## Abstract

The paper suggests that several factors, besides tight monetary policy, may well have contributed to the high real interest rates that have been observed in Israel. The paper examines the impact of unanticipated changes in nominal interest rate monetary policy shocks on a number of variables in Israel. The paper provides evidence that the inflation expectation measure derived from the capital markets tends to overstate the trend in actual inflation. The paper also provides statistical data on the economic indices of Israel.

## II. Effects of Monetary Policy in Israel^{1}

1. Since 1992 Israel has operated in an inflation targeting framework in which the primary goal of monetary policy is to achieve a specified inflation target. Typically, this target is specified as a range of acceptable outcomes. At times, the Bank of Israel may find that inflation is likely to exceed the target, and the question arises as to the extent to which interest rates would need to be raised in order to ensure that inflation comes in within or below target. In this connection, it is useful to have some sense of the order of magnitude of the short-run impact that such a monetary tightening would have on output. At other times, the Bank of Israel may find that inflation is likely to come in below (or at the lower end of) the inflation target at the same time as output is below its potential level. Under these circumstances, would a small, temporary cut in nominal interest rates likely have a positive effect on growth in the short term? Just as importantly, would it generate inflationary effects that are incompatible with the paramount goal of achieving the inflation target? How rapidly would these effects be realized, if at all? This paper seeks to contribute to the debate over these issues by examining empirically the influence of interest rates on prices, the exchange rate, output and other variables.

2. Section A of the paper provides some details on the behavior of inflation, interest rates and other key variables in the post-stabilization period. Section B reviews the relevant literature on the impact of monetary policy on output, inflation and the exchange rate in Israel and selected other countries. Section C contains the empirical work, which makes use of structural vector autoregression (VAR) involving a number of macroeconomic variables to investigate the interaction of key variables in Israel. Section D concludes.

### A. Inflation In Israel: Macroeconomic Background

3. Over the last decade Israel has transformed itself from a high- to a low-inflation economy. Annual CPI inflation peaked at 373 percent in 1984, before a successful stabilization program reduced its level to 5.2 percent in 1999 and to just 1.3 percent on a 12-month basis as of January 2000. The stabilization program undertaken by the authorities in mid-1985 consisted of several elements: a substantial reduction of the fiscal deficit -via cuts in expenditure and increases in taxation; a tightening of monetary policy, featuring increases in reserve requirements and the discount rate and the elimination of central bank financing of the budget deficit; the establishment of a nominal anchor for the sheqel (of NIS 1.5 per U.S. dollar); and the temporary suspension of wage indexation accompanied by a three-month freeze on the prices of virtually all goods and services.^{2} Inflation fell rapidly following the introduction of the package, dropping to just 20 percent on a 12-month basis by the second half of 1986. Since then the decline in the inflation rate has been very slow, with inflation generally remaining in the range of 15-20 percent between 1986 and 1991, dropping to about 10 percent between 1992 and 1997, and finally falling to the mid-single digits since mid-1997 (see Figure 1).

4. Figures 1 through 5 provide information on consumer price inflation, real output growth, inflation in imported goods prices (in U.S. dollar terms), interest rates, the nominal and real exchange rate versus the U.S. dollar, and the unemployment rate over the period 1989–99. As shown in Figure 1, real output growth has been fairly high over the period, averaging about 4½ percent on an annual basis, but has also been subject to considerable variation, reaching peaks of about 10 percent in the first quarter of 1992 and the third quarter of 1995, both of which were short-lived. The very rapid rates of growth achieved in the early 1990s reflect in part the very rapid influx of immigrants during that period. Figure 1 also suggests that since the early 1990s there has been some tendency for output growth and consumer price inflation to move together. This trend was particularly pronounced in period 1992–93, when inflation and output growth both declined sharply, and in the period since about 1996, when inflation and output growth have both been on a downward trend. With the exception of two isolated instances in 1991 and 1995, import price inflation has been moderate (Figure 2). From mid-1996 through mid-1999 import prices actually fell, which may in part explain the improved overall inflation performance during those years, but import prices have rebounded sharply since then with no attendant increase in domestic inflation.

**Israel: Changes in Import Prices, 1989-98**

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

**Israel: Changes in Import Prices, 1989-98**

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

**Israel: Changes in Import Prices, 1989-98**

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

5. Figure 3 shows the behavior of the nominal interest rate differential relative to the United States, which is defined here as the difference between the Bank of Israel’s discount rate and the Fed Funds rate in the United States. The nominal differential remained in the neighborhood of around 8 percent over much of the period, but since mid-1996 has been on a downward path (although this was temporarily reversed in late 1998 in the wake of the Russia/LTCM crisis). The chart also shows the behavior of the real Bank of Israel discount rate, defined here as the interest rate at time *t* minus the actual 12-month rate of consumer price inflation in period *t+4*. The nominal exchange rate has generally depreciated smoothly over the period 1989–99, although the Russia/LTCM crisis did lead to a sharp increase in the rate of depreciation at the end of 1998, part of which was subsequently offset (Figure 4). However, the real effective exchange rate (CPI-based) appreciated over much of the period, other than in the fourth quarter of 1998.^{3} Finally, following a steady rise in unemployment in the early 1990s, the unemployment rate declined steadily from a peak of 11.4 percent in the first quarter of 1992 to 6.4 percent in the second quarter of 1996 (Figure 5). Since then, and coincident with the slowdown in real output growth, the rate of unemployment has tended to increase.

