of 3½ percent in 1996, and assuming potential GDP growth of 1¼ percent during the period 1997-02, actual real GDP growth would have to average about 2 percent per annum during 1997-02. If potential total factor productivity growth recovers back to previous to its trend growth rate of 0.7 percent, potential GDP would expand at a rate of about 1½ percent.
D. Other Macroeconomic Evidence: Inflation and Private Consumption Behavior
This section examines additional macroeconomic evidence on potential output growth and the cyclical output gap, drawing on the time series behavior of inflation and private consumption during the period 1976Q1-1996Q2.
Standard models of short-term output and price determination associate movements in the inflation rate with movements in the cyclical output gap (though a short-term Phillips curve relationship). However, while inflation is (positively) correlated with the cyclical output gap, it is also assumed that inflation has no long-term effect on the level of-real GDP (vertical long-run Phillips curve).10 These ideas may be exploited in a bivariate decomposition of output into a “supply component”—which is driven by shocks with permanent effects on the level of output—and a “demand component”—which reflects the shocks with transitory effects on the level of output. In this exercise, movements in inflation rates are assumed to provide the necessary information to identify the transitory and permanent output shocks using a vector autoregressive model (VAR) (see Blanchard and Quah (1989) for details regarding model specification and estimation).
A vector autoregressive (VAR) model was applied to quarterly data on real GDP growth and GDP inflation rate data for the period 1976.Q1 to 1996.QII using 8 lags of each variable in the VAR. The estimated supply and demand components derived from the VAR suggest that the transitory output component at the end of the sample interval (i.e. in 1996.QII) is about 1 percent (Chart I-6). This estimate is below the output gap estimates obtained from the production function or linear trend approaches. However, there are three caveats attached to this exercise. First, the relationship between inflation and the cyclical output gap may be nonlinear in the sense that at low inflation rates prices become stickier and a larger portion of a given demand shock is reflected in cyclical output movements (see Clark, Laxton, and Rose (1995)). Given Switzerland’s experience of relatively low inflation rates—particularly since 1993—the presence of nonlinearity would clearly reduce the output gap downwards. Second, at least some of the inflation were likely correlated with supply shocks (e.g., sharp changes in oil prices). This would obfuscate the correlation between the demand component and inflation. And third, the moderate length of the available time series imparts some imprecision to the VAR results.
Switzerland: Bivariate Decomposition of Real GDP in Supply and Demand Components, 1978Q1 – 1996Q2
Citation: IMF Staff Country Reports 1997, 018; 10.5089/9781451807141.002.A001
Source: Staff estimates.Switzerland: Bivariate Decomposition of Real GDP in Supply and Demand Components, 1978Q1 – 1996Q2
Citation: IMF Staff Country Reports 1997, 018; 10.5089/9781451807141.002.A001
Source: Staff estimates.Switzerland: Bivariate Decomposition of Real GDP in Supply and Demand Components, 1978Q1 – 1996Q2
Citation: IMF Staff Country Reports 1997, 018; 10.5089/9781451807141.002.A001
Source: Staff estimates.An alternative macroeconomic indicator that could provide some evidence on the size of transitory output movements is the time series for private consumption (see Cochrane (1994) for a discussion). Most (successful) models of private consumption behavior predict that consumers seek to smooth their consumption spending with respect to transitory ups and downs in real income. More pertinently, according to the permanent income hypothesis of consumption, consumers would smooth consumption with respect to transitory real income movements, while adjusting real consumption to perceived movements in permanent income. A simple approach was adopted by regressing the logarithm of private consumption on the logarithm of real GDP. The residual (”spread”) of the regression is interpreted as a predictor of future transitory GDP movements. Accordingly, if the spread is positive—i.e. measured consumption is above measured real GDP—this would provide evidence that consumers view measured real GDP as below permanent or potential GDP.
The estimated spread from a linear regression of real private consumption on real GDP for the time period 1976Q1-1996 indeed suggests that consumers considers actual GDP to be below permanent GDP but not by a large margin (Chart I-7). Similar regressions with other (and perhaps more appropriate) real income concepts including real GNP and real disposable income (the latter is only available at annual frequencies) yield similar results. Two important caveats attach to this exercise. First, to the extent that liquidity constrained consumers account for a significant share of total consumption, private consumption would also reflect transitory movements in real incomes.11 And secondly, consumers in Switzerland may have reacted to the significant increase in job and income insecurity since 1990 by increasing their propensity for precautionary saving. This structural break would be reflected in co-movements of observed real consumption and income since 1990, thus undermining the identifying assumption of the present exercise.
Switzerland: Real Private Consumption as an Indicator of Potential Output Growth, 1976Q1 – 1996Q2
Citation: IMF Staff Country Reports 1997, 018; 10.5089/9781451807141.002.A001
Source: Staff estimates.Switzerland: Real Private Consumption as an Indicator of Potential Output Growth, 1976Q1 – 1996Q2
Citation: IMF Staff Country Reports 1997, 018; 10.5089/9781451807141.002.A001
Source: Staff estimates.Switzerland: Real Private Consumption as an Indicator of Potential Output Growth, 1976Q1 – 1996Q2
Citation: IMF Staff Country Reports 1997, 018; 10.5089/9781451807141.002.A001
Source: Staff estimates.APPENDIX
Data on real GDP, private consumption, the GDP deflator employment, number of registered unemployed persons, capacity utilization in manufacturing, and the resident population are taken from the data base of the Swiss Institute for Business Cycle Research.
