This Selected Issues paper studies Luxembourg’s economic growth performance of the past two decades with a view to shedding light on the growth prospects and fiscal implications. The paper investigates whether the recent weakness in activity is largely transitory or whether it heralds a new era of lower potential output growth. The paper also explains one option for reforming the pay-as-you-go pension pillar, which is to link pension benefits to the contributions base through a “solidarity factor.”

Abstract

This Selected Issues paper studies Luxembourg’s economic growth performance of the past two decades with a view to shedding light on the growth prospects and fiscal implications. The paper investigates whether the recent weakness in activity is largely transitory or whether it heralds a new era of lower potential output growth. The paper also explains one option for reforming the pay-as-you-go pension pillar, which is to link pension benefits to the contributions base through a “solidarity factor.”

I. Trend Versus Cycle in Macroeconomic and Fiscal Developments 2

A. Introduction and Summary

4. The current slowdown is unusual in terms of its depth, duration, and the role of the financial services sector (Figure I-1). Real GDP growth has fallen to about 1¼ percent starting 2001, from an average growth rate of 5½ percent during 1980–2000. The last period of similarly sluggish activity was around the period of the world cycle trough of 1981. Neither around the world cycle trough of 1991, nor around the European cycle trough of 1993 did Luxembourg’s annual real GDP growth ever fall below 3¼ percent.3 At the sectoral level, the weakening of value added has been broad based but particularly pronounced in financial services, driven by the correction in global equity markets. Real value added in this sector shrank for the first time since 1985, reflecting the fall in stock-market valuations—by about 60 percent between September 2000 and April 2003—according to the euro-area Dow Jones index (Figure I-2). Given the economy’s large exposure to financial services, real GDP growth has thus fallen more than in other euro-area countries.

Figure I-1.
Figure I-1.

Luxembourg: Recent Economic Developments, 1985-2003

(In percent)

Citation: IMF Staff Country Reports 2004, 124; 10.5089/9781451824278.002.A001

Sources: STATEC; IMF, World Economic Outlook; and Bloomberg.1/ Data for 2003 are IMF staff estimates.
Figure I-2.
Figure I-2.

Luxembourg: Equity Markets and Asset Prices, 1985-2003

Citation: IMF Staff Country Reports 2004, 124; 10.5089/9781451824278.002.A001

Sources: Datastream; Jaeger and Schuknecht (2003); and IMF staff estimates.1/ Average of Belgium, France, and Germany.

5. Not surprisingly, the budgetary situation has also deteriorated significantly. For the first time in two decades the general government budgetary balance is estimated to have recorded a deficit, equivalent to about 1 percent of GDP in 2003, although tax revenue has yet to reflect the full effect of the economic slowdown. This follows surpluses averaging 4½ percent of GDP during the boom years of 1998-2000. Furthermore, budgetary deficits are now projected to persist well into the second half of the decade, still reaching 1½ percent of GDP in 2006, according to the authorities’ latest Stability and Growth Program (SP) update.

6. A key issue is whether the recent weakness in activity is largely transitory or whether it heralds a new era of lower potential output growth. Identifying changes in trend (potential) growth early and accurately is of fundamental importance for adjusting policies in a timely manner. Accordingly, this chapter attempts to decompose recent output developments into their trend and cycle components and explores some fiscal implications. The main findings are that both structural and cyclical factors are at play in the current growth slowdown, with output and productivity trend growth rates declining already for some time. While the recent recovery in equity prices—by some 30 percent since April 2003—can be expected to boost financial sector value added and real GDP growth, asset prices remain well below previous peaks and will continue to reverberate for several years in the economy and the public sector accounts.

7. The remainder of the paper is organized as follows. Section B reviews recent output developments, distinguishing between cyclical and structural components but focusing on the latter; Section C analyzes the cyclical component in more detail; and Section D concludes.

B. Recent Macroeconomic Developments: Trend versus Cycle

8. Various methods can be used to decompose output developments into trend and cycle components. One approach develops a structural model of the economy to separate trend and cycle components of output changes. Such a model might assume that cyclical changes are caused mainly by demand-side shocks while trend changes are caused mainly by supply-side shocks (e.g., Blanchard and Quah, 1989). Another approach eschews assumptions about the structure of the economy and instead decomposes actual output into permanent and transitory movements using univariate statistical techniques. This is the approach that is adopted here because it is relatively simple to apply, parsimonious on the assumptions that it imposes, and well-suited to a very open economy such as Luxembourg.