**Israel: Nominal Interest Rate Differential vs. United States and Real Interest Rate, 1989-99**

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

**Israel: Nominal Interest Rate Differential vs. United States and Real Interest Rate, 1989-99**

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

**Israel: Nominal Interest Rate Differential vs. United States and Real Interest Rate, 1989-99**

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

**Israel: Nominal and Real Exchange Rate versus US dollar, 1989-99**

(1988:4=1)

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

**Israel: Nominal and Real Exchange Rate versus US dollar, 1989-99**

(1988:4=1)

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

**Israel: Nominal and Real Exchange Rate versus US dollar, 1989-99**

(1988:4=1)

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

**Israel: Unemployment Rate, 1989-99**

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

**Israel: Unemployment Rate, 1989-99**

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

**Israel: Unemployment Rate, 1989-99**

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

### B. Literature Review

6. The link from interest rate policy to economic activity and inflation has long been a subject of interest for both academic economists and policy makers. In principle, changes in interest rates can affect prices and real output through a variety of channels, which are known collectively as the “transmission mechanism.”^{4} In general, changes in policy-determined interest rates can affect real output through three channels. First, higher real interest rates affect demand directly through a variety of ways. For example, an increase in interest rates reduces fixed and inventory investment by firms and makes debt-financed expenditures (such as for consumer durables) by individuals more expensive. It also increases the return to saving and produces a negative wealth effect by reducing the value of household and firm assets (the future returns from which are now discounted at a higher interest rate), both of which will reduce consumer expenditure. Second, an increase in interest rates will tend to lead to an appreciation of the exchange rate in foreign currency markets, shifting relative prices in favor of imported goods, with negative repercussions for domestic output. Finally, changes in official interest rates can affect output by altering individuals’ expectations about the future course of the economy, which obviously will have implications for their spending and investment decisions.

7. Similarly, the transmission mechanism from real interest rates to inflation works through more than one channel. First, as noted above, an increase in domestic interest rates will tend to lead to an appreciation of the exchange rate, making imported goods cheaper (and possibly forcing domestic producers of tradables to reduce their prices as well). This will have a direct, negative impact on the domestic price level. Second, to the extent that higher interest rates lead to a reduction in domestic output and demand (for the reasons discussed above), this will tend to put downward pressure on domestic employment and hence wages, also contributing to a reduction in prices. Finally, as was the case with output, increases in interest rates may have an impact on the expectations of individuals and firms regarding the future course of the economy, which would also affect the price level and inflation.

8. Studies of the transmission mechanism, and of the impact on monetary policy on output and prices more generally, tend to concentrate on two related questions. First, what impact, if any, does monetary policy (and especially unanticipated changes in monetary policy) have on key variables? Second, how long is the lag between changes in monetary policy and changes in output, inflation, and other variables of interest? Summarizing the now substantial body of work addressing these questions, Christiano and others (1998) state

[T]here is considerable agreement about the qualitative effects of a monetary policy shock in the sense that the inference is robust across a large subset of the identification schemes that have been considered in the literature. The nature of this agreement is as follows: after a contractionary monetary policy shock, short-term [nominal] interest rates rise, aggregate output, employment, profits and monetary aggregates fall, the aggregate price level responds very slowly, and various measures of wages fall, albeit by very modest amounts. In addition, there is agreement that monetary policy shocks account for only a very modest percentage of the volatility of aggregate output; they account for even less of the movements in the aggregate price level. (page 4).

The increase in domestic interest rates would also be expected to lead to an appreciation of domestic currency in foreign exchange markets, a result that is confirmed by Eichenbaum and Evans (1995) for the United States, Clarida and Gertler (1997) and Smets and Wouters (1999) for Germany, and Cushman and Zha (1997) for Canada.

9. With respect to the lags in the transmission mechanism, Christiano and others (1998) find that it takes about two quarters for a monetary policy shock to affect output in the United States, with the largest decline occurring about 18 months after the shock occurs. They also find that it takes about 18 months for the general price level (the GDP deflator) to begin to decline. Similarly, policy simulations undertaken by the Bank of England (1999) find that a monetary contraction leads reasonably quickly to a decline in output that reaches a maximum (of between 0.2 percent and 0.4 percent for each 100 basis points increase in interest rates) after about five quarters. Inflation does not begin to decline until the second year following a monetary contraction, and the maximum impact—a decline of between 0.2 and 0.4 percent in the inflation rate for each 100 basis points increase in interest rates—does not occur until the ninth quarter after the policy shock. Eichenbaum and Evans (1995) find the U.S. dollar exchange rate appreciates for a long period of time following a monetary policy tightening, with the maximum impact being felt 24 to 39 months following the monetary contraction, depending on the particular exchange rate being examined. By contrast, Cushman and Zha (1997) find that for Canada, a monetary tightening leads to an immediate appreciation of the nominal and real exchange rates vis-à-vis the U. S. dollar, but that most of the appreciation has worn off after 12 months.

10. In simulations conducted using the macroeconometric models maintained by the central banks of a number of euro-area countries (reported in BIS, 1995), cuts in nominal policy interest rates in Germany, Italy, and France are found to have their maximum impact on real output in the year following the interest rate cut, with a temporary 100 basis point cut in short-term interest rates leading to an increase in output of 0.4-0.5 percent of GDP. The maximum impact on inflation (which as in the United Kingdom is on the order of 0.2-0.4 percentage point for a 100 basis point interest rate cut) occurs two years after the cut in Germany and France, but in the same year as the interest rate cut in Italy. The difference presumably reflects the fact that the interest rate cut affects the exchange rate much more rapidly in Italy than in the other two countries: the maximum impact of an interest rate cut is felt in the same year as the cut in Italy, but only in the following year in Germany and France.