The data on the real capital stock are constructed as the cumulative values of investment in machinery and equipment and non-residential structures using the data in Lüscher and Ruoss (1996). The annual depreciation rates for machinery and equipment and non-residential structures were fixed at 20 percent and 4 percent, respectively. The measurement of the data on labor input also follows the methodology in Lüscher and Ruoss (1996), and the SNB provided series on the number of part-time workers and annual hours worked.
Total factor productivity is derived as the residual from the production function (5) in the text, where the capital stock is measured in effective units, i.e. after adjusting for capacity utilization.
References
Chapter I
Akerlof, George A., William T. Dickens, and George L. Perry, 1996, “The Macroeconomics of Low Inflation,” Brookings Papers on Economic Activity: 1, (Washington: Brookings Institution), pp. 1–76.
Blanchard, Olivier Jean, and Danny Quah, 1989, “The Dynamic Effects of Aggregate Supply and Demand Disturbances,” American Economic Review, Vol. 84, pp. 655–73.
Clark, Peter, Douglas Laxton, and David Rose, 1995, “Asymmetry in the U.S. Output-Inflation Nexus: Issues and Evidence,” IMF Working Paper No. 95/76 (Washington: International Monetary Fund).
Cochrane, John H., 1994, “Permanent and Transitory Components of GNP and Stock Prices”, Quarterly Journal of Economics, Vol. 109 (February), pp. 241–65.
Federal Statistical Office, 1992, Les Scénarios de I’Évolution Démographique de la Suisse, 1991-2040 (Bern: Federal Statistical Office).
Harvey, Andrew C., 1989, Forecasting, Structural Times Series Models and the Kalman Filter (Cambridge: Cambridge University Press).
Harvey, Andrew C., and Albert Jaeger, 1993, Detrending, Stylized Fact, and the Business Cycle, Journal of Applied Econometrics No. 8, September, pp. 231–47.
Lüscher, Barbara, and Eveline Ruoss, 1996, “Entwicklung der potentiellen Production in der Schweiz,” Geld, Währung und Konjunktur, Quarterly Bulletin No. 14 (1), (Zürich: Swiss National Bank, March), pp. 61–74.
Marty, Rudolf, 1996, “Wachstumstrend, Produktionspotential und Output-Lücken,” Konjunktur No. 59 (July/August), (Zürich: Konjunkturforschungsstelle der ETH Zürich), pp. 15–26.
Organization for Economic Cooperation and Development, 1996, OECD Economic Surveys 1995-96: Switzerland (Paris: OECD).
Swiss National Bank, “Swiss Monetary Policy in 1995,” Geld, Währung und Konjunktur, Quarterly Bulletin No. 12 (4) (December), (Zürich: Swiss National Bank), pp. 272–74.
Prepared by Albert Jaeger.
The time period for Germany starts in 1960 to exclude the rapid growth period in the 1950s following the rebuilding from World War II.
The reasons are unclear as to why Switzerland’s real GDP growth rate since 1976 has been markedly lower than in other industrial countries with similar per capita income levels and despite the sustained high Swiss national savings rate. However, the following factors and/or caveats are often put forward in this context. First, Switzerland’s terms of trade improved significantly since 1975 (by some 25 percent), suggesting that real GDP growth underestimates the rise in living standards. Second, the large supply shocks since the mid-1970s combined with heavily regulated labor and product markets may have inhibited the Swiss economy’s growth potential. Third, the sharp rise in women’s labor force participation rate since the mid-1970s was accompanied by a distinct trend to part-time work. And fourth, measurement problems related to the national accounts statistics could have led to downward biases in output measures.
The decomposition is based on a Cobb-Douglas production function assuming factor income shares of 0.7 and 0.3 for labor and capital income, respectively.
If the smoothing constant of the Hodrick-Prescott filter is fixed at the widely used value 1,600 (for quarterly data), the Hodrick-Prescott filter will represent an “optimal filter” (in the sense of yielding decompositions in potential output and output gap that are congruent with the statistical properties of the GDP series) if the variance η(t) is zero and the ratio between the variances of ζ(t) and GAP(t) is equal to 1,600 (Harvey and Jaeger (1993)).
The econometric software package STAMP (Structural Time Series Analyser, Modeller and Predictor) was used for estimating the unobserved component models reported in Table 3.
However, the Hodrick-Prescott filter estimates are very sensitive to the choice of data frequency and/or smoothing constant (as should be expected from the above discussion of the UC model). For example, applying the Hodrick-Prescott filter to quarterly GDP data spanning 1976.Q1-1996.Q2 and using a smoothing constant of 1,600 gives an output gap of almost zero, in line with unpublished results reported by the FFA.
The pronounced procyclicality of the labor force participation rate may suggest that some real GDP fluctuations in Switzerland are characterized by hysteresis, i.e. cyclical output fluctuations lead to changes in the size of the labor force and—through this channel—”spill over” into changes in potential output. However, standard decompositions of GDP into trend and cyclical output components rule out the possibility of hysteresis a priori. Preliminary work based on an UC model that allows the cyclical component to affect potential output with a lag, suggests that hysteresis models of GDP growth could be germane for Switzerland during this period.
Recent SNB work by Lüscher and Ruoss (1996) arrives at an estimate of potential output growth rate of 1¾ percent during 1991-95, mainly reflecting their assumption that total factor productivity follows a deterministic trend throughout the period 1976-95.
However, Akerlof, Dickens, and Perry (1996) have argued—based on U.S. evidence—that the long-run Phillips curve bends to the left (using output gaps as the slack indicators) at low inflation rates.
Previous staff work has estimated that the share of liquidity constrained consumers in Switzerland could amount to about 50 percent.