9. Output trend developments are less stable for small economies than for large economies. The reason is that various supply-side shocks concern small economies more, notably domestic and foreign changes in taxation and regulation as well as shifting locational preferences and opportunities of economic agents. More importantly, in small, open economies such shocks are likely to interact in a reinforcing manner: for example, changes in regulation can dampen financial sector and output growth which, in turn, reduces migration, with further output losses. Both vicious and virtues growth circles are possible. These circles are difficult to model. As a partial remedy to the issues raised by migration, this section scrutinizes separately the trend developments in output and labor productivity growth.

10. Among the univariate statistical measures to decompose output the Ouliaris ideal band-pass filter (2001) has several advantages over other methods. Output changes may be driven by many processes, some operating at irregular or very high frequencies (e.g., seasonal cycles), others at an intermediate frequency (business cycles), and some at a low frequency (changes in trend growth rates). The fundamental idea of univariate methods is to decompose the processes into these three components using a filtering procedure. Filters typically assume a business cycle length of anywhere between 2 to 8 years. The Ouliaris filter has several advantages over other filters.

  • Unlike other filters (e.g., Hodrick and Prescott, 1997), the Ouliaris filter operates entirely in the frequency domain.

  • Unlike the Baxter and King (1999) band-pass filter, the Ouliaris filter is not affected by leakage from the zero frequency component of nonstationary series, nor does it involve a loss of observations at either end of the series—for policymaking, recent data are particularly important.

  • The Ouliaris filter is less likely to introduce spurious cycles, which is a particular problem with the Hodrick-Prescott filter. The smoothing usually applied by practitioners assumes that the series under investigation shares the statistical properties of U.S. time series on which the Hodrick-Prescott filter was based.

  • The Ouliaris filter is statistically consistent, unlike the Baxter and King filter, and does not suffer from end-of-sample bias, unlike the standard Hodrick-Prescott filter.

11. Ouliaris filter estimates suggest that the potential output growth rate of the Luxembourg economy might be about 4 percent per annum but that has been declining over the past decade (Figure I-3). An estimate of recent potential output growth based on the 1960–2002 ESA 79/95 spliced time series is about 4 percent; based only on the 1981-2002 ESA 95 data it is close to 5 percent. However, given the variance that supply-side shocks introduce into estimates of potential output—as shocks to productivity interact with migration or commuting in a reinforcing manner—it might be more appropriate to focus on the average estimate for potential output growth over several decades. Both for the spliced 1960–2002 ESA 79/95 and the 1981–2002 ESA 95 time series average potential growth rate amounts to about 4 percent. Nonetheless, a cause for concern is that potential output growth has declined by about one fifth (ESA 95 data) to one quarter (ESA 79 data) percentage point per annum since the peaks reached in the late 1980s or early 1990s.

Figure I-3.
Figure I-3.

Luxembourg: Actual and Trend Real GDP Growth Rates, 1961-2002

(In percent)

Citation: IMF Staff Country Reports 2004, 124; 10.5089/9781451824278.002.A001

Sources: STATEC; IMF, World Economic Outlook; and IMF staff estimates.

12. The slowdown in output trend growth is mainly driven by labor productivity trend growth and there is no compelling evidence for a sustained turnaround thus far (Figure I-4).4 During 1981–2002, labor productivity trend growth averaged about 2½ percent. However, after reaching a peak of about 3½ percent in 1989, it has been decelerating to around 1⅓ percent recently. Like real GDP trend growth, labor productivity trend growth fell during the 1990s, on average by one fifth percentage point.5 Therefore, most of the drop in potential output growth seems to be due to labor productivity.

Figure I-4.
Figure I-4.

Luxembourg: Actual and Trend Labor Productivity Growth Rates, 1980-2002

(In percent)

Citation: IMF Staff Country Reports 2004, 124; 10.5089/9781451824278.002.A001

Sources: STATEC; IMF, World Economic Outlook; and IMF staff estimates.1/ Differentials from output-weighted average productivity level for Germany, France and Belgium.

13. The decrease in labor productivity trend growth in Luxembourg is not shared to a similar extent by neighboring countries. The trend growth differential between Luxembourg on the one hand and Belgium, France, and Germany on the other averaged about ¾ percentage points per annum during 1981-2002 (Figure I-4). However, it peaked at about 1¼ percentage points in 1989 and has since almost vanished.

