Israel
Selected Issues and Statistical Appendix

The estimated potential output growth decelerated from 4.9 percent in the third quarter of 2000 to 4.2 percent in the first quarter of 2001. In the Israeli context, the sharp and exogenous nature of shocks at the end of 2000 complicated the task of estimation. The paper presents potential output estimates using the Hodrick-Prescott (HP) filter and the unobserved components approach. The potential output and the nonaccelerating inflation rate of unemployment estimates have been obtained simultaneously. The statistical data are also presented in the paper.

Abstract

The estimated potential output growth decelerated from 4.9 percent in the third quarter of 2000 to 4.2 percent in the first quarter of 2001. In the Israeli context, the sharp and exogenous nature of shocks at the end of 2000 complicated the task of estimation. The paper presents potential output estimates using the Hodrick-Prescott (HP) filter and the unobserved components approach. The potential output and the nonaccelerating inflation rate of unemployment estimates have been obtained simultaneously. The statistical data are also presented in the paper.

Estimation of Potential Output in Israel1

A. Introduction

1. The Israeli economy has undergone a series of structural changes in the last two decades. First, inflation has been gradually reduced from triple digits in the early 1980s to low single digits, hitting zero percent in 2000. Second, it has experienced an influx of immigrants from the former Soviet Union since 1989, amounting to nearly 800,000 to date (relative to the total population as of end-1989 of 4.7 million). Finally, against the background of a history of extensive military spending on technology and the government incentives, the high-tech sector has flourished over the past couple of years, putting Israel among the leading centers of the high-tech activity. The outstanding performance of this sector boosted economic growth in 2000: real GDP grew by 6 percent, following the stagnation during 1997–99.

2. The immigration wave led to a significant increase in unemployment rate from an average of 5.7 percent during the 1980–88 to over 11 percent in 1992. It declined gradually to 6.6 percent in 1996 before heading up to 9 percent during the slump in economic activity in 1997–99, Although the unemployment rate is still substantially higher than the pre-immigration period, the absorption of immigrants has been rather successful considering the fact that currently immigrants constitute over 17 percent of the labor force. Two main factors account for this success: increased flexibility in the labor market led to a wage moderation in the early 1990s; and the change in the output mix towards the skill-intensive sectors matched with the skill composition of immigrants, many of whom had been highly educated.

3. These developments suggest that the supply side of the Israeli economy has gone through a major evolution over the past decade. This makes it very important to estimate potential output and the nonaccelerating inflation rate of unemployment (NAIRU) for Israel, which are central to the conduct of both monetary and fiscal policies. Estimated values of these variables will provide valuable insights for assessing the cyclical position of the economy as well as the policy stance in recent years, and also help guide the future course of macroeconomic policies.

4. Difficulties inherent in estimating potential output are quite well-known. The significant structural changes mentioned above complicate the task further, since the frequently used filtering techniques for estimating potential output perform poorly under such conditions. This study presents potential output estimates using Hodrick-Prescott (HP) filter and the unobserved components approach. In the latter approach, potential output and the NAIRU estimates are obtained simultaneously.

5. The organization of the paper is as follows. Section B briefly presents the two alternative methods used in this paper. The estimates for potential output and the output gaps are presented in Section C. Finally, Section D summarizes the main findings.

B. Methodology

6. The methods used in the literature for estimating potential output can be basically classified into two main categories; purely statistical smoothing methods and structural models. For the former, the most widely applied technique is the HP filter, which has disadvantages of being a rather mechanical trend-removal tool based on an ad-hoc choice of a smoothing parameter and the end-point bias affecting the level of the filtered series. The methods covered in the second category involve decomposition consistent with the economic underpinnings of potential output. A widely used method is the production function approach postulating a simple two-factor Cobb-Douglas production function. This method could make use of inflation information by inserting an exogenous NAIRU series into the production function together with other input components. However, even if this is done, it fails to recognize the relationship between the output gap and the unemployment gap, where the latter is defined as the gap between actual unemployment and NAIRU (i.e. the cyclical component of unemployment). Another particularly important class of models is known as the structural time series models (STM) or unobserved components (UC) models. In the literature, a range of models have been proposed to derive potential output using the UC approach, based on either inflation information and/or mutual dependence of cyclical output and cyclical unemployment. This paper implements Apel and Jansson (1997) model which makes use of both the inflation information and the mutual dependence between the cyclical output and the cyclical unemployment. The technicalities of the methods employed in this study are given below.

