APPENDIX I: Data
35. The following provides additional information regarding the data used, including the sources and nature of estimates and projections. Figure A1 shows a number of these series, including their HP filters where relevant.
a: labor share of income is calculated from ISTAT data by scaling upwards the share dependent employment income in value added (at market prices, excluding financial intermediation services indirectly measured, FISM), by the ratio of total employment to dependent employment.
Y, K, E: output, net capital stock, and employment data are from ISTAT, forecasts are based on current WEO projections.
util: survey measure of production capacity in use in industry, provided by the Bank of Italy. It is assumed to return to its sample average in 2002 and beyond, which at 92 percent is above the level of the first quarter 2002, but below the average of 2001 (around 93 percent). Potential utilization is based on the HP filter of this series.
WP: working age population data from the OECD, growth rates are as projected by ISTAT.
HW: hours worked in industry from ISTAT. For 2002 and beyond, it is assumed constant at the 2001 level. Potential hours worked is the HP filter of this series.
u and labor force: unemployment rate and labor force are from the ISTAT labor force survey. Forecasts are based on WEO projections.
pr: participation rate is the ratio of the labor force to WP. Under current WEO projections, this ratio grows steadily to 64 percent by 2007.
NAIRU: see Appendix II.
W: wages are the compensation rate of the business sector (annual salary per employee, in euros); data are from the OECD.
APPENDIX II: NAIRU Estimates
36. This appendix describes the construction of the NAIRU used in this paper, and compares it with OECD estimates.
37. The NAIRU is estimated using a simple method described by Giorno and others (1995).25 This starts by defining the NAIRU as the level of unemployment above (below) which inflation is falling (rising): 26
where: W is the nominal wage level, U is the actual unemployment level, and D is the first difference operator. An estimate of α can be obtained by applying the approximation that the NAIRU is constant between any two consecutive periods, in which case:
Combining equations (Al) and (A2) provides an estimate of the NAIRU:
38. The resulting series is smoothed to eliminate erratic components.27 Figure A2 compares actual employment with this (initial) estimate of the NAIRU (labeled NAIRU1). Further smoothing is conducted to produce the final estimate of the NAIRU (labeled NAIRU2)—these modifications reflect the view that the true NAIRU is not likely to have declined in the early 1990s (as implied by NAIRU), and that labor market reforms are likely to have led to a decline in the NAIRU in the mid 1990s (somewhat earlier than suggested by NAIRU1).
39. Four comments regarding the NAIRU estimate are warranted:
First, the new estimates imply an unemployment gap (the difference between unemployment and the NAIRU) that is relatively close to that produced by the OECD, especially in 2001 (Figure A3).
Second, the unemployment gap from 1980 to 2001 is estimated to have been positive on average, due largely to observations after 1994 (the average gap until that time was only 0.1 percentage point). The gap was especially large from 1997 to 1999. While this was a period when reforms were leading to greater labor market flexibility, it was also a time of sizable adverse shocks to labor demand—following from tighter monetary policy and sizable fiscal consolidation necessary to help meet the Maastricht criteria. Hence, it was possible for unemployment to remain high, and even rise, at a time when the NAIRU was thought to be on the decline.
Third, these estimates are comparable to those implied by the bivariate model of the NAIRU presented by Boone and others (2002). Their model produces a range of estimates depending on the specification of the volatility of the NAIRU relative to the unemployment gap. Using a range of values for this volatility parameter that they argue is “reasonable” produces a range of NAIRU estimates spanning those presented above; these estimates also imply a persistent unemployment gap over the past decade or more.
Fourth, staff project that on current policies, the NAIRU will decline only in the coming years, from 8.8 percent in 2001, to 8.7 percent in 2002, and 8.6 percent thereafter.
Ahn, S., 2002, “Competition, Innovation and Productivity Growth: A Review of Theory and Evidence,” OECD Economics Department Working Paper, no. 317.
Baxter, M. and R.G King, 1995, “Measuring Business Cycles: Approximate Band-Pass Filters for Economic Time Series,” NBER Working Paper no. 5022(Cambridge, Massachusetts: National Bureau of Economic Research).