14. The deceleration of labor productivity is likely driven by total factor productivity (TFP). The reason is that during the 1990s the average annual growth rate of fixed capital formation was a healthy 9 percent, relative to real GDP growth of about 6¼ percent. IMF (2002) concluded that there was a deceleration of TFP growth in Luxembourg in the second half of the 1990s which seemed to have affected mostly the non-financial sector. However, the slowdown has now also spread to the financial sector, where average TFP growth declined to 1½ percent per annum in 1996–1999 against 2¼ percent per annum in the first half of the 1990s.

15. The decline in TFP growth might be caused by changes in the regulatory environment or might reflect a maturing of the financial services industry. Changes in the regulatory environment are difficult to pinpoint, although the regulatory burden on the financial services sector certainly has not decreased. Falling returns to learning—consistent with an Aghion and Howitt (1999) learning by doing (LBD) endogenous growth model—might be a more promising explanatory factor. Luxembourg grew out of the steel crisis helped by the rapid development of a strong financial sector during the second half of the 1970s and the first half of the 1980s. Initially, LBD capitalized on institutional-policy advantages in a strongly competitive environment, driving TFP growth. But as the financial sector gained market share and matured, the relative importance of capital and labor in the production process increased, while that of LBD and TFP growth diminished. Moreover, the fall in the rate of firms entry in the financial industry might be another force driving the drop in the trend growth rate of the financial industry.6

16. Labor force growth has supported output trend growth strongly thus far, suggesting that the slowdown in productivity has not had an obvious effect on migration and commuting. Population growth accelerated from 0.3 percent per annum in the 1980s to 1.4 percent per annum during 1990-2002; over the same period, domestic employment growth accelerated from an annual average rate of 1.4 percent to an annual average of 3.6 percent (Figure I-5), driven by a rising number of commuters. This rapid expansion of employment likely was related to relatively prudent wage setting as real unit labor costs mostly declined.

Figure I-5.
Figure I-5.

Luxembourg: Labor Market Developments, 1985-2003

(In percent)

Citation: IMF Staff Country Reports 2004, 124; 10.5089/9781451824278.002.A001

Sources: STATEC; Eurostat; IMF, World Economic Outlook; and IMF staff estimates.1/ Data for 2003 are IMF staff estimates.

17. However, the rising unemployment among residents and the recent lack of wage moderation are causes for concern (Figure I-5). The wage moderation did not carry through the latest slowdown, as real unit labor costs rose significantly in 2001 and 2002. Given that the economy is open and small, and thus a price taker in a global market, it is unlikely that unit labor costs can rise significantly without denting corporate profitability and competitiveness. A further concern is the secular increase in structural unemployment, from 1.4 percent in the 1980s to about double that amount recently. While the structural unemployment rate might appear low in a European context, the actual unemployment rate stands already at about 4½ percent and a wider definition of unemployment that includes people on subsidized work schemes and training is close to 6 percent.

18. Rising actual unemployment is likely to translate into higher structural unemployment because of various institutional features of the labor market, which also explain the coexistence of healthy employment growth and rising unemployment. In particular, the high replacement rate of unemployment benefits, the fairly broad targeting of those benefits (younger workers need only little work experience to obtain benefits and older workers can receive support for up to 2 years), the eligibility conditions and unlimited duration of minimum income support (RMG), and the frequent recourse to early retirement tend to make domestic workers more expensive relative to commuters. Thus, healthy employment growth can be flanked by rising unemployment because commuters have lower reservation wages than residents.7 Furthermore, with unemployment of residents not projected to fall significantly through 2006, a future re-estimation might reveal a ratcheting up of structural unemployment that is not yet apparent in 2002–03 data. The reason is that once residents lose employment, the incentives to accept a job fall over time: skills and therefore attainable wages depreciate, while jobless support is generous and long lasting.

19. Overall, both output trend growth and labor productivity trend growth have been declining but remain at levels that suggest that much of the recent economic slowdown is cyclical. The latest estimate for potential output growth is at least 4 percent per annum, relative to actual growth rates of about 1¼ percent in 2001–03; for potential labor productivity growth it is around 1⅓ percent, relative to a significant actual productivity decline in 2001-03. The difference between output and labor productivity is labor force growth of about 2½ percent annually, driven by migration and commuting.

