The Dual-Mandate Debate: What Do Central Banks Really Target?

Abstract

Inflation targeting, a monetary policy framework, is criticized for its narrow mandate of safeguarding price stability only and neglecting other equally important macroeconomic variables. This negligence, according to the critics, might have had a role in the unprecedented, real business-cycle fluctuations observed in the past. Hence, they advocate for mandating central banks with equally emphasizing employment and output growth along with inflation. Theoretical claims aside, the literature does not present any empirical evidence on how to determine whether a central bank adheres to a single or a dual mandate. This study is aimed at filling this gap by analyzing the reaction functions of various central banks, including the ones targeting inflation and the ones with no specific targets. Using the panel data from OECD countries, our findings question the prevalent theoretical misunderstanding in the literature: the central banks with no specific targets (the dual-mandate monetary policy regimes) appear to be targeting the rate of inflation only, whereas the central banks that are thought to have a single-mandate seem to be targeting inflation, output growth, and unemployment. These results are significant, both statistically and economically, and question the baseless criticism of the inflation-targeting regime for neglecting employment and output growth.

1. Introduction

Inflation targeting (IT), a monetary policy regime, has been adopted by over 30 central banks around the world. The Bank of Canada (BoC) is one of the pioneers that adopted IT in 1992 when the Canadian federal parliament mandated the BoC to maintain price stability by targeting a low and stable inflation rate of 2%.1 Ever since, the BoC has maintained a band of 1–3% for its annual inflation rate. The BoC is required, constitutionally, to seek the renewal of its mandate of price stability from the Canadian federal parliament every five years.

In the fall of 2018, I received an email from a friend and a mentor of mine, seeking my signature on a petition that requested the Canadian Federal Finance Minister at the time to switch the BoC’s single-mandate policy of inflation targeting to a dual/multiple mandate(s), targeting the inflation rate, output growth, and full employment simultaneously.

The said petition garnered a positive response and was signed by 61 Canadian economists, including myself. It was submitted to the Canadian Ministry of Finance for further consideration. Subsequently, a group of the petitioning economists were invited to meet the Finance Minister to present their views regarding the Bank of Canada’s mandate.

As the signatory of the petition, I tended to agree with my colleagues on the plausibility of a dual/multiple mandate(s) for the BoC. However, the same petition also stimulated a research question: is there any empirical evidence that could support or refute the claim of the movers of this petition? Indeed, this paper is an attempt to verify this claim. However, before we embark on empirically assessing whether a dual/multiple mandate(s) will bear fruitful outcomes for the economy in general, we first need to verify the premise of this single- vs. dual-mandate debate: Can we categorize central banks into single- and dual-mandate CBs based on their choice of various monetary policy regimes? In other words, if a central bank has adopted inflation targeting as its monetary policy regime, can we classify it as having a single mandate? Similarly, for a central bank that does not explicitly adhere to a particular monetary policy regime, is it acceptable to classify it as having a dual mandate or multiple mandates? Of course, we are very keen to finding out which monetary policy regime, a single or a dual mandate, can lead to a better economic outcome. However, due to time and space constraints, we just focus on assessing the basic premise of the debate, which is the classification of the central banks into single- and dual-mandate CBs.

The rest of the paper proceeds as follows: Section 2 discusses the advantages and disadvantages of a single-mandate monetary policy regime that have basically caused the division among macroeconomists, namely those who support a single-mandate monetary policy regime (inflation targeting) and those who oppose it. Section 3 formulates the hypothesis for this study. Section 4 presents the model utilized for our empirical assessment. Section 5 presents the data and methodology. Section 6 analyzes the regression results. Section 7 concludes the paper.

2. Single Mandate Versus Dual Mandate: Advantages and Disadvantages

While there is considerable literature on single-mandate monetary policies of inflation targeting, there is hardly any literature that highlights the advantages associated with a dual- or multiple-mandate monetary policy framework. Therefore, we are going to rely on the IT literature to deduce the potential advantages and disadvantages of a single mandate. Looked at from the opposite side, the advantages of IT could possibly be considered, to some extent, as disadvantages of a dual mandate and so on. Here are some of the advantages and disadvantages of a single mandate that the proponents and opponents of inflation targeting often cite in the literature:

