Inflation targeting in practice: further evidence.
Since the early 1990s, a monetary policy strategy known as "inflation targeting" has been formally adopted by a number of industrialized countries. Under this approach, a monetary authority officially announces an inflation target level or target range, which generally is consistent with low or zero inflation, to be achieved within a specified time frame.(1) Both conceptual and practical issues of inflation targeting have received considerable attention in the literature. This paper focuses on the latter issue by drawing lessons from the experiences of three representative countries - New Zealand, Canada, and the U.K. - which have recently been studied by Ammer and Freeman (1995), McCallum (1996), Dueker and Fischer (1996), and Mishkin and Posen (1997).
At a superficial level, recent inflation rates of 2% or less per annum across those three inflation-targeting countries lend credence to the new policy rule in operation. Moreover, in light of vector autoregression (VAR) simulations, most of the existing studies reveal substantial effect of inflation targeting on economic performance. Dueker and Fischer (1996), however, argue that conclusions based solely on unilateral evidence can be misguided, given that many other countries have also shared parallel disinflationary experiences even without employing any formal targeting framework. Their argument stems from the difficulty of disentangling effects due to inflation targeting from those due to other nonpolicy factors, such as a global absence of inflationary pressures attributable to recent falls in oil prices. This distinction is particularly crucial for the three small open economies.
This paper attempts to provide further insights into the above unresolved issue. In contrast to existing cross-country case studies, this study involves a two-step approach that can offer more direct inferences on the performance of inflation-targeting countries relative to some non-inflation-targeting counterparts. The empirical work begins with employing the methodology of Vahid and Engle (1993), which helps remove stochastic trend and cyclical components in observed data that the three targeting countries have in common with some nontargeting neighbors. Subsequently, inflation-targeting regimes are evaluated by applying VAR simulations to the filtered country-specific data.
In line with the country-pairing strategy employed earlier by Dueker and Fischer (1996), this study adopts a bilateral framework by pairing Canada with the U.S.; the U.K. with Germany; and New Zealand with Australia. The selection of the first two pairs is consistent with Engle and Kozicki (1993), who find evidence of common trends in their output series among those of the G-7 countries. Another reason for choosing the U.S. and Germany is their different official policy rules, which focus on targeting money growth or interest rates instead of inflation. Australia has followed the example of New Zealand by instituting an official target since mid- 1993. The Reserve Bank of Australia, however, has not announced any specific targeting time frame like those well defined by its New Zealand counterpart. Due to its late (official) entry into the targeting regime relative to New Zealand, inferences can be drawn at least through 1993.
Juxtaposed against the observed data series, the country-specific time series clearly reveal that a great deal of unilateral evidence on regime-shift effects is in fact an artifact of cross-country comovements that are consequently removed. VAR simulations conditional on the country-specific data further confirm that the new monetary regimes have not been instrumental in reducing inflation beyond the levels already experienced in some nontargeting countries.
This paper proceeds as follows. The next section offers an overview of the historical background, and presents VAR simulation results based on observed data. Section III discusses the bilateral data decomposition results. Section IV reports simulation results after controlling for trend and cyclical comovements in the three bilateral systems. The final section provides a summary and concluding remarks.
II. BACKGROUND, ORIGINAL DATA, AND BENCHMARK RESULTS
Figure 1 illustrates the three inflation-targeting countries' target ranges (dashed lines) and the historical inflation patterns measured by fourth-quarter changes in the price index that the central banks have adopted for policy guide. Bernanke et al. (1999) offer detailed discussions of the underlying policy frameworks. Briefly, in each case, the monetary authority has opted to target the underlying trend of an inflation measure, which is the consumer price index (CPI) for New Zealand and Canada, and retail prices for the U.K. While the central banks of the first two countries have established target ranges that dropped in a stepwise manner over time, the Bank of England has chosen less ambitious targets, holding the increase in retail prices within a relatively wide 1-4% range. Figure 1 shows that inflation rates indeed fell precipitously in the early 1990s, and the three central banks have attained their stated objectives. The somewhat less impressive performance for New Zealand in recent years corresponds to much higher inflation levels (reaching 5% in 1995) experienced by its neighboring country, Australia. Increased inflationary pressures experienced by New Zealand have led its new coalition government to widen the inflation target bandwidth from 0-2% to 0-3% since December 1996.