11. A number of papers have examined empirically the impact of interest rates on other variables in Israel. Azoulay and Elkayam (1999) develop a structural model of the Israeli economy using quarterly data covering the period 1988 to 1996. They find that an increase in real interest rates (defined as the nominal interest rate minus expected inflation, itself a function of lagged inflation rates) leads to a temporary reduction in inflation. In addition, an increase in real interest rates leads in the short term to a reduction in output growth but in the medium term to higher output growth. However, this latter result likely reflects the fact that the authors hold the nominal exchange rate constant in their analysis: the fixed nominal exchange rate combined with the decrease in the inflation rate induces a real depreciation of the sheqel that serves to spur output growth. Leiderman and Bar-Or (1999) estimate a reduced-form equation for inflation (using data from the period 1990 to 1998) that includes current and lagged real interest rates, here defined as the nominal rate minus expected inflation derived from the difference between the yield on inflation-indexed and noninflation-indexed sovereign debt. They find that a permanent 1 percentage point increase in real interest rates leads to a decline of about 1.7 percentage points in the inflation rate.^{5} Leiderman and Bar-Or also find that the pass-through from the nominal exchange rate to the price level is about one-half or that a 1 percent depreciation of the sheqel vis-à-vis the dollar leads to a 0.5 percent increase in the price level. However, the extent of the pass through is sensitive to the business cycle, with the impact of nominal depreciations on the price level being somewhat lower when the economy is at less than full employment.

12. Swagel (2000) estimates a structural VAR involving real output, the real exchange rate, and the margin between real interest rates in the United States and Israel, among other variables.^{6} He finds that a 100-basis point increase in real interest rates leads to a decline of about ¼ percent in the price level, to an appreciation of about 1 percent in the sheqel-dollar exchange rate, and to a decline of about 0.2 percent in real output. He also finds that about 20 percent of the impact of a depreciation of the sheqel is passed through to higher prices. Using a different approach—the “narrative approach” as developed by Romer and Romer (1989)—De Fiore (1998) examines the impact of three episodes of monetary tightening in Israel during the 1990s.^{7} Comparing actual levels of industrial production, the exchange-rate, and other variables after the tightening to those that would have been forecasted in the absence of the policy tightening, she finds that monetary policy had relatively small effects on industrial production and that a tightening of monetary policy was always followed by an appreciation of the nominal exchange rate, although only in two of the three episodes that she studied did an appreciation of the real exchange rate also occur.

13. Beenstock and others (1994), using a structural model of the Israeli economy covering the period 1962-90, find that some 80 percent of the impact of a nominal devaluation is passed through to higher prices. Idiosyncrasies in their model result in interest rate policy having little impact on real output. Leiderman and Razin (1988) use a VAR to examine the interaction of the trade deficit, the exchange rate, the inflation rate and the money supply. They find that nominal exchange rate shocks have temporary effects on the inflation rate and permanent effects on the price level.

### C. Modeling the Economy Using Vector Autoregression Methods

14. VARs have become the tool of choice for investigations of the monetary transmission mechanism, and are the technique adopted in this paper as well. One of the advantages of the VAR approach is that it eliminates the need to specify a specific structural model of the economy by modeling every endogenous variable in the system as a function of its own lagged values and those of the other endogenous and exogenous variables, as well as any deterministic variables included in the model. A VAR with *k* endogenous variables, and *p* lags can be expressed in the form:

where Z_{i}; is a *k* × 1 vector of endogenous variables, Π_{i} is a *k* × *k* matrix of regression coefficients, *A* is a *k* × *k* matrix whose diagonal elements are normalized to equal 1 but which may have arbitrary of diagonal elements, and ε_{t}, is a *k* × 1 vector of orthogonal error terms.

15. The analysis of the model consists of three stages. In the first stage, a decision is made about the variables and number of lags to include in the model, and estimation is conducted using ordinary least squares. Once the model is estimated, the resulting Π matrices can be inverted to express *Z*_{t} as a function of current and lagged regression errors *u*:

Here the row i, column j element of *G*_{s}, *G*_{tj,s} gives the response of *Z*_{i,t+s}, to a unit shock to *u*_{j,t}. In practice, the meaning of the regression errors *u*_{t} is likely to be ambiguous because the possibility of contemporaneous correlation among variables included in the model means that each of the *u*_{i,t} may potentially reflect the effects of all of the fundamental shocks ε_{i,t} that are assumed actually to be driving variations in the economy. The second step of the analysis is therefore to make assumptions about which variables are contemporaneously correlated, and use these assumptions to decompose the regression errors into the orthogonal (i.e., independent) errors ε_{t}. Accordingly, although the use of VARs eliminates the need to specify a full model of the economy, it is not the case that the technique allows the researcher to be completely agnostic about how the variables of interest interact. The final step of the analysis is then to construct impulse response functions that chart the responses of variables over time to unit increases in the elements of ε_{t}.^{8} We interpret a non-zero realization of the orthogonal error term related to the nominal interest rate differential as an unanticipated, temporary monetary policy shock: an increase or decrease in the interest rate differential that could not have been predicted based on the information available at time *t-1*. Even though the increase or decrease in the rate may persist for more than one period because of the autoregressive structure of the model, the shock itself is temporary because the error terms are assumed to be serially uncorrelated.

16. The specification examined in this paper includes nine variables, and can be thought of as encompassing five equations (although in keeping with the VAR structure, all endogenous variables are regressed on lagged values of all the variables in the system):

where *XR* equals the nominal exchange rate vis-à-vis the U.S. dollar;

*PISR*equals the consumer price index in Israel;*PUSA*equal the consumer price index in the United States;*RDIFF*equals the nominal interest rate differential between the Israel and the United States;*W*equals the nominal wage in Israel;*PM*equals the price of Israeli imports in U.S. dollars;*M*equals the money stock;*Y*equals real output in Israel; and*U*equals the unemployment rate in Israel.

Thus, Equation 3 describes the evolution of the nominal exchange rate vis-à-vis the US dollar as a function of prices in the United States and Israel and the nominal interest rate differential. In Equation 4, the price level in Israel is modeled as a markup over wages and imported goods prices (themselves a function of the exchange rate and the dollar price of imported goods). In Equation 5, money demand is modeled as a function of prices, real output, and the interest rate differential, while wages are modeled in Equation 6 via a Phillips’ Curve relationship with prices and the unemployment rate. Finally, Equation 7 states that real output is a function of the real exchange rate and the real interest rate differential. The model thus contains nine variables: consumer prices in Israel and the United States; the nominal differential between the Bank of Israel’s discount rate and the Fed Funds rate in the United States; the nominal NIS-U.S. dollar exchange rate; Ml; nominal wages; the seasonally-adjusted rate of unemployment;^{9} real GDP; and the unit value (expressed in dollar terms) of imported goods.