The Hodrick-Prescott Filter

7. The Hodrick-Prescott filtering procedure is designed to extract the trend component of time series data. Potential output is extracted by minimizing an objective function that has as its arguments the squared differences of actual output from potential output and the squared differences of the potential output growth from one period to next. The parameter λ determines the degree of smoothness of the trend.

1TΣt=1T(lnytlnytp)2+λTΣt=2T1[(lnyt+1plnytp)(lnytplnyt1p)]2

8. The estimated potential output series depends on the chosen value of λ, which is arbitrary to a certain extent. The estimated potential output series matches with the actual output series if λ is set equal to zero, whereas it corresponds to a constant rate growth path connecting the first and last data observations if λ is ∞. In general, higher values of λ lead to smoother growth of potential output, thereby increasing the fluctuations in output gaps. Another shortcoming of the HP filter is its high end-sample bias, stemming from the nature of the minimization problem that makes the smoothed series to always converge to the original sample at both ends of the series.

System estimates of potential output and the NAIRU (STM/UC)

9. Apel and Jansson (1997) proposes an unobserved components model that estimates potential output and the NAIRU simultaneously. This method entails estimating a system of equations which explicitly incorporates the co-variation restrictions on cyclical output and cyclical unemployment suggested by theory. The system has two elements: (1) a Phillips curve which includes “supply shock” variables and incorporates a constant-inflation restriction on the NAIRU; and (2) an Okun’s law relationship which translates cyclical unemployment movements into cyclical output movements. Taken together, these elements give an unobserved components model consistent with the common definitions of the NAIRU and potential output, and thus can be used to derive economically interpretable measures of these key unobservable economic variables.

10. The empirical model consists of the following equations:

Δπt=Σl=14ρlΔπtl+Σj=01ηj(utjutjn)+Σk=04ωkztk+ɛtpc(1)
ytytp=Σt=01ϕl(utlutln)+ɛtol(2)
utn=utln+ɛtn(3)
ytp=α+yl1p+ɛlp(4)
ututn=Σm=12δm(utmutmn)+ɛtp(5)

where

πt: the log difference of CPI

ut: the unemployment rate

utn: the NAIRU

zt: supply shock proxies (normalized so that zt =0 means supply shocks are absent)

yt: the log of real GDP

ytp: the log of potential output (GDP)

ɛtpc , ɛtol , ɛtn , ɛtp and ɛtc are assumed to be IID and mutually uncorrected with constant variances.

11. Equation (1) introduces a version of Gordon (1997)’s so-called “triangle” Phillips model as an identifying restriction within the unobserved components system, while acknowledging the unobservability of the NAIRU. The NAIRU estimates will of course depend on the development of actual inflation. Equation (2) is an Okun’s law relation incorporating the identifying restriction of the mutual dependence of cyclical unemployment and cyclical output fluctuations, while acknowledging the unobservability of both the potential output and the NAIRU. Equations (3)(4) characterize the NAIRU as a random walk process and potential output as a random walk with drift. Finally, equation (5) specifies cyclical unemployment as a pure autoregressive process.

12. In order to estimate this UC system, it is necessary to re-write it in state-space form. The state-space representation of this system is as follows:

Measurement system:

[ytutΔπt]=[10ϕ1ϕ2011000ρ1ρ2][ytputnututnut1ut1n]+[00...0...0s1...sk][z1tzkt]+[etol0etpc](6)

State (or transition) system:

[ytputnututnut1ut1n]=[1001000000δ1δ20010][yt1put1nut1ut1nut2ut2n]+[α0000]+[etpetnetc0](7)

13. Kalman filter is applied to the state-space system given by equations (6) and (7) to derive the unobserved components, namely potential output and the NAIRU, for a given parameter set of the system. Then, the maximum likelihood method is applied to obtain a new set of parameter estimates. This procedure is repeated until convergence is achieved.

C. Empirical Results

The Hodrick-Prescott Filter

14. The sensitivity of the trend component derived by the HP filter to the last observations of the actual series is well known. The data series for Israel contains a sharp deceleration of output growth towards its end, and therefore, the end-point problem results in a fairly low estimated potential growth rate for recent quarters. A widely applied solution to circumvent this problem is to extend actual output series into the future, either using forecasts or assuming a move back to a plausible potential output growth path. In this paper, two versions of HP filter estimation are presented. The first version extends the sample series to 2001 Q2-Q4 using WEO forecasts (seasonally adjusted GDP series),2 and the second version assuming a quick recovery in 2001 Q2 and a 4.4 percent annualized growth rate for 2001 Q3-Q4. The 4.4 percent growth is roughly the average growth rate for the period 1992-2000.