To facilitate comparison, dynamic simulations are first performed on the basis of an unrestricted VAR fitted to historical data. In line with Ammer and Freeman (1995) and Mishkin and Posen (1997), the model consists of four variables: output, inflation, a long-term interest rate, and a short-term interest rate closely associated with monetary policy. The output variable is measured by log GDP level; inflation is measured by log CPI level; the long-term interest rate is measured by the 10-year government bond yield; the short-term rate is measured by the overnight or discount rate.
B. Data Representations
This study focuses on quarterly data of the six countries under consideration over the period 1975-1996.(2) The beginning period is selected in order to avoid possible structural breaks associated with the first major oil shock and changes in foreign exchange regimes in the early 1970s. As an initial step in the econometric work, the appropriate specifications for the variables are examined. This is done by performing the augmented Dickey-Fuller (ADF) and Phillips-Perron (PP) t-tests for unit roots. For both tests, the orders of the autoregressive lag included in regressions are determined using a sequential procedure suggested by Perron (1989), beginning with a maximum lag order k = 8.
Table 1 reports ADF and PP test results for the 24 data series, four for each of the six countries. As shown in the "Levels" category, the unit-root null hypothesis is invariably rejected for all variables in levels. "Differences" shows test results for data in first differences. In this [TABULAR DATA FOR TABLE 1 OMITTED] [TABULAR DATA FOR TABLE 2 OMITTED] case, the null can be rejected at the 10% significance level or higher for all variables except the CPI. For the latter, there is only weak evidence of stationarity in the PP tests.
Perron (1989), however, argues that the presence of structural change in data can significantly lower the power of standard unit-root tests. Given the focus on major policy shifts, this possibility deserves attention. Importantly, Figure 1 reveals noticeable reductions in both the levels and variability of inflation under the new regimes, which might have plausibly affected our statistical findings. From this perspective, the sequential procedures developed by Zivot and Andrews (1992) and Lumsdaine and Papell (1997) are followed to allow the ADF regression for a single and multiple a priori unknown breakpoints, respectively.
Table 2 displays results of the modified ADF t-tests, which are the minimum ADF t-statistics among all possible breakpoints in the samples. All variables are in levels, except for the CPI, which is in first differences. The table shows results for the case of one possible trend break and for the case of two breaks. With few exceptions, almost all of the periods corresponding to the single-break statistics are also identified under the multiple-break method. Moreover, most ADF t-statistics are smaller than their counterparts in Table 1. Given the finite-sample critical values, nevertheless, the null hypothesis of a
unit root can not be rejected in any case in favor of stationarity along a broken trend.
Table 2 reveals several interesting observations regarding the periods that yield the minimum ADF t-statistics. First, in the case of either one- or two-break method, the periods for the output series of New Zealand, Canada, and the U.K. are closely associated with the respective "official" dates of regime shifts in the early 1990s. On the other hand, while the potential break for New Zealand's price series occurs around its first inflation-target announcement, those for the other two inflation-targeting countries occur around the late 1970s and the early 1980s when dramatic changes in inflation occurred. Interestingly, the locations of break candidates for the three inflation-targeting countries are also closely related to those for their non-targeting counterparts.
Overall, despite the allowance for possible structural breaks, statistical results in Table 2 reinforce that, for all countries, the output and interest rate data series are nonstationary in levels but stationary in first differences. The CPI data, on the other hand, are stationary only in second differences. For subsequent empirical work, we therefore express the two interest rates and the output variables in first differences, and took CPI in second differences, which represent changes in inflation. In addition, the output and CPI data are multiplied by 400 to represent annualized percentage rates of change.
C. Benchmark Simulation Results
Based on the data specifications for the four variables described above, dynamic simulation exercises proceed as follows. For each country, a fourth-order unrestricted VAR is first fitted to the data sample running from the first quarter of 1975 through the period just before the time of first inflation-target announcement, i.e., 1990:1 for New Zealand, 1991:1 for Canada, and 1992:3 for the U.K.(3) Resulting parameter estimates are then used to generate unconstrained forecasts of the four variables from the period of policy shift forward through 1996:4. Effectively, the simulated paths represent counterfactual series that correspond to what the variables would have followed in the absence of inflation targets.