17. In the model, the primary transmission mechanism from interest rates to inflation is the exchange rate: an increase in the interest rate differential between Israel and the United States leads to an appreciation of the exchange rate (through Equation 3), reducing the price of imported goods and the overall price level. In addition, higher interest rates affect output through two channels: directly, through the influence of interest rates on investment and consumption; and indirectly, through an appreciation of the exchange rate that makes imports less expensive and exports more costly. Lower output, in turn, would put downward pressure on wages, which should further reduce the price level. The relative openness of the Israeli economy suggests that the transmission mechanism from interest rates to prices and output should, operate fairly rapidly, as the tradables sector accounts for a large share of GDP and the consumption basket.

18. Most previous studies of the impact of monetary policy on inflation and real variables in Israel have examined the effect of increases in real interest rates. By contrast, in this chapter we examine the impact of a tightening of nominal rates. The decision to concentrate on the impact of an increase in the nominal interest rate differential rather than the real rate differential reflects the fact that the policy instrument available to the authorities is in fact the nominal interest rate. Presuming that the authorities have the ability to manipulate the real interest rate begs the question of the impact of changes in the interest rate on inflation, which in fact is one of the key issues examined here. However, including the nominal interest rate in the VAR framework introduces complications of its own. Specifically, by including the nominal interest rate differential in the VAR, we are modeling it as an endogenous variable that responds to changes in other variables over time. At the same time, we want to treat the Bank of Israel’s discount rate (and the associated differential versus the exogenous Fed Funds rate) as a policy variable that is manipulated by the authorities, and to chart the response of other variables to an autonomous shift in the differential. It is important to recognize at the outset that there may be some tension between the way the nominal interest rate differential is modeled and the way we wish to interpret it. This problem does not arise in studies that use the real interest rate differential as a variable, because even if the nominal differential is exogenous, the real differential will be endogenous due to the endogeneity of inflation.

19. Augmented Dickey-Fuller Tests were conducted to determine the order of integration of each of the variables. While the nominal interest rate differential was found to be stationary, all other variables were found to be integrated of order 1. Subsequently, two-step Engle-Granger tests were conducted to look for cointegrating relationships among the other variables included in the model. The results of these tests (which are not reported here) indicate the presence of cointegrating relationships in all of the equations of the model except for that linking Israeli prices, wages, and the unemployment rate. Thus, all of the endogenous variables included in the model are either stationary or are cointegrated with at least one other variable with the exception of the unemployment rate.^{10} This left three options: to include the unemployment rate in the VAR anyway; to drop the unemployment rate entirely from the model; or to difference the unemployment rate to make it stationary.^{11} In the event, all three options were pursued and produced results that were generally equivalent. In what follows we report only the version in which the first difference of the log of the unemployment rate is included in the VAR.

20. Based on the assumption that developments in Israel have limited impact on the economy of the United States and other large industrial economies, the U. S. price level and the unit value of imports in dollar terms are treated as exogenous in the analysis. In other words, while price developments in the United States are allowed to have an impact on the Israeli economy, changes in the modeled variables are treated as having no impact on domestic or export prices in the United States.

21. The model was estimated using quarterly data covering the period 1988–98, to restrict attention to the post—stabilization period.^{12} In addition to the variables listed above, a constant term and three seasonal dummies were included in the regression. In addition, following the specification contained in Swagel (2000) a dummy term was included to allow for a break in the mean starting in the first quarter of 1992 as a result of the wave of Russian immigration. With the exception of the dummies and the nominal interest rate differential, all variables were expressed in natural logarithms.^{13}

22. As noted above, creating impulse response functions that chart the impact on a set of variables from innovations in another requires moving from the vector of regression errors *u(t)* to the vector *ε(t)* of orthogonal error terms (or “innovations”) that are assumed actually to be driving variations in the economy. Carrying out this decomposition of error terms requires the imposition of restrictions on the contemporaneous interactions among variables. One common set of restrictions is to apply a recursive ordering of the variables. Instead, following Bernanke (1986) and Sims (1986), we make assumptions about which variables are likely to display contemporaneous correlation, and use these restrictions to decompose the errors. Specifically, we assume that prices respond to innovations in the exchange rate and in wages within the same quarter, while output is affected contemporaneously by unemployment innovations.^{14} The exchange rate is assumed to respond within the same quarter to innovations in all variables except the unemployment rate, while wages respond to within-quarter innovations in only the price level and unemployment. Thus, all variables have a contemporaneous impact on the price level either directly (in the case of wages and the exchange rate), or indirectly via their effect on wages and/or the exchange rate. Current period money demand is directly affected by innovations in prices, output, and the interest rate differential, while the interest rate differential and the unemployment rate respond to innovations in other variables with a one period lag. A *x*^{2} test of the 8 overidentifying restrictions has a p-value of 43 percent, meaning that the restrictions are not rejected at a moderate significance level.

23. The estimated contemporaneous coefficients (with standard errors in parentheses) are as follows:

*Prices*

*Exchange rate*

*Wages*

*Output*

*Money demand*

*Interest rates*

*Unemployment*

Where *XR* equals innovations in the exchange rate;

*P*equals innovations in the price level;*RDIFF*equals innovations in the nominal interest rate differential;*W*equals innovations in the nominal wage;*U*equals innovations in the unemployment rate;*Y*equals innovations in real output; and*M*equals innovations in nominal M1.