15. The growth rates for the potential output series and corresponding output gaps derived from these two versions of the Hodrick-Prescott filter are presented in Figure 1, Figure 2 and Table 1. In the first version, the potential growth estimate increases smoothly from roughly 4.7 percent in 1990 to 5.7 percent in 1994, which corresponds to the period of large immigration. Then, potential growth decelerates smoothly to 3.3 percent in 2000. Since the economy experienced a sharp slowdown close to the end of the sample period and the WEO forecasts assume weak output growth, the HP filter gives rather weak potential growth estimates for recent years. In line with these low potential growth estimates, the output gap derived from the HP filter is closed by the first quarter of 2000, and during the second and third quarters actual output is above potential. On average, potential output is estimated to be 0.9 percent above the actual output in 2000. Given that signs of overheating in the economy were absent during this period, these results point to the danger of using mechanical filtering techniques lacking economic underpinnings.

Figure 1.
Figure 1.

Potential Output and the Output Gap: HP Filter 1

(Sample extended with 2001 forecasts)

Citation: IMF Staff Country Reports 2001, 134; 10.5089/9781451819519.002.A001

Figure 2.
Figure 2.

Potential Output and the Output Gap: HP Filter 2

(Sample extended with 4.4 percent annualized growth)

Citation: IMF Staff Country Reports 2001, 134; 10.5089/9781451819519.002.A001

Table 1.

The Estimates for Quarterly Output Gap: HP Filter 1, HP Filter 2 and STM/UC Model

article image

The growth rate for the seasonally adjusted data from the same quarter, previous year.

16. The second version of HP filter gives higher potential output growth rates for recent years. The potential output growth declines from 5,7 percent in 1994 to 3.8 percent in 2000. This method indicates that the output gap was closed in 2000 on average, but in the last quarter it moved back to a negative 2.2 percent.

17. The substantially different results derived from two versions of the HP filter, based on different extensions of the actual GDP series for 2001 Q2–Q4, underscore the end-period bias of the HP filter. This is a crucial shortcoming, given that the extension of the sample is arbitrary to some extent and depends crucially on the judgment of the forecaster.

System estimates of potential output and the NAIRU (STM/UC)

18. For the STM/UC approach, the model is estimated using the quarterly data for the period 1988:3-2000:4. Price inflation is the seasonally adjusted CPI. Both the GDP growth and unemployment rates are seasonally adjusted. The set of variables used as supply shock proxies are the real effective exchange rate, the relative price for imports, the relative price for oil, the productivity and a range of dummies capturing the periods for VAT rate changes and purchase tax changes. Productivity is derived as the industrial production per hour worked on a seasonally adjusted basis. The real effective exchange rate is the trade weighted CPI based real effective exchange rate with respect to the trading partners. All variables except the dummies are in logarithms.

19. The CPI inflation enters the estimation in its first difference since this matches with the constant inflation nature of the NAIRU without having an implicit long-run equilibrium inflation rate. Using the level of the CPI inflation in the Phillips-curve specification leads to an interpretation that, in the absence of supply shocks and when the unemployment rate is equal to the NAIRU, inflation converges to some (constant) long-run equilibrium level. This does not match with the characteristic of the period examined. The inflation target announced by the Bank of Israel declined from a range of 7–10 percent in the early 1990s to a range of 3-4 percent in 2000. Therefore, the targeted inflation rate was in a gradual decline during the estimation period, likely reducing long-run equilibrium inflation gradually as well.

20. The estimated drift coefficient of potential output,α , indicates an annual growth rate of roughly 4.7 percent. The variance estimate for the random walk process generating potential output is insignificant. This suggests that the deterministic trend dominates the evolution of the potential output series. In contrast, the variance estimate for the NAIRU is significantly higher, indicating a substantial time variation of the NAIRU over the estimation period. Figure 2 shows potential output, the output gap, and the cyclical unemployment series, all estimated by the UC model.