Figure 2A presents simulation results for New Zealand. Vertical gridlines mark the beginning period of the new policy regime. Long dashed lines represent simulated values, encapsulated by two standard error or 95% confidence bands (short dashed lines) generated by Monte Carlo simulations with 1,000 draws, based on a method described in Doan (1992). As shown in the top frame, the large gap between the actual and simulated series within the first Year of the simulation horizon depicts a greater than 5% loss of output growth associated with Reserve Bank of New Zealand's new disinflation policy.
The second frame displays results for New Zealand's inflation variable, with negative figures representing inflation reductions. Immediately after the initial inflation-target announcement, the actual data series follows a dramatic drop and subsequently hovers below its simulated counterpart through the end of 1991. For the remainder of the targeting period, the two time paths are qualitatively indifferent from each other.
The third frame depicts results for the long-term bond series. Similar to the inflation series, negative values in 1991 reflect drops in the interest rate level during the earlier periods of inflation targeting. Since then the bond series tends to be either below or within the 95% confidence interval of its simulated path. A similar pattern can be realized from the discount rate, as illustrated in the bottom frame. This last observation suggests that New Zealand's low inflation levels throughout much of the 1990s were not associated with deliberately tight monetary policy actions, which would in principle require increases in the discount rate at least during the onset of the reactive policy regime.
Figures 3A illustrates corresponding scenarios for Canada. Compared with those for New Zealand, inflation changes and output growth exhibit smaller variations at the beginning of the targeting regime. They nevertheless remain statistically lower than their simulated counterparts. During the early 1990s, both Canadian interest rate series become negative, indicating decreasing levels. Except for a few blips, the actual interest rate series tend to stay below their simulated paths.
Figure 4A displays simulation results for the U.K. As a comparison, evidence of reductions in inflation and output growth is much weaker than that for the other two countries. Policy effects, however, are more discernible in the bottom two frames, which show dramatic declines in the actual interest rates within the first year of inflation targeting. Furthermore, as in the case of New Zealand, data variability tends to reduce over time across the new regime. This last observation implies that the U.K. and New Zealand programs may have succeeded in achieving stability in prices as well as in other variables.
The policy effects that are evidenced in panel A of Figures 2-4 appear to be weaker than comparable results reported by Ammer and Freeman (1995) and Mishkin and Posen (1997). It is important, however, to understand that for ease of illustration, those earlier studies employ levels specifications. In order to maintain comparability with the empirical work that follows, the simulation results reported above, on the other hand, are based on data in differences. By comparison, a one-period change in a variable expressed in differences would, ceteris paribus, lead to a permanently different level for that variable in all succeeding periods. For instance, the rapidly declining and subsequently low inflation levels observed in Figure 1 correspond to the large negative values for the change in inflation series immediately after the inception of targeting regimes, as shown in Figures 2-4. Lessons drawn from observed data, however, may be illusory in the sense that many non-inflation-targeting countries have also shared similar experiences. In particular, Dueker and Fischer (1996) assert that policy effects become ambiguous once the economic records of these targeting countries are compared with those of their neighbors. Moreover, based on simulations with interest rate data, Kahn and Parrish (1998) argue that there has been no systematic change in the way monetary policymakers in targeting countries have responded to changes in economic conditions.
Given these perspectives, the next section seeks to remove the time-series dynamics in observed data that the three targeting countries have in common with their nontargeting neighbors. Australia has been chosen for New Zealand, the U.S. for Canada, and Germany for the U.K. The first of each country pair is relatively larger, implying a greater tendency for it to exert economic influences, such as inflationary pressures, on its targeting counterpart.