20. The key issue is whether migration and commuting will be high enough to combine with the relatively low labor productivity trend growth rate to continue to produce trend output growth of 4 percent annually. Modeling the interaction between output growth and migration or commuting is difficult. The evidence in the literature on migration would suggest that as productivity trend growth declines and unemployment rises, the expansion of the labor force should decelerate as well, further lowering output trend growth. This feedback has not been considered in the estimation of potential output growth presented here, because of the absence of standard methodologies to do so. But the important point is that policies can limit any such feedback, for example, by avoiding tax and contribution rate increases as well as rising structural unemployment.8

C. Business Cycle Characteristics

21. This section studies the business cycle characteristics of various economic variables. Using, for example, real GDP growth, the definition of business cycle statistics is as follows. The duration of a cycle is the number of periods between two consecutive peaks (or troughs) of growth rates. The amplitude or deepness of a recession is the growth contraction from peak to trough in percentage points, and mutatis mutandi, for the amplitude of an expansion. Steepness is the rate of change of output growth from peak to trough, or vice versa. Cumulative movements of output growth are the output losses from peak to trough relative to the previous period. As cumulative movements of output are approximated using a linear procedure, a measure of the excess in cumulated movements is also presented.

22. The section also looks at neighboring countries, given the role Luxembourg plays as a European financial center. Accordingly, indicators of economic and financial activity for the euro area, Belgium, France, Germany, the United Kingdom, and the United States are also included. The sample period covers 1980–2002, but data for fiscal variables for Luxembourg are only available on a consistent basis starting in 1990 (see Data Sources), which limits the robustness of the conclusions and poses problems for obtaining significant results in some circumstances.

23. The methodology for determining business cycle characteristics avoids any simple detrending or filtering to avoid altering the statistical properties of the cycle. It is therefore not directly comparable to other work that dates cycles based on deviations from fixed sample means (e.g., Jaeger and Schuknecht, 2003). The main reason for this strategy is to avoid altering the statistical properties of the cycle (King and Rebelo, 1993 and Harding and Pagan, 1989). A local peak in the Harding and Pagan (2002) methodology that is adopted here requires yt-yt-2>0, yt-yt-1>0, yt+1-yt<0, and yt+2-yt<0 with the y series either in levels or in rates of change. Given that this paper is mostly interested in growth developments, the y series is in rates of change, unless otherwise noted.

24. The key features of peak to trough (PT) and trough to peak (TP) developments suggest (Table 1):

  • The duration of cycles in real asset prices is much longer than the duration of cycles in real output. The duration of cycles in Luxembourg output and fiscal variables is similar, between 5-6 years.9 Cycles in real value added in the financial sector (i.e., fees and commissions) appear roughly as long as cycles in real asset prices (BEDEFR), lasting about 10 years.10

  • Cycles in real output and fiscal variables are asymmetric, meaning that developments during the PT phase cannot simply be viewed as a mirror image of developments during the TP phase.11 In particular, contractions are shorter and steeper than expansions, a characteristic that is very pronounced for real revenue growth. Cycles of financial variables are considerably less asymmetric.

  • The amplitudes of cycles in financial variables and in real revenue growth are much larger than the amplitude of cycles in real activity.

  • Real GDP growth exhibits strong early output expansions, as has been found for U.S. real GDP growth.12 This is reflected in the positive value of the “excess” statistics during the upswing phase. Real revenue growth shares this feature.

Table I-1.

Luxembourg: Business Cycle Characteristics

(In percent, unless otherwise indicated)

article image
Sources: IMF, WEO; Datastream; Jaeger and Schuknecht (2003); and IMF staff estimates. PT stands for Peak-to-Trough and TP for Trough-to-Peak.

Annual growth rates.

Ratio to GDP.

Refers to Belgium, France, and Germany real asset prices average.

Refers to BEDEFR plus the U.K. and U.S. real asset prices average.

Cycles could not be statistically identified because of the short sample.

