Israel: Selected Issues

Abstract

Israel: Selected Issues

Low Inflation in Israel—Should We Worry About It?1

A. Introduction

1. In a context of global disinflationary pressures, inflation in Israel has slipped into negative territory. This paper aims to better understand why this has happened. In particular, the paper aims to answer two questions.

  • Is the decline in inflation solely due to external shocks (the fall in energy prices and the appreciation of the shekel) or have domestic factors contributed as well?

  • What is the transmission channel of external shocks on inflation in Israel? Do lower energy prices only affect non-core inflation, or do they also have an impact on core inflation? Is the impact confined to first-round effects only, or are there also second round effects—with lower inflation expectations leading to lower inflation?

2. To answer these questions, we employ a semi-structural empirical model. We apply a slightly modified version of the model developed in Arnold, Chen, and Christiansen (2015) to Israel (see Section C and Appendix I).

3. The empirical results suggest that the fall in inflation is largely due to external factors. As discussed in further detail below, it appears that oil price changes and nominal effective exchange rate (NEER) fluctuations have been contributing negatively to inflation for the past two years. In contrast, domestic factors, such as the output gap, were mostly contributing positively.

4. We also find that, the impact on headline inflation from external shocks can be sizable and relatively long-lasting. The analysis suggests the pass through from a NEER shock is large. For example, a 10 percentage point depreciation of the NEER increases headline inflation by 0.5 percentage points after 2 quarters of the shock. An oil price shock tends to have a smaller impact; however, the effect tends to be more persistent, lasting for several years.

5. The paper is organized as follows: Section B presents some stylized facts related to low inflation in Israel and the monetary policy response so far. Section C discusses the empirical model and the results. Section D concludes.

B. Inflation and monetary policy responses: Stylized Facts

6. Inflation in Israel has declined from 3.5 percent in 2011 to minus 1.0 percent in March 2015. It has been below the lower end of the Bank of Israel’s (BOI) target band of between 1 and 3 percent since mid 2014. This decline is not unique to Israel, but has also been observed in both major economies (e.g., the United States and the euro area) and small open economies (e.g., Czech Republic, Poland, Sweden, and Switzerland).

A03ufig1

CPI Inflation

(Y/Y percent change)

Citation: IMF Staff Country Reports 2015, 262; 10.5089/9781513594675.002.A003

Sources: Haver analytics and Fund staff calculations.

7. The drop in inflation has occurred despite robust economic growth. In the last three years, GDP growth has averaged about 3 percent (Figure 1). Unemployment has fallen steadily, and the output gap—if any—has likely been modest.

Decomposing inflation

8. In our analysis we distinguish between core and non-core inflation, and between first and second round effects.

  • We decompose inflation into core (headline inflation excluding energy) and non-core inflation (energy) components. Domestic factors can only influence core inflation, while external factors can affect both core and non-core.

  • We will look at both first round and second round effects. First round effects are the impact of shocks on inflation, while second round effects are the impact of shocks on inflation expectations.

9. Using these decompositions we can break down the contribution of external factors into direct first-round effects on headline inflation (e.g., oil price shocks impacting fuel prices), indirect first-round effects operating through core inflation (e.g., oil price shocks impacting firms’ input costs), and second-round effects where external factors impact on inflation expectations which then feeds through into prices.

10. Figure 1 (upper right panel) shows that both core and non-core inflation have declined. The contribution from core inflation has been declining since early 2014, while falling energy prices have been contributing negatively to headline inflation since mid-2014. This corresponds to the decline in global commodity prices, especially oil prices, over the same period (Figure 1, middle left panel).

Figure 1.
Figure 1.

Drivers of low inflation: stylized facts

Citation: IMF Staff Country Reports 2015, 262; 10.5089/9781513594675.002.A003

Sources: IMF WEO, IMF GAS database, Haver analytics, and Fund staff calculations.1/ A negative change indicates appreciation of the shekel.

11. The fall in core inflation may have been due to exchange rate appreciation. The shekel appreciated by nearly 20 percent in nominal effective terms between the late 2012 and mid-2014 (Figure 1, middle right panel). This appreciation translated into falling import prices, which could have an impact on both core and non-core inflation.