Svensson (1997) calls inflation targeting a “potential commitment mechanism”: due to its focus on price stability and sharing a specific target with the public, IT puts the credibility of policymakers on the line, should they deviate from their commitment, and, therefore, can be held accountable. Bernanke and Mishkin (1997) think that inflation targeting can solve the time-inconsistency problem often shrouding monetary policy. They also call inflation targeting the sole protector of price stability from the “pass through” of unexpected shocks into inflation, which, in turn, would keep nominal interest rates stable due to the stable inflation expectations. Rudebusch and Svensson (1999) credit inflation targeting with the “locking-in” feature of expectations in addition to the reduction in volatility of output gaps. Schaechter et al. (2000) praise inflation targeting for bolstering the motivation for the necessary institutional reforms needed for taming inflation. Truman (2003) sees inflation targeting as building monetary policy’s credibility and anchoring inflation expectations better than other monetary policy regimes. Goodfriend and King (2005) praise inflation targeting for its flexibility, which enables policymakers to focus on medium-to-long-term horizons without worrying about short-term shocks. Batini and Laxton (2007) call inflation targeting an efficient policy because the cost of policy failure under inflation targeting is much lower than the loss due to the policy failure under exchange rate targeting, for instance, where reserve losses may lead to a financial crisis. De Mendonça and De Guimarães e Souza (2012) believe that inflation targeting helped developing countries achieve a significant reduction in inflation rates. Yamada (2013) assigns a superior performance to inflation targeting than exchange rate targeting in lowering inflation rates.

Inflation targeting has had its share of criticism as well, where the opponents consider some unnecessary constraints being imposed on the policymakers: The first, and a common critique, that we have also mentioned at the start of this paper, is that the IT monetary policy is an overly narrow approach with a single-objective focus, the price-level stability, thus limiting policymakers’ discretion. There seems to be no discretion at all when it comes to addressing the other important macroeconomic variables, such as output and unemployment, particularly when this narrow, single-objective approach of taking swift actions in fighting inflation may lead to some serious repercussions on output and employment. Some opponents, such as Arestis and Sawyer (2003), even go on to question whether any central bank has the ability to control inflation. They cite international supply shocks, fiscal policy, and domestic wage negotiations as some of the major factors that could breed sudden episodes of inflation but are mainly out of the control of a central bank, rendering their actions to curb inflation ineffective. If an inflation-targeting central bank, for example, applies its conventional tool of contracting its money supply to fight inflation, we would expect the output to fall, which will further induce negative supply shocks, thus worsening, and not enhancing, the economic state. Some critics also blame inflation targeting for granting too much discretion to central banks, leading to the “moving the goalposts” phenomenon, which may endanger the credibility of a central bank. Others point to the fact that most of the IT-adopting central banks were experiencing higher inflation rates, and it was appropriate for them to adopt inflation targeting in order to fight the higher inflation rates at the time. This fact would support the view of inflation targeting being a disinflationary tool. The proponents of inflation targeting, however, do not consider inflation targeting as a mere stabilizing policy and believe that inflation targeting could do more harm than good if implemented as a disinflationary tool in an economy that is experiencing higher inflation rates. Instead, they claim that inflation targeting helps “lock in” the already-gained low levels of inflation rates. Some opponents see this as a replacement for other nominal anchors, specifically an exchange rate regime. Since a central bank cannot have multiple anchors at one time, an IT-adopting central bank must not engage in exchange rate management but instead should adopt a free-floating exchange rate mechanism. Now, this laissez-faire approach may suit those who enjoy having hard currencies or a stock pile of them, but countries with soft currencies do not have the luxury of leaving their exchange rate unguarded in the face of speculative attacks, supply shocks, and abrupt capital flows, which would only result in an extreme volatility propagating into their macroeconomies.

Lopez (2005) credits inflation targeting with the desired welfare outcomes by stabilizing and safeguarding the households’ purchasing power. Brito and Bystedt (2010) as well as Saqib and Aggarwal (2017) question the viability of inflation targeting for emerging economies, as they find no evidence of any improvement in the macroeconomic performance of these emerging economies, such as the behaviors of inflation and output growth, following the adoption of inflation targeting. Yamada (2013) concludes with the ineffectiveness of inflation targeting on economic performance among developed and developing economies. Lavoie and Seccareccia (2012) criticize inflation targeting for its negative impact on labor markets due to the fact that with a narrowly focused mandate, inflation-targeting adoption may breed rising wages that, in turn, could lead to adverse effects on overall employment, thus adversely affecting the macroeconomic performances of the adopting economies as opposed to enhancing it. Valera et al. (2018) do not see any significant impact of IT adoption on output growth among a survey of 113 studies on inflation targeting. Seccareccia and Khan (2019) conclude that meeting the inflation target for an IT-adopting central bank has a trade-off in the form of chronically high long-term unemployment. Finally, Khasawneh (2020) and Boujlil et al. (2020) blame narrow monetary-policy regimes, such as inflation targeting, for the commercial banks’ bad loans as well as for the overall public debt instruments.