25. Further analysis reveals the cross correlation among business cycle phases of real GDP, fiscal variables, and financial variables. Here cross correlations between economic time series refer to cross correlations between their estimated cycle phases, i.e., cross correlations between time series’ upswing (TP) and downswing (PT) phases, determined according to the Harding and Pagan (2002) methodology. Significant contemporaneous correlations include those between phases of (Table 2): (i) real GDP and real revenue growth; (ii) real GDP growth in Luxembourg and the euro area; (iii) real value added of Luxembourg’s financial sector and the euro-area Dow Jones equity market index;13(iv) the general government balance-to-GDP ratio on the one hand and the euro area real GDP growth and real asset price growth in neighboring countries on the other hand; and (v) world real asset prices and both real asset prices in neighboring countries and the Dow Jones equity market index.14 Furthermore, lead-lag tests of cross correlations between business cycle phases reveal that equity prices (DJEUROGA) lead asset prices (WORLD) for two to three years (Figure I-6); and that financial sector value added (FSRG) leads the general government balance-to-GDP ratio (GGGDP) for up to two years.

Table I-2.

Luxembourg: Correlation of Business Cycle Phases 1/

(Growth rates, except for general government balance)

article image
Sources: IMF, WEO; Datastream; Jaeger and Schuknecht (2003); and IMF staff estimates.

Only significant estimates are shown (* refers to 95 percent significance and ** to 90 percent significance).

Refers to Belgium, France, and Germany real asset prices average.

Refers to BEDEFR plus the U.K. and U.S. real asset prices average.

Figure I-6.
Figure I-6.

Lead-Lag Relations: Asset Prices, Financial Sector, and General Government Deficit, 1990-2002 1/

Citation: IMF Staff Country Reports 2004, 124; 10.5089/9781451824278.002.A001

Sources: STATEC; Eurostat; IMF, World Economic Outlook; and IMF staff estimates.1/ FSRG stands for real value added in the financial sector; GGGDP for the general government budgetary balance (in percent of GDP); DJEUROGA for the DJ Euro Stoxx index; and WORLD for an index of world real asset prices. The horizontal lines delimit a 95 percent confidence interval.

26. The current downswing is unusual also because of the extent of asset price weakness, which will continue to reverberate for some time (Table 3). Recently, all important variables were in a downswing phase, including real GDP growth both in Luxembourg and the euro area, financial sector value added, equity and asset prices, and real revenue (Table 3). Given the close contemporaneous relation between equity prices and financial sector value added, the recent recovery of equity prices can be expected to boost financial sector activity and, with some lag, also to improve the public sector accounts. Past experience would suggest that the recovery could be fairly strong early on but the current setting is unusual. Furthermore, considering the extent of the fall in equity prices since their peak, output and real revenues will likely grow at a more moderate pace than in the 1990s.

Table I-3.

Luxembourg: Business Cycle Phases 1/

Peaks and Troughs (1=yes; 0=no)

article image

Upswings and Downswings (1=upswing; -1=downswing)

article image
Sources: IMF, WEO; Datastream; Jaeger and Schuknecht (2003); and IMF staff estimates.

Annual growth rates; all data are in real rather than nominal terms, unless otherwise noted.

Index in nominal terms.

Ratio to GDP, not annual growth rate; variables in nominal terms.

Refers to Belgium, France, and Germany real asset prices average.

Refers to BEDEFR plus the U.K. and U.S. real asset prices average.

27. A fundamental issue is how the asymmetries in the real GDP cycles and the large amplitudes of asset price cycles should be taken into account in setting fiscal policy. Contractions are short while expansions are long, sending potentially confusing signals about trend output and therefore revenue growth rates. Furthermore, during asset price booms revenues might appear stronger than they fundamentally are. Conversely, during periods of asset price corrections, declines in the revenue-to-GDP ratio would not be surprising: the amplitude of real revenue cycles is much larger than that of real GDP cycles. Fiscal policy needs to take this into account and, considering also the length of asset price cycles, take a cautious approach in determining expenditure policy, using a planning horizon that stretches perhaps further into the future than is currently the case.15 A deeper analysis of the implications of asymmetries in cycles and equity prices for the design of various fiscal rules lies beyond the scope of this paper.

D. Conclusion

28. The main findings are that both cyclical and structural factors play an important role in the current growth slowdown and that, in setting policies, past output and fiscal revenue growth rates should therefore not be extrapolated. The unusually long period of sluggish activity renders estimates of output trend growth very uncertain. An estimation methodology that counts among the most robust to “end-point-problems” reveals that potential output and labor productivity growth have been declining by at least one fifth percentage point per annum over the past decade. While the economy’s potential output growth rate remains high relative to that of neighboring countries, this is no longer the case for the potential labor productivity growth rate. As productivity slows, so could job creation, wage growth, and therefore the expansion of the labor force, with adverse consequences for potential output growth.