Bentolila, S. and G. Saint-Paul, 1998, “Explaining Movements in the Labour Share,” CEPR Discussion Paper no. 1958 (London: Centre for Economic Policy Research).
Bloom, D.E., D. Canning, and J. Sevilla, 2002, “Technological Diffusion, Conditional Convergence, and Economic Growth,” NBER Working Paper no. 8713 (Cambridge, Massachusetts: National Bureau of Economic Research).
Boone, L., M. Juillard, D. Laxton, and P. N’Diaye, 2002, “How Well Do Alternative Time-Varying Parameter Models of the NAIRU Help Policymakers Forecast Unemployment and Inflation in the OECD Countries,” IMF Working Paper, forthcoming.
Brandolini, A. and P. Cipollone, 2001, “Multifactor Productivity and Labour Quality in Italy, 1981-2000,” Bank of Italy, Economic Research Discussion Paper no. 422.
De Masi, P. R., 1997, “IMF Estimates of Potential Output: Theory and Practice,” IMF Working Paper 97/177(Washington: International Monetary Fund).
Decressin, J., and others, 2001, Selected Euro-Area Countries: Rules-Based Fiscal Policy and Job-Rich Growth in France, Germany, Italy, and Spain, IMF country Report. No.01/203(Washington: International Monetary Fund).
Elmeskov, J., 1993, “High and Persistent Unemployment: Assessment of the Problem and Its Causes,” OECD Economics Department Working Paper No.132.
European Commission, 1999, “Comparison of Trend Estimation Methods,” issues paper for the EPC Working Group on output gaps, II/346/99/EN.
Giorno, C., P. Richardson, D. Roseveare, and P. van den Noord, 1995, “Estimating Potential Output, Output Gaps and Structural Budget Balances,” OECD Economic Studies, 24, pp. 167–209.
Hall, R.E. and C.I. Jones 1999, “Why do Some Countries Produce So Much More Output Per Worker than Others?” The Quarterly Journal of Economics, 114:83–116.
Mc Morrow, K. and W. Roeger, 2001, “Potential Output: Measurement Methods, ‘New’ Economy Influences and Scenarios for 2001-10—A Comparison of the EU15 and the U.S.,” European Commission Economic Papers No.150.
Scarpetta, S., P. Hemmings, T. Tressel, and J. Woo, 2002, “The Role of Policy and Institutions for Productivity and Firm Dynamics: Evidence from Micro and Industry data,” OECD Economics Department Working Papers No.329.
Willman, A., 2002, “Euro Area Production Function and Potential Output: A Supply Side System Approach,” ECB Working Paper Series No.153.
Prepared by Christopher Kent.
But rising to 2.0 percent in 2000.
De Masi (1997) provides a survey regarding the application of this approach within the IMF. Mc Morrow and Roeger (2001) describe the production function approach and compare it with other approaches to estimating potential output.
Willman (2002) examines the more general constant elasticity of substitution (CES) production function. He presents evidence based on the euro area that suggests the Cobb-Douglas function provides a good approximation and, moreover, that output gap estimates are relatively insensitive to alternative parameterizations and functional forms of the underlying production function.
The labor share does vary over time—rising from 0.71 in 1970 to 0.77 in 1975 and then declining steadily to 0.62 by 2001. Estimates of the gap and potential output growth presented below are, however, broadly unchanged if instead the labor share is assumed equal to the level of 2001, Giving greater weight to the capital input in this way reduces the output gap in 2001 by only 0.06 percentage points, and increases potential growth by 0.1 percentage points (by 2007) relative to results presented in Tables 3 and 4. For a discussion of the determinants of the labor share and its evolution in the OECD see Bentolila and Saint-Paul (1998).
Another possibility is to acknowledge differences in the quality/skill of different labor inputs (as done, for example, in Brandolini and Cipollone, 2001, discussed below).
In short, this was done by regressing TFP measured as per equation (1) on util, a constant, and a time trend, and then testing whether the coefficient on util was significantly different from a=0.7.
Of course the same problem applies also to the beginning of the sample, but for policy purposes the focus is on the current output gap and future potential output
This is the standard approach used by the IMF and the OECD when applying the production function methodology (European Commission 1999).