As an aid in interpreting these results, the first equation states that innovations in the price level are found to depend positively on innovations in the exchange rate and negatively on innovations in wages as well as on the error term *e _{P}*; the second states that innovations in the exchange rate are found to depend on innovations in prices (positively), wages (positively), the interest rate (negatively), output (negatively), and money (positively), as well as on the error term

*e*.

_{XR}24. Most of the contemporaneous correlations are of the expected sign, although many are not statistically significant. The low t-statistics imply that there is a large region of parameter space that implies a covariance matrix for the VAR innovations similar to that found in the data. Nevertheless, there are some interesting results. Innovations in the domestic price level are found to contemporaneously correlate positively with innovations in the nominal exchange rate (i.e., with depreciations of the sheqel) and vice versa, while innovations in the nominal interest rate differential (which can be interpreted as a tightening of monetary policy) are found to correlate negatively with innovations in the exchange rate (or positively with exchange rate appreciations). Money demand innovations are found to correlate positively with real output innovations and negatively with real interest rate innovations. Two puzzling correlations involve the unemployment rate, which is found to correlate positively (albeit weekly and not significantly) with real output and wages. These findings could reflect the changing composition of the labor force with the shift of production toward more high technology areas, as wages and output rise overall but lower-skilled workers experience greater unemployment. The negative (albeit statistically insignificant) correlation between wage shocks and price level shocks in Equation 8 is also counterintuitive, although price level innovations do have a positive contemporaneous impact on wages in Equation 10.

25. Figures 6 and 7 illustrate the response of the exchange rate, money demand, inflation and output growth to an unanticipated monetary tightening, as represented by a positive realization of *e _{RDIFF}*. The figures represent the dynamic effects of the contemporaneous correlations identified in Equations 8 through 14, augmented by the impact on each of the variables of changes in lagged variables. The results are consistent with the characterization of the transmission mechanism in the description of the model. As expected, higher nominal interest rates lead to an appreciation of the sheqel that tends to drive down the price level. In addition, higher interest rates reduce the demand for money. More specifically, the results suggest that a one percentage point (100 basis points) increase in the nominal interest rate differential between Israel and the United States leads to an immediate appreciation of the nominal exchange rate of 0.5 percent and an immediate 1 percent decrease in the demand for money. By the first quarter following the monetary tightening, the exchange rate has appreciated by nearly ¾ of a percentage point, while money demand has declined by about 1.7 percent. Some 12 quarters following the interest rate increase, the nominal exchange rate has appreciated by about ¾ percentage point, and money demand has fallen by about 1¾ percent.

**Israel. Impulse Responses of Variables to I Percentage Point Increase in Nominal Interest Rates**

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

**Israel. Impulse Responses of Variables to I Percentage Point Increase in Nominal Interest Rates**

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

**Israel. Impulse Responses of Variables to I Percentage Point Increase in Nominal Interest Rates**

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

**Israel. Impulse Responses of Inflation and Real Output Growth to 1 Percentage Point Increase in Nominal Interest Rates**

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

**Israel. Impulse Responses of Inflation and Real Output Growth to 1 Percentage Point Increase in Nominal Interest Rates**

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

**Israel. Impulse Responses of Inflation and Real Output Growth to 1 Percentage Point Increase in Nominal Interest Rates**

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

26. After an initial increase, the inflation rate declines with the annual rate of inflation falling by about ¼ percentage point three quarters following the shock. The impact of a monetary tightening on *inflation* is essentially zero by the tenth quarter following the shock. The initial jump in the inflation rate following an increase in nominal interest rates may reflect reverse causality: the monetary authorities are likely to respond to an increase in the inflation rate by raising nominal interest rates, resulting an a positive short-run correlation between the two. The fact that the impact on inflation turns negative by the second quarter—with the maximum impact being felt in the third quarter—suggests that the transmission mechanism from interest rates to inflation operates somewhat more rapidly in Israel than in France, Germany, the United States and the United Kingdom, where other studies have found that it can take up to two years for an interest rate hike to have its maximum impact on the price level. The more rapid operation of the transmission mechanism in Israel could reflect a number of factors, including the very open nature of the economy (and the consequent importance of the exchange rate in determining the price level) and the relatively high level of wage indexation that persists as a legacy of the high inflation past.

27. Real output growth is positive in the first three quarters following an increase in interest rates (Figure 7), but turns negative by the fourth quarter following the monetary tightening. Again, this may reflect in part reverse causality, in the sense that the authorities may tighten monetary policy when output growth accelerates, leading to a short-run positive correlation between interest rates and output growth. The fact that it takes until the fourth quarter following a tightening for output growth to turn negative suggests that the transmission mechanism from interest rates to real output growth is somewhat slower than that from interest rates to inflation, a finding that differs from those of studies of other countries, where the transmission mechanism from interest rates to output has typically been found to operate more rapidly than that to prices. The maximum impact of a tightening of monetary policy on output occurs 7–9 quarters following the policy shock (during which the growth rate of real output is about 0.4 percent reduced), which is slightly slower than in other countries, where the maximum effect was found to occur after 5–6 quarters. The impact of monetary policy on real output growth is transitory, falling essentially to zero after about 28 quarters.

28. Figure 8 provides information on the implied impulse responses of real variables. A 1 percentage point increase in nominal interest rates leads to an immediate increase in the real interest rate differential^{15} (that briefly exceeds 1 percentage point) and to a sharp appreciation of the real exchange rate. Some 12 quarters following the nominal interest rate rise, the real exchange rate has appreciated by about ½ percent and the real interest rate differential has increased by about 2/3 percentage point.