12. The Bank of Israel has responded with a series of easing measures. Since mid-2012, the BOI lowered its policy rate by a total of 2.4 percentage points to 0.1 percent. Interest rate cuts in August and September 2014 brought an end to the shekel appreciation, and triggered a sharp, if temporary, depreciation of the shekel. When upward pressure on the shekel resumed, the BOI cut interest rates by 0.15 percentage points in March 2015. In addition, the BOI intervened in the foreign exchange market by a cumulated amount of USD 4.9 billion since 2014.2

13. These measures seem to have succeeded in stabilizing inflation expectations. Even though headline inflation was still declining, shorter-term inflation expectations bottomed out in late 2014 and have risen slightly since then, with 2-year ahead expectations at 1 percent in March 2015.

14. With current policies, inflation is expected to converge back to the target in 2016. The year-on-year decline in oil prices is expected to slow during the latter part of 2015, providing sufficient support for inflation converging to the target. Moreover, the current policy mix is very accommodative reflecting the near-zero interest rates and broadly neutral fiscal policy, despite a near-zero output gap and declining unemployment.

A03ufig2

Simulated CPI Inflation

(Y/Y percent change)

Citation: IMF Staff Country Reports 2015, 262; 10.5089/9781513594675.002.A003

Source: Fund staff calculations.Note: The simulated inflation path is generated based on an estimated open economy Philips curve, with the assumptions on output gap and oil prices from the WEO and GAS and an unchanged shekel exchange rate against the USD.

C. Empirical Approach and Results

15. Identifying the drivers of inflation and the channels through which they operate is important for determining the appropriate policy response. When expectations are well anchored, temporary external shocks should not have permanent effects on inflation, and policymakers can “look through” these shocks as inflation will converge back to target as the shock dissipates. However, if there is a series of external shocks in the same direction or the shocks are very persistent, then policymakers may need to respond to ensure inflation expectations remain well anchored. Thus, it is important to understand the factors driving low inflation and whether those factors have second-round effects that raise the risk of low inflation becoming entrenched in inflation expectations.

16. As noted above, we employ a semi-structural empirical model to identify various domestic and external factors driving inflation. We use a modified version of the model from Arnold, Chen, and Christiansen (2015). The model uses the identity of headline inflation as the sum of core and non-core inflation to inform the estimation of a hybrid open economy Phillips curve equation for core inflation and non-core inflation equation based on external factors. This allows us to differentiate between domestic and external factors’ contributions to inflation, including the persistence of any shocks, as well as their direct and indirect first-round effects on headline inflation. The model also includes an equation for inflation expectations to assess the second-round effects of domestic and external factors. A more detailed discussion of the model and its empirical specification can be found in Appendix I.

Empirical Results

17. The empirical results suggest that external factors have indeed been significant drivers of inflation. Historical decompositions confirm that the external factors accounted for the largest contribution to CPI inflation (Figure 2).

  • External drivers. Oil price changes and NEER fluctuations have generally had a larger impact than the output gap, particularly in combination. NEER appreciation in 2013 and first half of 2014 had sizable and persistent negative effects on inflation. For instance, the appreciation of the shekel pulled down CPI inflation by more than 0.7 percentage points in 2014Q2. However this impact is fading away following the shekel’s depreciation in the second half of 2014. Even as the drag on inflation from past NEER appreciation wanes, the decline in oil prices between mid-2014 and January 2015 has become a significant source of disinflationary pressure. The analysis suggests the contributions from oil prices to CPI inflation have fallen by 1.6 percentage points to -0.6 percent in 2015Q1 compared to 2012Q1

  • Domestic factors. The impact of domestic factors has been modest. The key domestic factor we examine is the output gap, with positive output gaps contributing positively to inflation both before the global financial crisis and in its aftermath. More recently, the contribution from output gap has declined, turning slightly negative in the last two quarters, and is rather small. In particular, the contributions to CPI inflation from the output gap have declined by 0.6 percentage points to 0.0 percent in 2015Q1 compared to 2012Q1.

A03ufig3

Contributions to CPI Inflation

(Percentage points)

Citation: IMF Staff Country Reports 2015, 262; 10.5089/9781513594675.002.A003

Sources: Bank of Israel, Haver analytics and Fund staff calculations.1/ CPI inflation is adjusted to account for the constants in the regression framework and any insignificant external factors (see Appendix II). “Other domestic” are the residuals and base effects from the core and expectations equations, while “Other external” are the base effects and residuals from the non-core inflation equation.

18. Second round effects have been modest. The model allows us to decompose the sum of the external shocks impact on headline inflation into direct first-round effects (through non-core inflation), indirect first-round effects (through core inflation), and second-round effects (affecting core inflation through inflation expectations) (Figure 3). The direct effects (blue bars) from external factors largely correspond to oil price changes, though NEER fluctuations also contribute to the direct effects at times. Usually though, external factors have their largest impact through indirect effects (orange bars). There is also evidence of second-round effects (red bars), which have recently contributed negatively to headline inflation, but the magnitude of these effects is small.