III. DATA DECOMPOSITION RESULTS
To obtain relevant country-specific data, this section applies Vahid and Engle's (1993) common-trend/common-cycle approach to 12 bilateral data systems (four variables and three country pairs). Under this framework, common stochastic trends are characterized by cointegration among integrated variables, while common "cycles" are characterized by serial correlation of common features among the residual stationary components of these variables. Intuitively, these are conditions that make time series move together in the long- and short-run horizons, respectively. Given this representation, Engle and Issler's (1995) two-step procedure is used to search for common dynamics within a bilateral setting. For each of the 12 data systems, the first step is to identify common trends by determining the number of cointegrating rank; the second step is to search for the number of common cycles using canonical correlation analysis.
The number of cointegrating rank is identified by performing two alternative statistics developed by Johansen (1988), and Johansen and Juselius (1990): the maximum likelihood ([[Lambda].sub.max]) and the likelihood ratio trace ([[Lambda].sub.trace]) statistics. The former tests the null hypothesis that the number of independent cointegration vectors is r against the alternative of r + 1 cointegrating vectors; the latter tests the null that the number of independent cointegrating vectors is less than or equal to r against a general alternative.
In light of the unit-root test results presented in section II, cointegration tests are performed on the log levels of GDP, the inflation series expressed as first differences of the log CPI, and the bond and discount rates in levels. Test results for the 12 bivariate systems covering the period 1975:1-1996:4 are reported in Table 3. In all cases, the null hypothesis of r = 0 can be rejected at conventional statistical levels, supporting the existence of a single common trend for each bilateral system. The [TABULAR DATA FOR TABLE 3 OMITTED] test statistics for testing r = 0 are weaker for the Canada-U.S. system, but nevertheless are statistically significant at the 10% level.
Nadal-De Simone (1996), however, argues that policy regime changes may potentially affect long-run relations among variables. This possibility is explored by performing two tests suggested by Hansen (1992). The statistics, Sup-F and Mean-F, are essentially extensions of Andrews' (1993) Quandt likelihood ratio methods for testing parameter instability in stationary systems given a single a priori unknown breakpoint.
As reported in Table 4 (top), both Sup-F and Mean-F statistics indicate strong evidence of structural instability in cointegrated systems. Parameter constancy is rejected for the majority of cointegrating relations except for (i) the inflation systems for the New Zealand-Australia and the Canada-U.S. pairs, and (ii) the short-term interest rate systems for the New Zealand-Australia and the U.K.-Germany pairs. Interestingly, most of the identified breaks occur in the 1980s rather than the 1990s, when the targeting strategies officially began. Given these test results, the cointegration models are augmented with a level-shift dummy variable corresponding to the identified structural breaks. For the output data, a dummy variable is also used to account for a possible change in the slope of common trend.
The second step in the data decomposition process involves identifying common "cyclical" or temporary features. This is done by applying canonical correlation analysis to the detrended data, which capture variations around identified common (broken) stochastic trends. Before performing such analysis, it is instructive to investigate whether there are structural breaks in the detrended data. There is, however, no direct method available for testing structural breaks within canonical regression. As a result, Andrews's (1993) Sup-F and Mean-F statistics are computed using least squares to test for structural instability in the detrended time series.
As shown in Table 4 (bottom), except for the bond rates for the New Zealand-Australia and the U.K.-Germany systems, the Andrews tests indicate little evidence of structural instability. Therefore, except for those two systems detected for structural breaks, canonical regressions are run over the entire sample period. For the remaining systems, the full sample is split at the identified break dates and squared canonical correlations are computed separately.
Table 5 displays canonical regression results. In most cases, including those estimated with split samples, test statistics show support for one canonical correlation and thus one common cycle. Taken together, the cointegration and canonical correlation test results for the 12 bilateral systems confirm that for each of the four variables, one common stochastic trend and one common cycle are shared by each of the three country pairs.
The relevant data of interest are obtained by subtracting the identified common cycle component from the detrended data series. In essence, the residual series capture idiosyncratic stationary dynamics relative to those shared by the other country within a bilateral framework, with the allowance for breaks in common trends. Moreover, for the New Zealand and U.K. bond rates, common cycle data are obtained by splicing the time-series segments extracted from the split-sample canonical regressions described above.