**Israel. Impulse Responses of Variables to 1 percentage Point Increase in Nominal Interest Rates**

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

**Israel. Impulse Responses of Variables to 1 percentage Point Increase in Nominal Interest Rates**

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

**Israel. Impulse Responses of Variables to 1 percentage Point Increase in Nominal Interest Rates**

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

29. Moving to the impact of a shock to the exchange rate, the results indicate that a 1 percentage point increase in the exchange rate (i.e., a depreciation of 1 percentage point in the sheqel-dollar exchange rate) leads in the period immediately following to an increase of 0.1 percent in the price level. The pass-through of the exchange rate depreciation to the domestic price level proceeds for several following quarters, reaching a cumulative 40 percent (that is, an increase of 0.40 percentage points in the price level) after 12 quarters. The impact on inflation dies out by the twelfth quarter following the shock. The depreciation has a mild, positive impact on real output, albeit with a lag, with output growing by a cumulative one-third of a percentage point after 12 quarters (see Figure 9).

**Israel. Impulse Responses of Variables to a 1 Percent Depreciation of the Sheqel**

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

**Israel. Impulse Responses of Variables to a 1 Percent Depreciation of the Sheqel**

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

**Israel. Impulse Responses of Variables to a 1 Percent Depreciation of the Sheqel**

Citation: IMF Staff Country Reports 2000, 062; 10.5089/9781451819502.002.A002

30. Again, these results are consistent with the transmission mechanism from interest rates to output outlined in the description of the model. First, as expected, higher real interest rates lead to an appreciation of the exchange rate and a decline in real money balances (while both *M* and *PISR* fall, money demand falls more rapidly than the price level so *M/PISR* declines by about 1½ percent). Higher interest rates are also associated with a decline in real output growth, meaning that the net impact of all these factors on output growth is indeed negative. Moreover, the impact of an exchange rate appreciation on output growth is negative, while positive liquidity shocks are expansionary. These results suggest that the specification of the model captures the impact of monetary policy on real output relatively well.

31. Overall, the results of the VAR suggest that monetary policy can affect real interest rates, the real exchange rate, and real money demand. While the short-run impact of monetary policy on real output in the VAR is anomalous—with monetary tightening leading to faster output growth—increases in interest rates are also associated with lower real output growth in the medium term. However, the estimated elasticities are relatively small: for example, it would take nearly a 200 basis point shock to nominal interest rates to achieve an immediate 1 percent appreciation of the sheqel versus the dollar. Over a longer-term horizon, uncovered interest parity holds approximately, as the increase in the interest rate differential in the periods following the shock is approximately equal to the appreciation of the exchange rate of the sheqel.

32. As expected, the results also suggest that monetary policy has an impact on reducing inflation and the price level. However, once again the estimated impact of a monetary tightening on the price level is rather small, with a 100 basis point tightening of interest rates leading to a decline in the inflation rate of about 0.3 percentage point three quarters later. Such a finding may appear difficult to reconcile with the sharp decline in inflation that has occurred over the last decade and that has largely been attributed to the tight monetary policy pursued over that period. To be sure, other factors besides monetary policy have been at work in influencing the decline in inflation that occurred over the last decade. Among these may be variables that are not well captured by the VAR setup but are likely to be of importance, including improvements in the credibility of macroeconomic policy or reductions in inflationary expectations that are independent of changes in the variables included in the study. Although the sample period was chosen to minimize the degree of structural change in the economy, structural changes may also have played a role in reducing the rate of inflation. For example, it has been conjectured that the widening of the sheqel’s exchange rate band and the Bank of Israeli’s policy of not intervening in the foreign exchange market has played an important role in containing inflation by reducing the sensitivity of prices to exchange rate movements.^{16} To argue that other factors have also influenced the decline in inflation in no way implies that monetary policy did not also play an important role. More generally, the analysis conducted in this paper examines the impact of temporary and unanticipated shocks to monetary policy. The impact of permanent changes to the systematic components of monetary policy that seek to achieve a permanent, secular decline in the rate of inflation may be quite different from the short-term effects examined here.

33. The results of the VAR are quite similar to those obtained by Swagel (2000), who finds that a one percentage point increase in *real* interest rates leads to a sustained decrease of about ¼ percentage point in the price level, a 1 percentage point appreciation of the exchange rate, and a 0.2 percent decline in the growth rate of output. He also finds that about 20 percent of the impact of a depreciation is passed through to higher prices. The results are similar as well to those of Bufman and Leiderman (1997), Leiderman and Bar-Or (1999) and Azoulay and Elkayam (1999), who also find that higher real interest rates lead to declines in the price level. The rate of pass-though is similar to that obtained by Leiderman and Bar-Or (1999), who report an average pass-though of about 45 percent. Leiderman and Razin (1988) find that a nominal exchange rate shock leads to a temporary increase in inflation and a permanent increase in the price level, which is also consistent with the results obtained here. Beenstock and others (1994) find that some 80 percent of the impact of a nominal devaluation is passed through to higher prices, twice the rate of pass-through found in this study. The higher rate of pass-through they find compared to that obtained in this study and in Swaget (2000) and Leiderman and Bar-Or (1999) could constitute some evidence in support of the hypothesis that changes in the Bank of Israel’s exchange rate policy have led to a reduction in the sensitivity of inflation to movements in the exchange rate. De Fiore (1998) finds that interest rate increases have only a minor impact on real output, and that monetary tightening leads to an appreciation of the nominal exchange rate, consistent with the results obtained here. In general, the results are also consistent with those of studies of the transmission mechanism in G7 countries (see Christiano and others (1998) and BIS (1995), for example, as well as the other references in paragraphs 9 and 10) with respect to the magnitude of the impact of an interest rate shock on inflation and output although, as noted above, the transmission mechanism appears to operate somewhat more rapidly with respect to prices and more slowly with respect to output in Israel than in these other countries.

34. Concerns about the treatment of the nominal interest rate differential as both an endogenous and a policy variable are alleviated to some extent by the empirical results. By assumption, shocks to the other variables have no contemporaneous impact on the nominal interest rate differential. However, even lagged values of the other endogenous variables have little feedback on the nominal interest rate differential: four quarters after a 100 basis point tightening of nominal rates, the nominal interest rate differential stands at 103 basis points. Even after 40 quarters, the differential remains above 70 basis points, suggesting that the nominal interest rate is only weakly affected by the other variables (consistent with its interpretation as a policy variable).