21. The potential output growth estimated by the UC model is variable but generally high during 1992–96, a period of large immigration. The estimates lie roughly in the range of 4.5-5.6 percent. During 1997–99 characterized by tight monetary and fiscal policies and weak economic growth, potential growth decelerated gradually from 5 percent to 4.3 percent. The buoyant growth brought about by the high-tech sector in 2000 lifted the potential growth to 4.8 percent. The high productivity growth due to the structural change of the Israeli economy from the traditional sectors towards the high-tech sector offers an explanation for this improvement. The combined effects of the recent Intifada, the decline in the NASDAQ, and the slowdown in the US economy, led to a sharp slowdown in the fourth quarter. The estimated potential output growth decelerated sharply from 4.9 percent in the third quarter of 2000 to 4.2 percent in the first quarter of 2001. While the substantial slack in the economy during the 1997–99 period was largely eliminated by the third quarter of 2000, the estimated output gap has rapidly increased to 3.9 percent of GDP by the first quarter of 2001, following a sharp decline in activity (Table 1).

D. Conclusion

22. This paper presents the estimates for the potential output and the output gap series for the Israeli economy based on the HP filter method and the STM/UC method that makes use of the Philips curve and Okun’s law relations in the context of a system of equations (Apel and Jansson, 1997). The main shortcoming of the HP filter is its reliance on mechanical filtering technique lacking economic underpinnings. The STM/UC method estimates potential output and the NAIRU simultaneously, making use of both the inflation information and the mutual dependence between cyclical output and cyclical unemployment. Moreover, the method takes into account exogenous factors that may influence inflation.

23. Two versions of the HP filter differing in the way the sample is extended to cope with the end-point bias are implemented. The results indicate the sensitivity of the HP filter to the end-point adjustments. Given this and the lack of economic underpinnings of the HP filter method it seems justifiable to interpret the HP filter estimates as a secondary indicator of output gap, and rely more on the STM/UC method.

24. The Israeli economy experienced a sharp slowdown in the fourth quarter of 2000 due to the combined effects of the Intifada, the decline in the NASDAQ and the slowdown in the U.S. economy, following the robust growth in the previous quarters mainly driven by the high-tech sector. The estimates from the two versions of the HP filter and STM/UC methods indicate a negative output gap for the fourth quarter of 2000, expanding further in the first quarter of 2001. However, the magnitude of the gap is substantially different, being −1.1 and −2.2 percent for the first and second versions of HP filter, respectively, versus −3.9 percent for the STM/UC method (all for the first quarter of 2001). For 2000 as a whole, the estimated output gap was 0.9 percent for the first version of the HP filter while the second version indicated that the gap was closed. On the other hand, the STM/UC method estimated an output gap of −1.6 percent. Given the absence of any indicator showing that the economy was overheating during 2000, one may argue that the first version of the HP filter underestimated the potential output growth.

25. The evidence suggested by the STM/UC method shows that the output gap which was almost closed in the third quarter of 2000 widened rather sharply in the fourth quarter of 2000 and the first quarter of 2001. The estimated potential output growth decelerated from 4,9 percent in the third quarter of 2000 to 4.2 percent in the first quarter of 2001. Given mainly the supply-driven nature of the shock to the economy, namely the closure of the territories leading to the absence of the Palestinian labor force and the sharp slowdown in the high-tech sector, it is plausible that not only actual, but also potential output growth has slowed down. Needless to say, potential output and the output gap estimates derived from any method are subject to uncertainty. In the Israeli context, the sharp and exogenous nature of shocks at the end of 2000 complicates the task of estimation further; the deceleration in the potential output growth might be more than estimated by the model, since the estimated potential tends to change smoothly due to the low variance estimate for the random walk process generating potential output. If this is indeed happening, it narrows down the actual output gap compared to the estimated gap, a point that needs to be taken into account in policy analysis. However, given the magnitude of the output gap estimated by the STM/UC model, it seems plausible that, even after discounting this factor, a substantial gap has been created by the sharp deceleration of output at the end of 2000.

Figure 3.
Figure 3.

Potential Output, Output Gap and the Cyclical Unemployment3 Derived from the STM/UC Model

Citation: IMF Staff Country Reports 2001, 134; 10.5089/9781451819519.002.A001

References

  • Apel, M., and P. Jansson, 1997, System Estimates of Potential Output and the NAIRU, Economics Department, Sveriges Riksbank (Stockholm, Sweden).

    • Search Google Scholar
    • Export Citation
  • Apel, M., and P. Jansson, and P. Jansson, 1999, “A Theory-Consistent System Approach for Estimating Potential Output and the NAIRU,” Economic Letters, No. 64, pp. 271275.

    • Search Google Scholar
    • Export Citation
  • Apel, M., and P. Jansson, and P. Jansson, 1999 “System Estimates of Potential Output and the NAIRU,” Empirical Economics, No. 24, pp. 373388.