The patterns of transformed data for the three inflation-targeting countries over the 1985-96 period are plotted in panel B of Figures 2-4. In contrast to the original differenced data (in left panels), these new series represent deviations away from the common trend and cyclical dynamics that the targeting countries share with their non-targeting neighbors. If the historical data of a variable follow closely to those of its neighboring counterpart, then the transformed series would lie close to zero. Except for the long-term bond series, this is apparently the case, reflecting high synchronization in economic activity within the three country pairs.
By comparison, the amplitudes of deviations away from zero are much larger for the bond series than those for the discount rates. This finding reflects that short-term interest rates, which are more directly managed by central banks, are more synchronized across countries than long-term interest rates. The long-term bond rates, on the other hand, are to a larger extent affected by market expectations as well as other factors, such as policy credibility. For instance, while the idiosyncratic bond series of New Zealand tends to be persistently lower during the inflation-targeting period than earlier, the Canadian and British [TABULAR DATA FOR TABLE 4 OMITTED] [TABULAR DATA FOR TABLE 5 OMITTED] series continue to fluctuate around a zero mean. Such comparative evidence is indicative of lower inflationary expectations, and thus higher credibility, for New Zealand's monetary regime (relative to its Australian counterpart).
As a whole, the transformed country-specific data reveal two salient features. First, in contrast to the patterns exhibited in the original differenced data, there is no apparent reduction in data variability associated with the new policy regimes. Rather, the new inflation series become slightly more variable in the 1990s, reflecting larger changes in inflation for the three inflation-targeting countries relative to their neighbors. Second, discernible changes in the unilateral data at the times of policy shift to inflation targeting are largely absent from the idiosyncratic time series. Motivated by such prima facie evidence, the following section reevaluates the three inflation-targeting programs in light of the new data set.
IV. SIMULATION RESULTS WITH COUNTRY-SPECIFIC DATA
Once again the four-variable VAR simulation exercises as described in Section II are repeated for the three inflation-targeting countries, only this time country-specific data are used. Panel B of Figures 2-4 depicts the simulated paths, which are superimposed on the transformed data series.
Figure 2B illustrates simulation results for New Zealand. Two observations stand out. First, the inflation series (relative to that of Australia) lies within the 95% confidence band of the simulated series throughout much of the targeting period. The output series, on the other hand, tends to outpace its simulated path from 1993 onward, when the Reserve Bank of Australia began to adopt a similar targeting program.
Evidence from the two interest rates is somewhat mixed, however. While the discount rate series lies mostly within the 95% confidence band of its simulated path, the bond series appears to be persistently lower through 1995. As discussed above, this finding, which appears to be even more pronounced than that realized in Figure 2, may be interpreted as evidence of relatively high policy credibility for the Reserve Bank of New Zealand.
Figure 3B displays corresponding results for Canada. Except for the bond rate, all idiosyncratic data series are well encapsulated in the 95% confidence bands of their simulated counterparts. Similar to that for New Zealand, the bond series lies significantly below its simulated path in the earlier years of inflation targeting. Beyond that, most of the policy consequences observed in panel A are nonexistent.
Similar results for the U.K. are realized from Figure 4B. There is, however, an interesting observation not found in the other two countries. That is, the U.K.'s new output, inflation and bond series are significantly above their corresponding simulated values during the first year of targeting-policy adoption. Thereafter, the output, inflation, and discount rate series are largely indifferent from the simulated paths. The idiosyncratically high British bond rates before 1993 are also observed by Huh (1997) in light of comparisons with German data.
In contrast with the benchmark results in panel A of Figures 2-4, the country-specific data deliver overall much less meaningful evidence associated with anti-inflation policy actions. A lack of qualitative difference between the idiosyncratic inflation series and their simulated paths implies that the inflation-targeting rule has been less instrumental as a policy guide than once realized in section II, or in earlier studies, for example Ammer and Freeman (1995) and Mishkin and Posen (1997). Yet, under the perspective that long-term bond rates reveal inflation expectations as well as policy credibility, some comparative evidence not apparent in unilateral data emerges. More specifically, the bond series for New Zealand (relative to that for Australia) during much of the targeting regime has been idiosyncratically lower than the levels extrapolated from the historical past. In the case of Canada, on the other hand, it has been roughly the same. Finally, for the U.K., it has deviated widely, reflecting the extent of volatility in market expectations.