To sum up our literature review, on the one hand, there are macroeconomists who praise the single-mandated monetary policy regime, namely inflation targeting, as having advantages, such as IT providing a potential commitment mechanism, thus lending credibility to policymakers and solving the time-inconsistency problem. They also praise IT for being the sole protector of price stability from the “pass through” of unexpected shocks; IT also has “locking-in” and “anchoring” features that help reduce the output gap volatility and motivate institutional reforms. Finally, the IT proponents regard it as an efficient and superior-performing policy. As for the opponents, on the other hand, a single-mandate monetary policy regime such as inflation targeting offers an overly narrow approach, often neglecting some very important macroeconomic variables, output and employment. This narrow approach, according to the opponents, can also lead to negative supply shocks, further exacerbating economic activities. Finally, the authors who advocate for a dual- or multiple-mandate monetary policy blame the single-mandate CBs for the “moving the goalposts” phenomenon, which could affect their credibility.

3. The Hypothesis

In this study, we intend to empirically examine whether a single-mandate central bank indeed adheres to its single mandate of targeting one specific variable, particularly the rate of inflation, and whether a dual- or multiple-mandate central bank indeed has more than one target, particularly output growth and employment, in addition to the rate of inflation.

In other words, we want to find out—among our sample of OECD countries—which central bank has a single monetary policy goal and who has multiple monetary policy goals. An IT-adopting central bank, often labeled as obeying a single-mandate monetary policy regime, should be targeting inflation only, whereas a non-IT-adopting central bank, considered to be obeying a dual-or multiple-mandate monetary policy regime, should target other macroeconomic variables as well, along with the rate of inflation. To elaborate further though, most prominent central banks, led by the Federal Reserve and European Central Bank (ECB), claim to have a dual-mandate monetary policy, emphasizing equally price-level stability as well as employment or output growth; this duality does not show up in their theoretically stated goals. The following evidence should suffice to refute their dual-mandate claim, at least theoretically:

(a) In a speech, delivered on 27 August 2020, the Chair of the Federal Reserve, Jerome Powell, stated that “In conducting monetary policy, we will remain highly focused on fostering as strong a labor market as possible for the benefit of all Americans. And we will steadfastly seek to achieve a 2 percent inflation rate over time.” One can easily notice here that while the price-stability mandate is clearly stated as a 2% inflation rate, the other equally important mandate, employment, gets no specific goal. The following excerpt regarding the second part of the Fed’s dual mandate, employment, renders the dual-mandate claim almost meaningless, theoretically:

The other part of the Fed’s dual mandate is maximum employment. The concept of maximum employment can be thought of as the highest level of employment that the economy can sustain over time. Of course, measuring this concept is hard because the level of maximum employment varies over time with business conditions, demographics, labor market regulations, and other factors. Rather, the Fed considers a wide range of employment indicators to estimate the shortfalls of employment from its maximum level. In short, the Fed does not have a numerical target for the level of employment; rather, the Fed analyzes economic conditions using a wide range of data to design policies that achieve maximum employment.2

(b) Ahmad and Brown (2017) examine the ECB’s “Two-Pillar Monetary Policy Strategy”, where they assess the ECB’s response—when designing its monetary policy and setting the policy rate—to the rate of inflation and money supply growth during the great financial crisis (GFS) of 2008. They find the rate of inflation to be the only significant variable influencing the ECB’s response, whereas monetary growth does not appear to be significant in their empirical findings.

A better way, therefore, to validate the above-mentioned claims would be to look for the empirical evidence. This can be achieved by estimating the central banks’ reaction functions in finding out the likelihood of their adherence to their claim: if we find a single variable, namely the rate of inflation, statistically and economically appearing as significant, we can classify that central bank as having a single-mandate regime, and if we find more variables, in addition to the rate of inflation, statistically and economically appearing as significant, we can classify that central bank as following a dual-mandate policy framework.