    • Search Google Scholar
    • Export Citation
  • Gordon, R. J., 1997, “The Time-varying NAIRU and its Implications for Economic Policy,” Journal of Economic Perspectives, No. 11, pp. 1132.

    • Search Google Scholar
    • Export Citation
  • Hodrick, R. J., and E. C. Prescott, 1981, “Post-War U.S. Business Cycles: An Empirical Investigation,” Carnegie Mellon University Discussion Paper No. 451.

    • Search Google Scholar
    • Export Citation
  • Okun, A. M., 1962, “Potential GNP: Its Measurement and Significance,” Proceedings of the Business and Economics Section, American Statistical Association, pp. 98104 (Washington D.C.).

    • Search Google Scholar
    • Export Citation

STATISTICAL APPENDIX

Table 1.

Israel: GDP by Expenditure Components in Current Prices, 1995-2000

article image
Sources: Central Bureau of Statistics, Monthly Bulletin of Statistics.
Table 2.

Israel: GDP by Expenditure Components in Constant Prices, 1995-2000 1/

article image
Sources: Central Bureau of Statistics, Monthly Bulletin of Statistics.

Estimates in 1995 prices derived by chaining each category’s growth rate computed at different base year prices; hence totals do not reflect the sum of their components.

Table 3.

Israel: Investment, 1995-2000

article image
Sources: Central Bureau of Statistics, Monthly Bulletin of Statistic.
Table 4.

Israel: Consumption, 1995-2000

article image
Sources: Central Bureau of Statistics, Monthly Bulletin of Statistics.
Table 5.

Israel: Gross Private Income and Savings, 1995-2000

article image
Sources: Central Bureau of Statistics, Current Briefings in Statistics; and data provided by the Bank of Israel.

Including contribution to National Insurance Institute.

Table 6.

Israel: National Savings, Foreign Savings, and Investment, 1995-2000

article image
Source: Data provided by the Bank of Israel.

Total income defined as GNP plus unilateral transfers from abroad.

Table 7.

Israel: Industrial Production Indices, 1995-2000

article image
Sources: Central Bureau of Statistics, Monthly Bulletin of Statistics; and data provided by the Bank of Israel.

Includes sectors not listed below.

Table 8.

Israel: Labor Market Indicators, 1994-2000

article image
Sources: Bank of Israel, 1999 Annual Report; Central Bureau of Statistics, Monthly Bulletin of Statistics; and data provided by the Bank of Israel.

Aged 15 and above.

Includes unreported foreign workers.

Table 9.

Israel: Employment and Labor Input by Industry, 1994-2000

article image
Sources: Central Bureau of Statistics, Monthly Bulletin of Statistics; and data provided by the Bank of Israel.

Employment figures are annual averages; labor input figures are weekly averages.

The Central Bureau of Statistics definitions and sample changed in 1994. The rows from 1996 on follow the new classification.

Israeli employees.

Data do not sum to total due to an “unknown” category.

Table 10.

Israel: Real Wages, Labor Costs, and Productivity, 1994-2000

article image
Sources: Central Bureau of Statistics and data provided by the Bank of Israel.

Deflated by the consumer price index.

Business sector net domestic product per man-hour estimated from the expenditure side.

Ratio of labor cost per man-hour to labor productivity.

Table 11.

Israel: Real Wage Indices, 1994-2000 1/

article image
Source: Data provided by the Bank of Israel.

Average monthly wage per employee post at constant prices, based on employers’ contributions to the National Insurance Institute, deflated by the consumer price index.

Table 12.

Israel: Consumer Price Index and its Main Components, 1996-2000

(Percentage change, unless indicated otherwise)

article image
Source: Central Bureau of Statistics data, as prepared and compiled by the Bank of Israel.

In tables from previous years this component was calculated as “General Index excluding fruits and vegetables”.

Table 13.

Israel: Selected Price Indices, 1996-2000

(Percent increase during the period)

article image
Source: Central Bureau of Statistics data, as prepared and compiled by the Bank of Israel.

The weights of the components in the Consumer Price Index were changed in 2001.

Public transport, communication services, education, medical services, municipal taxes, electricity, fuel, water, and meat.

Table 14.

Israel: Bank of Israel Accounts, 1995-2000

article image
Sources: Central Bureau of Statistics, Monthly Bulletin of Statistics; and data provided by the Bank of Israel.