V. SUMMARY AND CONCLUSION
This paper has empirically reevaluated the experiences of three countries whose central banks have implemented explicit targets that aim at achieving price stability. Perceptible changes immediately after initial adoption of inflation targets are evidenced in the historical data of inflation and three other macroeconomic aggregates. In line with earlier studies, VAR simulations based on observed data confirm that inflation reductions would have been markedly smaller in the absence of policy reforms. Moreover, rapid disinflation and reduced output growth immediately following the shifts to inflation targeting resemble effects construed for a deliberate anti-inflation regime. There is also evidence of reduced variability in historical data.
These results aside, the paper proceeded to evaluate the assertion that the observed regime-shift effects have indeed been associated with a global disinflationary environment. To this end, Vahid and Engle's (1993) methodology was used to construct, for each of the three targeting countries, a new set of time series that depict idiosyncratic stationary dynamics relative to those of their nontargeting neighbors. Simulations conditional on these country-specific data revealed scant evidence of policy effects once evidenced in unilateral data. Evidence further supports that the sustained reduction of inflation rates in those targeting countries could have been secured in the absence of the new regime.
Given the relatively smaller sizes of the three inflation-targeting economies, the sharp contrast between the findings based on observed data and those based on filtered data can be attributed to cross-country synchronization of economic activity. Recognizing that recent disinflationary trends among inflation-targeting countries have merely been a happy coincidence should, nonetheless, by no means be interpreted as an argument against the desirability or feasibility of inflation targeting per se. Rather, the key insight from this cautionary tale is that before drawing any empirical inferences on the merit of any policy, one should examine historical evidence more circumspectly. In the present case, it is important to understand the extent to which economic fluctuations are international or domestic phenomena.
Despite the discovery of some features of interest, inferences in the present paper have been confined to a simple bilateral setting in an attempt to facilitate comparisons with earlier work. As a result, its findings should best be viewed as illustrative rather than conclusive. A fuller assessment of the potential effect of increased global integration on the efficacy of a policy rule requires a richer framework under a multilateral setting.
ADF: Augmented Dickey-Fuller PP: Phillips Perron VAR: Vector autoregression
This is a revised version of a paper presented at the 1998 Conference of Pacific Rim Allied Economics Organizations, Bangkok, Thailand, January 17, 1998. The author appreciates helpful comments from session participants, particularly Francisco de Asis Nadel-De Simone, and two anonymous journal referees.
1. There is controversy over the precise meaning of this targeting approach. Bernanke and Mishkin (1997) argue that the approach is better represented as a policy framework with some flexibility in policy actions rather than a policy rule. Moreover, while countries like the U.S. and Germany have had inflation-focused monetary policies, they have not made any explicit acknowledgement and are not commonly regarded as countries practicing this strategy.
2. The price and interest rate data were obtained from International Financial Statistics, and GDP data from Organization for Economic Cooperation and Development Quarterly National Accounts and Main Economic Indicators.
3. The VAR lag order is in line with the maximum order used by Ammer and Freeman (1995), and experimentation with other lag orders produced similar results. In addition, to maintain compatibility with data from other countries in simulations and data decomposition, the CPI instead of the retail price index was used for the U.K. Using the latter alternatively did not alter the qualitative results reported here.
Ammer, John, and Richard T. Freeman, "Inflation Targeting in the 1990s: The Experiences of New Zealand, Canada, and the United Kingdom," Journal of Economics and Business, 47:2, 1995, 165-192.
Andrews, Donald W. K., "Tests for Parameter Instability and Structural Change With Unknown Change Point," Econometrica, 61:4, 1993, 821-856.
Bernanke, Ben S., and Frederic S. Mishkin, "Inflation Targeting: A New Framework for Monetary Policy," Journal of Economic Perspectives, 1:2, 1997, 97-116.
Bernanke, Ben S., Thomas Laubach, Frederic S. Mishkin, and Adam S. Posen, Inflation Targeting: Lessons from International Experience, Princeton University Press, Princeton, N.J., 1999.