Baxter and King (1995) show that the HP filter tends to give a disproportionate emphasis to the end-points of the cycle (the first and last 3-4 observations), if no corrective measures are applied. Mc Morrow and Roeger (2001) also find this when estimating potential output for EU countries by applying an HP filter to real GDP. They find that a forecasting error of plus or minus 0.5 percentage points alters the estimate of the output gap by around 0.2 percentage points. Moreover, this sensitivity is similar across EU countries and does not appear to be strongly related to the cyclical position.
It is worth distinguishing the two roles played by these projections. The first is to help mitigate the impact of the end-point problem on estimates of the output gap up to 2001. The second is to provide inputs to form a projection for future potential output growth.
The growth of employment over the late 1990s occurred despite weaker growth in labor productivity in part because of wage moderation (Decressin and others, 2001). Also, labor market reforms allowed for more flexible use of labor—including the use of atypical contracts (see Staff Report)—supporting greater employment of women, and of youth.
This argument is similar to one made in the authorities new medium-term program (Documento di Programmazione Economico-Finanziaria, DPEF, July 2002), although the authorities assume annual employment growth to slow to only 1.6 percent. Though partly cyclical in nature, there is already evidence of slower employment growth in 2002.
Indeed, to the extent that on-the-job experience/training is facilitated by the initial level of education, the trend rise in schooling will help to reinforce this process.
Cases (ii) and (iii) may be closely related, since the relaxation of constraints on these forms of employment is likely to have facilitated the entry of otherwise less experienced/productive persons into the workforce.
Ahn (2001) provides a comprehensive review of the empirical literature in this area and confirms that the link between product market competition and productivity growth is positive and significant.
Calculations are from Decressin (2002)—kindly provided by the author—based on OECD business sector data. Table 1 and 2 data are not exactly comparable, since the former are based on economy—wide measures of growth. Nevertheless, at least for Italy the difference in measured TFP growth is not significant.
Laxton (1999) also finds evidence of catch-up of levels of labor productivity in Italy to that of the United States over the longer term. These findings of TFP convergence are similar in spirit to the finding of “conditional convergence" (see for example, Barro and Sala-i-Martin, 1995) based on catch-up of the capital stock to steady state levels.
Trend TFP growth as implied by applying the HP filter up to 2004 (that is, the correction for the end-point problem suggested by Baxter and King, 1995) is only 0.5 percent per year, while extending the filter to 2007 implies trend TFP growth in 2001 of only 0.7 percent. Both of these are lower than obtained by ignoring the end-point problem—that is, applying the HP filter up to only 2001.
To estimate the NAIRU, the EU adopt a combined Kalman filter and Phillips curve approach, whereby the deviation of unemployment from the NAIRU is negatively related to the change in wage inflation, controlling for other temporary shocks to wage inflation. One feature of their approach is that the unemployment gap is restricted to have a mean of zero over the sample period (so as to also ensure a symmetrical output gap over the sample).
As in this chapter, the EU estimate trend TFP growth by using the HP filter and extending the sample period beyond 2001 using projections.
The DPEF does not specify participation rates. These are calculated by assuming that the projections forworking age population are the same as those used by the staff (see Appendix 1). Just over half of the difference in projected employment growth rates is accounted for by the lower unemployment rate projected by the authorities, the remainder by their higher participation rate projection.
Using the authorities projections for participation within the staff model would imply an increase in the estimated 2001 output gap of around 0.4 percentage points. This arises from the fact that assuming a higher future participation rate reduces the gap between the current actual and trend participation rates to near zero (the staff estimate this gap to be 0.5 percentage points in 2001).
The OECD (2002) argue that rapid employment growth (toward the Lisbon target levels) would suppress productivity growth because it implies bringing into the workforce a large number of persons not previously employed (especially from the South), for whom the productivity level is on average lower than that of the existing workforce.
This method was originally espoused by Elmeskov (1993) who showed that the estimates were similar to those from comparable methods based on the alternative Okun’s law or Beveridge curve relationships.
Wage inflation is used, since the link to unemployment gap is more direct than it is for inflation of goods and services.
This is done by first replacing outlying observations—arising in a few periods when wage inflation is almost constant between two years—with linear interpolations, and then applying a moving average filter to the series. Also, the sample period is extended to 2003 using forecasts to help avoid the end-point problem.