A03ufig4

First-Round and Second-Round Effects of External Factors

(Percentage points)

Citation: IMF Staff Country Reports 2015, 262; 10.5089/9781513594675.002.A003

Sources: Bank of Israel, Haver analytics and Fund staff calculations.

19. The impact of exchange rate shocks rapidly tapers off after about five quarters, while the impact of oil price shocks is more persistent. The model can also be used to generate impulse responses of inflation to external shocks, as illustrated in Figure 4. Indeed, the impulse responses of inflation to external shocks illustrate that the impact from an oil price shock tend to be much more persistent compared to a shock to NEER, while the peak impact from a NEER shock can be twice as large.

  • Oil price shock. Figure 4 illustrates the impact of a one-period 10 percentage points decline in the oil prices. As expected, the peak impact (about -0.18 percentage points) occurs at the time of the shock, and inflation will be 0.49 percentage points lower 1 year following the shock. Moreover, the results indicate that the oil price shock has a very long lasting effect.

  • NEER shock. The figure also shows the impulse response to a one-time 10 percentage points appreciation in the NEER. Unlike with the oil price shock, the peak impact occurs one quarter after the shock. Also, the effect is less persistent despite the peak impact being more than twice as large (-0.50 percentage points). Moreover, inflation will be 0.9 percentage points lower in 1 year time.

A03ufig5

Inflation Responses to NEER and Oil Shocks

(Percentage points)

Citation: IMF Staff Country Reports 2015, 262; 10.5089/9781513594675.002.A003

Source: Fund staff calculation.Note: Impulse response of CPI inflation to an oil price or NEER appreciation shock of 10 percentage points. The x-axis represents quarters after the shock.

D. Conclusion

20. External factors have been largely responsible for low inflation in Israel. The analysis shows that domestic factors, particularly the output gap, have played only a limited role in the recent episode of low inflation. External factors, including the recent sharp decline in oil prices and NEER fluctuations, have been the key drivers of disinflation.

Appendix I. Main Empirical Methodology and Results1

Empirical approach

1. We closely follow the empirical methodology in Arnold, Chen and Christiansen (2015). First, we separate headline inflation into core inflation and non-core inflation which allows us to distinguish between direct and indirect first-round effects of shocks. For example, an oil price shock will impact non-core inflation directly but might also affect core inflation through input prices. In addition, we assess effects of shocks to inflation that work through price setting and expectations, so-called second-round effects. This allows us to document how different domestic and external factors impact headline inflation through the different channels. Domestic factors are captured by the impact of the output gap on inflation. External factors are captured by oil price changes and nominal effective exchange rate (NEER) fluctuations.

2. We employ a semi-structural empirical modeling approach. We start with the identity of headline inflation as the weighted sum of core and non-core inflation. Core inflation is modeled as a hybrid open-economy Phillips curve equation, which includes lagged core inflation, forward-looking expectations, and domestic and external variables. Non-core inflation is linked to external factors only. Inflation expectations are modeled based on all domestic and external factors impacting headline inflation. This allows us to capture the second-round effects from shocks.

3. The system of equations is estimated using the Seemingly Unrelated Regressions (SUR) approach to account for possible correlation between the error terms across equations. Empirically, we use tradable and non-tradable inflation as proxies for core and non-core inflation, this would allow us to have longer time series for the empirical estimation. In particular, the system is estimated using quarterly data using data between 2005Q1 and 2014Q4.

System of SUR equations

4. Five equations represent the system of equations estimated using the SUR approach. Let πtCPI denote headline inflation, πtcore denote core inflation, πtnoncore denote non-core inflation, and Wtcore(Wtnoncore) denote the weight in the consumption basket of the core (non-core) components. Also, πtE2yr let denote (2-year ahead) inflation expectations formed at period t, and RULCt denote the real labor costs. External factors include NEER fluctuations (NEERt) and oil price changes (oiltall). The component of oil price changes due to global aggregate demand is denoted oiltD (see Arnold, Chen and Christiansen, 2015).

5. Headline inflation identity. Including this indentify in the estimation of the system means minimizing the weighted sum of squared error terms for equations (2) and (3), with the weights corresponding the weights of core and noncore inflation in headline inflation.