Davidson, Russell, and James G. MacKinnon, Estimation and Inference in Econometrics, Oxford University Press, New York, 1993.
Doan, Thomas, RATS User's Manual, Estima, Evanston, Ill., 1992.
Dueker, Michael, and Andreas M. Fischer, "Do Inflation Targets Redefine Central Bank Inflation Preferences? Results from an Indicator Model," in Koos Alders, Kees G. Koedijk, Clemens J. M. Kool, and Carlo Winder (eds.), Monetary Policy in Converging Europe, Kluwer, Boston, 1996, 21-37.
Engle, Robert F., and Joao Victor Issler, "Estimating Common Sectoral Cycles," Journal of Monetary Economics, 35:1, 1995, 83-113.
Engle, Robert F., and Sharon Kozicki, "Testing for Common Features," Journal of Business and Economic Statistics, 11:4, 1993, 369-395.
Hansen, Bruce E., "Tests for Parameter Instability in Regressions with l(1) Processes," Journal of Business and Economic Statistics, 10:3, 1992, 321-335.
Huh, Chan, "Inflation Targeting," Federal Reserve Bank of San Francisco, Economic Letter, 1997, No. 97-04.
Johansen, Soren, "Statistical Analysis of Cointegration Vectors," Journal of Economic Dynamics and Control, 12:2/3, 1988, 231-254.
Johansen, Soren, and Katarina Juselius, "The Full Information Maximum Likelihood Procedure for Inference on Cointegration - with Application to the Demand for Money," Oxford Bulletin of Economics and Statistics, 25:2, 1990, 169-210.
Kahn, George A., and Klara Parrish, "Conducting Monetary Policy with Inflation Targets," Economic Review (Federal Reserve Bank of Kansas City) 83:3, 1998, 5-32.
Lumsdaine, Robin L., and David H. Papell, "Multiple Trend Beaks and the Unit-Root Hypothesis," Review of Economics and Statistics, 79:2, 1997, 212-218.
McCallum, Bennett T., "Inflation Targeting in Canada, New Zealand, Sweden, the United Kingdom, and in General," Working Paper No. 5579, National Bureau of Economic Research, 1996.
Mishkin, Frederic S., and Adam S. Posen, "Inflation Targeting: Lessons from Four Countries," Economic Policy Review, (Federal Reserve Bank of New York), 3:3, 1997, 9-110.
Nadal-De Simone, Francisco, "Inflation Targeting and Monetary Targeting in a Small Open Economy: the Behavior of Price Variables," Working Paper Reserve Bank of New Zealand, WP96/4, Aukland, 1996. Newey, Whitney K., and Kenneth D. West, "A Simple Positive-Definite Heteroskedasticity and Autocorrelation Consistent Covariance Matrix," Econometrica, 55:3, 1987, 703-708.
Osterwald-Lenum, Michael, "A Note with Quantiles of the Asymptotic Distribution of Some Important Maximum Likelihood Cointegration Test Rank Statistics," Oxford Bulletin of Economics and Statistics, August 1992, 461-472.
Perron, Pierre, "The Great Crash, the Oil Price Shock and the Unit Root Hypothesis," Econometrica, 57:6, 1989, 1361-1401.
Phillips, Peter C. B., and Bruce E. Hansen, "Statistical Inference in Instrumental Variables Regression with I(1) Processes," Review of Economic Studies, 57:1, 1990, 99-125.
Rao, C. D., Linear Statistical Inference and Its Applications, Wiley, New York, 1973.
Vahid, Farshid, and Robert F. Engle, "Common Trends and Common Cycles," Journal of Applied Econometrics, 8:4, 1993, 341-360.
Zivot, Eric, and Donald W. K. Andrews, "Further Evidence on the Great Crash, the Oil-Price Shock, and the Unit-Root Hypothesis," Journal of Business and Economic Statistics, 10:3, 1992, 251-270.
|Printer friendly Cite/link Email Feedback|
|Publication:||Contemporary Economic Policy|
|Date:||Jul 1, 1999|
|Previous Article:||Assessing the efficiency of public schools using data envelopment analysis and frontier regression.|
|Next Article:||E-money and payment system risks.|