29. One important caveat to bear in mind is that potential output growth estimates need to be treated particularly cautiously for Luxembourg; and they should not be used to gauge the amount of slack in the economy. It might be sensible to talk about the existence of a potential output growth rate over the medium run but less so over the long run. The reason is that the labor force variable cannot be considered sluggish over a long horizon. Other reasons for treating potential output growth estimates with care also relate to the openness of the economy, notably its sensitivity to domestic and foreign changes in taxation and regulation and to shocks to the demand for financial services. Furthermore, given the relation between commuting and migration on the one hand and economic conditions on the other, output gaps cannot be reliably gauged.

30. The characteristics of the latest economic slowdown are unusual, notably because of the extent of asset price weakness, which will continue to reverberate for some time. Typically, Luxembourg experiences strong output expansions in the recovery phase. The upswing in equity prices over the past year is consistent with such a scenario. However, the latest slowdown is characterized by weak real GDP growth, sluggish financial sector value added, and an unusually large correction in equity prices from previous peaks. This points to a slower recovery and likely will forestall a return to the real GDP growth rates of the 1990s barring another, unusual acceleration of equity prices. Since the general government budgetary balance continues to react over several years to slowdowns in value added of the financial sector, probably reflecting partly the lag with which corporate taxes are collected, further fiscal weakness can be expected for some time. Also, a return to the tax revenue growth rates of the 1990s appears unlikely.

Data Sources

All data are annual. The data on Luxembourg are from STATEC, the Luxembourg statistical office. Euro-area real GDP is from Eurostat. DJ Euro Stoxx data is from Bloomberg. Real asset prices are from Jaeger and Schuknecht (2003). Data generally cover the following periods (acronyms in parentheses):

Real GDP (GDPG), real value added in the financial sector (FSRG), euro area GDP (EUROAG), and Belgium-France-Germany real asset prices (BEDEFR), US real asset prices (US) and World real asset prices (WORLD): 1980-2002.

DJ Euro Stoxx (DJEUROGA): 1987-2002.

Central government balance and general government balance to GDP (CCGDP and GGGDP, respectively), central government revenue (RG): 1990–2002.

Central government revenue cash (RGC): 1980-2002.

The deflator for revenues is the GDP deflator.

References

  • Blanchard, O. and D. Quah, 1989, “The Dynamic Effects of Aggregate Demand and Supply Disturbances,The American Economic Review, 79, pp. 65573.

    • Search Google Scholar
    • Export Citation
  • Cogley, T. and J. Nason, 1995, “Effects of the Hodrick-Prescott Filter on Trend and Difference Stationary Time Series: Implications for Business Cycles,Journal of Economic Dynamics and Control 19, pp. 235278.

    • Search Google Scholar
    • Export Citation
  • Harding, D. and A. Pagan, 2002, “Dissecting the Cycle: A Methodological Investigation,Journal of Monetary Economics, Vol. 49, pp. 36581.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hodrick, R. J. and E.C. Prescott, 1997, “Post-War U.S. Business Cycles: An Empirical Investigation,Journal of Money, Credit, and Banking, Vol. 29, pp. 116.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Jaeger, A. and L. Schuknecht, 2003, “Asset Price Cycles and Fiscal Policy Behavior,IMF, mimeo.

  • Keynes, J. M. (1936). The General Theory of Employment, Interest, and Money, London, MacMillan.

  • Kim, C-J. and C. R. Nelson (1999). “Friedman’s Plucking Model of Business Fluctuations: Tests and Estimates of Permanent and Transitory Components,Journal of Money, Credit, and Banking, pp. 122.

    • Search Google Scholar
    • Export Citation
  • King, R. G., and S. T. Rebelo, 1993, “Low Frequency Filtering and Real Business Cycles,Journal of Monetary Economics, Vol. 33, pp. 40538.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Nadal De Simone, F., 2002, “Common and Idiosyncratic Components in Real Output: Further International Evidence,IMF Working Paper 02/229.

  • Neftcy, S. N. (1984). “Are Economic Time Series Asymmetric over the Business Cycle?,Journal of Political Economy 92, pp. 30728.

  • Ouliaris, S., 2001, “On Extracting Business Cycles From Nonstationary Data,” International Monetary Fund, manuscript.

  • Ramsey, J. and P. Rothman (1996). “Time Irreversibility and Business Cycle Asymmetry,Journal of Money, Credit, and Banking 28, pp.121.