(1)πtCPI=Wtcoreπtcore+Wtnoncoreπtnoncore
  • Core inflation. We model core inflation using a hybrid open-economy Phillips curve (Gali and others, 2001; Svensson, 1998), with the constraint that the coefficients on the backward and forward-looking inflation terms sum up to 1 (Mavroeidi, 2005). To control for potential endogeneity or multicollinearity issues when estimating the system, we use one period lagged NEER changes and oil price shock. In addition, we use a fitted value for the real unit labor costs as described in equation (5).
    (2)πtcore=α21+α22πt1core+(1α22)πt1E2yr+α23RULCt+α24NEERt+α25oiltD++ɛtcore
  • Non-core inflation. Non-core inflation is a function of changes in oil prices and the NEER, using predicted NEER values from equation (6) when estimating the model.
    (3)πtnoncore=α31+α32πt1noncore+α33NEERt1+α34oiltall++ɛtHICP
  • Inflation expectations. We assume inflation expectations are formed based on all available information at time t, including the domestic and external factors affecting core and non-core inflation, as well as the previous period’s inflation expectations.
    (4)πtE2yr=α41+α42πt1E2yr+α43RULCt+α44NEERt+α45oiltall++ɛtE2yr
  • Real unit labor costs. We assume real unit labor costs are a function the past output gaps and its own lagged values.
    (5)RULCt=α51+α52RULCt1+α53Ygapt1+α54Ygapt2+ɛtRULC

6. Empirically, we control for potential endogeneity or muticollinarity issues by using either fitted or lagged values of the independent variables. Within a given quarter, the NEER changes could be affected by contemporaneous feedback from inflation (or inflation expectations). Thus, we use one period lagged NEER changes. Given trade consists of an important share of Israeli growth, change in oil prices may in part reflects global business cycle which is correlated with Israeli real unit labor costs. As a result, we use lagged oil price change. Finally, any potential measurement errors associated with the real unit labor costs may be correlated with the contemporaneous error terms in the inflation equations. Thus, in both equation 1 and 3, we use fitted value of RULC which is modeled through equation 5. Moreover, equation 5 allows us to distinguish part of the RULC is driven by changes output gap which allows us to identify the part that is driven by output gap.

Historical decomposition

7. The historical decomposition accounts for the contribution of domestic and external factors including their lagged effects. Following Burbidge and Harrison (1985), we estimate the individual contributions of each external and domestic factors to the headline inflation over the sample period. Equations 24 essentially describe core, noncore inflation and inflation expectations as AR(1) processes with some external (exogenous) shocks (i.e. NEER, oil prices and euro area inflating). Using backward substitution, core, noncore inflation and inflation expectations at each point in time can be represented as a function of initial values plus all the external shocks as well as the residuals in the model. Also, when constructing the decompositions, for the external factors we set the coefficient to zero if it is insignificant and a test of joint significance of the coefficients on that factor across equations (using the Wald test) is not significant at the 5 percent level.

πtcore=α21Σj=0t1α22tπ0core+(1α22)Σj=0t1α22jπt1jE2yr+α23Σj=0t1α22jRULCtj+α24Σj=0t1α22jNEERtj+α25Σj=0t1α22joiltjD+Σj=0t1α22jɛtjcoreπtnoncore=α31Σj=0t1α32j+α32tπtnoncore+α33Σj=0t1α32jNEERtj+α34Σj=0t1α32joiltjall+Σj=0t1α32jɛtjnoncoreπtE2yr=α41Σj=0t1α42j+α42tπ0E2yr+α43Σj=0t1α42jRULCtj+α44Σj=0t1α42jNEERtj+α45Σj=0t1α42joiltjall+Σj=0t1α42jɛtjE2yr

Results

8. As table 1 shows, most of the estimated parameters are significant with expected signs. Higher real unit labor costs pushes domestic inflation upwards, at the same time, it also has impact to increase expected inflation. Oil prices seem to have significant impact on core, non-core as well as inflation expectations. Interestingly, NEER only works through non-core inflation and inflation expectations. Our estimates suggest 10 percent deprecation implies inflation to increase by about 1.5 percentage points in the long run.

Table 1:

Estimated Coefficients

article image
1

Prepared by Jiaqian Chen.

2

The amount excludes the foreign exchange purchases aiming to offset the effect of natural gas production on the exchange rate.

1

The model mostly follows Arnold, Chen and Christiansen (2015) with some minor modifications. Thus, this appendix is copied largely verbatim from the Arnold, Chen, and Christiansen (2015).

Israel: Selected Issues
Author: International Monetary Fund. European Dept.