### D. Conclusions

35. This paper has examined the impact of unanticipated changes in nominal interest rates—monetary policy “shocks”—on a number of variables in Israel. The results suggest an unanticipated monetary tightening will generate temporary declines in both inflation and real output growth and a nominal and real appreciation of the exchange rate. The results also suggest that about 40 percent of the impact of an exchange rate depreciation are passed on in higher consumer prices. These results (and their magnitudes) are broadly consistent with those obtained by other authors in studies not only of Israel but also of a number of G7 countries. However, the maximum impact on inflation in Israel is felt just three quarters following the policy shock, much more rapidly than has been found in studies of France, Germany, the United Kingdom and the United States. On the other hand, the maximum impact on output is felt about two years after the interest rate hike, slightly later than in these other countries.

36. These results have some important implications for the conduct of monetary policy in Israel. First, the very rapid transmission of interest rate movements to inflation means the Bank of Israel is in a better position to use interest rate hikes to counteract inflationary impulses than may be the case for central banks in countries where the transmission mechanism takes longer to operate. In addition, the relatively small impact of interest rates on inflation means that when output is well below potential and inflation is expected to be below or at the bottom end of the inflation target, there may be scope for the central bank to implement a cautious easing of monetary policy to provide some mild support to output growth without necessarily compromising the inflation target.

37. Recent events contribute a sort of natural experiment (conveniently falling almost entirely outside the sample period for the empirical work) that broadly confirms the magnitudes of the estimated relationships among interest rates, inflation and the exchange rate. In November 1998 the nominal interest rate differential between Israel and the United States increased by 425 basis points, as the Bank of Israel raised the discount rate by 400 basis points in response to a spike in inflation following the Russia/LTCM crisis, while the U.S. Federal Reserve reduced the fed funds rate by some 25 basis points. Nine months later, the rate of consumer price inflation had fallen by about 2 percentage points (from 8.2 percent to 6.3 percent on a year-on-year basis), compared to the 1 percent decline predicted by the VAR. The underlying rate of inflation—which excludes housing and fruits and vegetables—fell by exactly 1 percentage point (from 7.5 percent to 6.5 percent on a year-on-year basis). In the first quarter following the policy tightening the nominal sheqel-dollar exchange rate appreciated by about 3.8 percent (from an average of NIS 4.23 per dollar in November to NIS 4.07 per dollar in February), compared to an estimated appreciation of 3.2 percent in the VAR. Of course, the experience of 1998–99 does not precisely replicate the experiment conducted in this paper. First, some of the increase in the nominal interest rate differential may have been anticipated by agents. Perhaps more importantly, while the experiment in this paper traces the impact of a shock to monetary policy in an environment free of subsequent shocks to any variable, in fact all variables experience shocks—and react to those occurring to other variables—all the time. Nevertheless, the broad similarity of the empirical estimates of the impact of monetary policy on inflation and the exchange rate to the rough ratios calculated from the experience of late 1998 and 1999 do provide some additional evidence in support of the empirical findings.

38. While the impact of monetary policy on real output is found to be small, the implied short-term trade off between output and inflation (the “sacrifice ratio”) appears less so: according to the empirical results, an increase in interest rates sufficient to reduce the inflation rate by, say, 5 percentage points would lead to a cumulative decline in output on the order of about 8 percent. Does this result imply that in the absence of the roughly 10 percentage point decline in the annual inflation rate that occurred between 1993 and 1999, output would be 15 percent higher today than its current level? Such an inference would not be warranted. First, it is worth reemphasizing that while monetary policy was the main driving force behind the decline in inflation, other factors not well captured by the VAR were no doubt also at work. More fundamentally, however, it is important to recognize that the empirical work in this paper looked only at the impact of a onetime shock to monetary policy. The impact of such a temporary shock is likely to be quite different from that of a systematic and sustained policy of disinflation. Finally, while the empirical results suggest that the short run output costs of reducing high inflation are relatively large, in the long run such costs are likely to be zero or even negative, as agents fully adapt to the new, low inflation environment. While the empirical results do provide some insight into some of the ways that systematic, tight monetary policy may have contributed to the secular decline in inflation, for example by helping to stabilize and appreciate the exchange rate, one cannot use the sacrifice ratio arising from a one-time shock to monetary policy to assess the impact of a protracted policy of disinflation on output.

## References

Azoulay, Eddie and David Elkayam, 1999,

*“A Model to Assess the Effect of Monetary Policy on Inflation in Israel, 1988 to 1996,” Bank of Israel, mimeo (in Hebrew)*.Bank for International Settlement, 1995,

(Basle: BIS).*Financial Structure and the Monetary Policy Transmission Mechanism*Bank of England, 1999,

*“The Transmission Mechanism of Monetary Policy,”*(May).*Quarterly Bulletin*Beenstock, Michael and others, 1994, “A Macroeconometric Model for Israel 1962-90,”

Vol. 11, pp. 413–462.*Economic Modeling*,Bernanke, Ben, 1986, “Alternative Explanations of the Money—Income Correlation,”

, Vol. 25, pp. 49–100.*Carnegie—Rochester Conference Series on Public Policy*Bufman, Guy and Leonardo Leiderman, 1997,

*“Monetary Policy and Inflation in Israel,” Paper presented at the Conference on Inflation and Disinflation in Tel Aviv, mimeo.*Christiano, Lawrence J. and others, 1998, “Monetary Policy Shocks: What Have We Learned and to What End?”,

.*National Bureau of Economic Research Working Paper No. 6400*Clarida, Richard and Mark Gertler, 1997,

*“How the Bundesbank Conducts Monetary Policy” in*, ed. by Christina D. Romer and David H. Romer (Chicago: University of Chicago Press).*Reducing Inflation: Motivation and Strategy*Cooley, Thomas F. and Stephen F. LeRoy, 1985, “Atheoretical Macroeconometrics: a Critique,”