    • Search Google Scholar
    • Export Citation
  • Razzak, W. A. (2001). “Business Cycles Asymmetries: International Evidence,Review of Economic Dynamics 4, pp. 23043.

  • Sichel, D. E. (1993). “Business Cycle Asymmetry: A Deeper Look”, Economic-Inquiry, pp. 22436.

1

See IMF Staff Country Reports No. 02/118, Box1.

2

Prepared by Francisco Nadal De Simone.

3

This is the case according to the (most reliable) ESA 95 data. The ESA 79 data series stretches from 1995 back to 1960 and, in some charts, has been spliced with the ESA 95 for 1996–2002. ESA 79 data show low growth rates for a few years in the 1990s but not for three years in a row, as is the case currently.

4

Consistent employment data back to the 1960s has not been available and thus productivity is analyzed only with the ESA 95 data series.

5

Average labor productivity growth was 1.4 percent during the period 1993–2002 while trend growth in labor productivity was 1.7 percent.

6

According to this LBD model in the Schumpeterian tradition, growth depends on the mix of research and development (R&D) and LBD. LBD improves the quality of products, and this is fully internalized by the firm because only the firm that solves practical production problems benefits from the experience. As in other Schumpeterian models, an increase in the real interest rate reduces growth via a reduction of R&D. Similarly, growth falls if the rate of entry of new firms or products declines. The real interest rate, however, has not trended up.

7

Another reason is the need to import skills that are unavailable locally.

8

Notice that congestion is a potential concern with respect to employment growth but the example of the large metropolitan areas in the world and of other, small open economies suggests that Luxembourg is still far away from hitting any major, natural long-term constraints.

9

Possibly due to the lack of full synchronization among the cycle phases of euro-area countries, the duration of the euro-area real GDP growth cycle is much longer than the duration of the Luxembourg real GDP growth cycle. Nadal-De Simone (2002) contains a description of business cycle characteristics of real GDP both in levels and in growth rates for France, Germany, Italy, and the United States. The duration of those countries’ cycles is about 3 years.

10

See also Jaeger and Schuknecht (2003) and Bernanke, Gertler, and Gilchrist (1999) for similar results. Here, the average cycle length for average real asset prices growth in Belgium, France and Germany (referred to as BEDEFR on Table 1) is 10 years; it is about 9.5 years if the UK and the US real asset prices are included (referred to as “World”). Notice that the index combines three different asset classes, equities, residential property, and commercial property, weighed using the shares of each component in private sector wealth; the private consumption deflator is used to convert nominal to real price indices. Data were kindly provided by Jaeger. Results on real asset prices in levels (not shown) are similar to those for asset prices in growth rates.

11

The considerable asymmetry in cycles in real economic activity is a well-known result in the literature. The literature on output asymmetry is vast and can be traced back to early references in Keynes (1936). More recent work includes the seminal piece by Neftcy (1984), as well as Seichel (1993), Ramsey and Rothman (1996), and Razzak (2001).

12

See Kim (1999).

13

Cycle phases of real value added in the financial sector might not exhibit a strong, positive correlation with phases of real asset prices because lending for personal or corporate real estate accounts only for a small proportion of the activity of Luxembourg banks—more than two thirds of their exposure is to other financial institutions.

14

With cycle phases of Luxembourg real GDP growth or real revenue growth the relation of the phases in the general government balance (in percent of GDP) is positive but insignificant. This might reflect that (particularly capital) expenditure has been expanded in line with GDP, notably to alleviate constraints on growth.

15

The end of the planning horizon of the current Stability and Growth Program is 2006.

Luxembourg: Selected Issues
Author: International Monetary Fund
  • View in gallery

    Luxembourg: Recent Economic Developments, 1985-2003

    (In percent)

  • View in gallery

    Luxembourg: Equity Markets and Asset Prices, 1985-2003

  • View in gallery

    Luxembourg: Actual and Trend Real GDP Growth Rates, 1961-2002

    (In percent)

  • View in gallery

    Luxembourg: Actual and Trend Labor Productivity Growth Rates, 1980-2002

    (In percent)

  • View in gallery

    Luxembourg: Labor Market Developments, 1985-2003

    (In percent)

  • View in gallery

    Lead-Lag Relations: Asset Prices, Financial Sector, and General Government Deficit, 1990-2002 1/