Vol. 16, pp. 283–308.*Journal of Monetary Economics*,Cushman, David O. and Tao Zha, 1997, “Identifying Monetary Policy in a Small Open Economy Under Flexible Exchange Rates,”

, Vol. 39, pp. 433–448.*Journal of Monetary Economics*De Fiore, Fiorella, 1998, “The Transmission of Monetary Policy in Israel,”

*IMF Working Paper 98/114*(Washington: International Monetary Fund).Eichenbaum, Martin and Charles L. Evans, 1995, “Some Empirical Evidence on the Effects of Shocks to Monetary Policy on Exchange Rates,”

Vol. 110, pp. 975–1009.*Quarterly Journal of Economics*,Fischer, Stanley and David W. H. Orsmond, 1999, “Israeli Inflation Targeting from an International Perspective,”

*Paper presented at the conference “Changes in the Structure of the Israeli Economy During the Past Decade,” mimeo*.Leiderman, Leonardo and Assaf Razin, 1988, “Foreign Trade Shocks and the Dynamics of High Inflation: Israel, 1978–85,”

Vol. 7, pp. 411–423.*Journal of International Money and Finance*,Leiderman, Leonardo and Hadas Bar-Or, 1999,

*“Monetary Policy Rules and Transmission Mechanisms Under Inflation Targeting in Israel,” Bank of Israel, mimeo*.Reifschneider, David and others “Aggregate Disturbances, Monetary Policy, and the Macroeconomy: The FRB/US Perspective,”

, Vol.1, pp. 1–19.*Federal Reserve Bulletin No. 85*Romer, Christina D. and David Romer, 1989, “Does Monetary Policy Matter? A New Test in the Spirit of Friedman and Schwartz,”

*NBER Working Paper No. 2966*.Sims, Christopher S., 1986, “Are Forecasting Models Usable for Policy Analysis?”,

, pp. 2–16.*Federal Reserve Bank of Minneapolis Quarterly Review*(winter)Smets, Frank and Raf Wouters, 1999,

*“The Exchange Rate and the Monetary Transmission Mechanism in Germany,”*.*De Nederlandsche Bank Staff Report*1999 No. 35Swagel, Phillip, 1999, “Inflationary Implications of EMU-Related Monetary Easing: An Econometric Estimate,”

.*Greece—Selected Issues*, SM/99/255Swagel, Phillip, 2000, “Inflation and Growth in Israel,”

*IMF Working Paper (forthcoming)*.

^{}2

See Fischer and Orsmond (1999) for more information on the stabilization program.

^{}4

See Bank of England (May 1999) and Reifschneider and others (1999) for nontechnical descriptions of the transmission mechanism of monetary policy.

^{}5

They state that when nominal interest rates and inflation expectations enter the regression separately, the coefficients on the two variables are similar to that obtained on the real interest rate variables, suggesting that an increase in nominal rates would have a roughly identical impact on inflation.

^{}6

A preliminary summary of his results is contained as an appendix to Fischer and Orsmond (1999).

^{}7

Specifically, she looks at episodes of monetary tightening in October 1991, September 1994, and May 1996, all of which she identifies as occurring in response to excessive inflation.

^{}8

See Bernanke (1986) or Sims (1986) for more details on the VAR methodology. See Cooley and Leroy (1985) for a critique of the methodology.

^{}10

The nominal wage is cointegrated with Israeli prices, the exchange rate, and import prices in the markup equation.

^{}11

The issue of whether stationarity or cointegration of variables is necessary in a VAR framework is unsettled, but most empirical studies have tended to opt for differencing when data are not stationary and cointegrating relationships are not present.

^{}12

The use of quarterly data was motivated by practical considerations. First, data on real output and unemployment are available only on a quarterly basis. Secondly, when the VAR was estimated using monthly data (including interpolated monthly observations for the output and unemployment series) the resulting impulse response functions did not display sensible behavior among variables of interest, presumably because of noise in the monthly series.

^{}13

Given the relatively short sample period and the large number of variables included in the estimation, selecting the appropriate number of lags requires striking a balance between accurately reflecting the dynamics of the system and preserving degrees of freedom. The 42 observations covered by the sample period could in principle allow the inclusion of up to two lags on each of the endogenous variables. In the version in which the unemployment rate is excluded from the analysis, up to three lags could be included. However, impulse response functions for the variables of main interest were not substantially affected by the number of lags included in the estimation, suggesting that the relationships of interest are robust to the exact specification adopted, and the results in the remainder of this section are based on a VAR with one lag.

Jarque-Berra tests for residual normality, Ljung-Box tests for residual auto correlation, and Lagrange multiplier tests for residual auto correlation and a residual autoregressive conditional heteroscedasticity did not reject the null hypotheses of normality and no residual autocorrelation at 5 percent significance levels, with the exception of the normality tests for the inflation differential, where normality of the residuals was decisively rejected. This rejection may reflect the short sample period. Alternatively, it could reflect the fact that the interest rate differential is not fully endogenous.

F and *x*^{2} tests rejected at least the 5 percent significance level the null hypotheses that the coefficients on the constant term, on the seasonal dummies, and on the dummy for the structural break in 1992 were equal to zero in each of the regressions, and these deterministic regressors were therefore retained.

^{}14

Modifying this assumption so that output is contemporaneously uncorrelated with unemployment has no impact on the results that follow. Likewise, allowing output shocks to contemporaneously affect unemployment does not meaningfully alter the results.

^{}15

With real rates defined as nominal rates minus the annualized rate of inflation in a given quarter.

^{}16

This argument asserts that while in the past producers could be certain that any depreciation of the sheqel would be permanent and therefore could immediately be reflected in prices, there is now a greater likelihood that depreciations will be temporary, and producers are therefore more hesitant to pass them along into prices.