4. The Model

For testing our hypothesis, as stated in the previous section, we use the following Taylor rule as the central bank’s reaction function:

i_t = α + π_t−1 + r* + β (π_t − π^T) + γ (y_t − y^P) + λ (u_t − u^T) + ε_t

(1)

where i_t represents the central bank’s key policy rate, π_t is the inflation rate; r* is the natural rate of interest; π^T represents the target rate of inflation; y_t is the output, whereas y^P is the potential output; u_t is the unemployment rate; and u^T is the target rate of unemployment. Finally, β, γ, and λ are the coefficients that measure the gaps between these variables, whereas ε_t is the error term. Two points warrant clarification here: First, the non-IT-adopting central banks in our data sample have no specific target rate of inflation, and so we use inflation rate only (without the gap) when estimating their reaction functions. Second, the unemployment gap is rarely given any significant importance, not to mention the disagreed-upon natural rate of unemployment (NAIRU), and so we use the unemployment rate only when estimating the reaction function. Given these constraints, our empirical model (the updated Taylor rule) simplifies to the following Base Model:

i_t = α + β π_t + γ y_t + λ u_t + ε_t

(2)

Noteworthy, it is quite plausible to assume that some central banks might not consider all the three variables—inflation, output growth, and unemployment—in their control matrix when designing their monetary policy. Instead, they may consider just two: inflation and output growth or inflation and unemployment, perhaps. Remember, as stated earlier in this section, some policymakers led by the Federal Reserve emphasize only price stability and employment when designing their monetary policy. It is, therefore, not uncommon for some central banks to estimate their reaction function using only two macroeconomic variables. We, therefore, use the following model to re-estimate the reaction function for the category of central banks that equally emphasize the rate of inflation and the employment:

i_t = α + β π_t + λ u_t + ε_t

(3)

We then substitute the employment with the output growth for the central banks that are interested in maintaining their output growth on a steady-state path along with price stability and re-estimate the reaction function as follows:

i_t = α + β π_t + γ y_t + ε_t

(4)

With these three empirical models—Equations (2)–(4)—we embark on our empirical exercise. Please note that in addition to these bivariate and multivariate empirical models, we also employ univariate models, where the policy rate is regressed on each control variable separately. Nonetheless, the results remain the same throughout the three specifications, albeit with some minor changes.

There is one last point with regard to our empirical settings: In addition to tweaking the empirical models, we also consider different clusters of our data sample: We first consider the entire sample when estimating the central banks’ reaction functions. We label it the “Grand Sample”. We then divide this “Grand Sample” into two sub-samples: the IT Group and the Non-IT Group. The former includes the central banks of our sample that opted for inflation targeting as their monetary policy framework, whereas the latter includes the central banks that did not adopt inflation targeting as their monetary policy framework. More importantly, this grouping should help us in concluding our single- vs. dual-mandate debate as to which group can be classified as obeying the single-mandate framework and which one has the dual-mandate regime in place.

5. The Data

We use the panel data for 33 OECD member countries: 17 of them have adopted the inflation-targeting regime, the IT-adopting countries, often labeled as having a single-mandate monetary policy of targeting the rate of inflation only; whereas the remaining 16 countries, the non-IT adopters, can be classified as having a dual- or multiple-mandate monetary policy regime in place.

Table 1. The OECD IT club members with their adoption date and target.

Country	Adoption Date	Inflation Target	Country	Adoption Date	Inflation Target
Australia	1993	2–3%	Mexico	2001	5.0% ± 1.5%
Canada	1991	1–3%	N. Zealand	1989	1–3%
Chile	1999	3% ± 1%	Norway	2001	2.5%
Czech Rep.	1997	2% ± 1%	Poland	1998	2.5% ± 1%
Finland ***	1990	2.5%	S. Korea	1998	3% ± 1%
Hungary	2001	3%	Spain ***	1990	2%
Iceland	2001	2.5%	Sweden	1995	2%
Israel	1997	1–3%	UK	1992	2%
			USA	2012	2%

Note: *** Finland and Spain left the IT club in 1999 after joining the European Union. Sources: Gemayel et al. (2011), Hammond (2012), the IMF, the World Bank, and major central banks’ websites.

below lists the 17 OECD countries that have adopted inflation targeting.3

As for the data sample, it is constructed from annual observations over a time horizon of 30 years, from 1990 to 2019. We exclude the years prior to 1990, as the vast majority of IT adopters joined the IT club in 1990 or later. We also exclude the period after 2019 due to the COVID-19 pandemic that certainly had a great impact on the design of monetary policies around the globe in the aftermath of the pandemic. Almost all central banks were scrambling to deal with the sudden halt to the economic activity in the couple of years following the COVID-19 pandemic and, thus, were resorting to similar monetary policies, namely addressing the liquidity trap that engulfed the world economy.

As for the variable choice, since we are interested in estimating the central bank’s reaction function, we stick to the macroeconomic variables commonly used in the Taylor rule model, namely the monetary policy rate (short-run interest rate), the rate of inflation (CPI), the output growth (GDP growth rate), and the rate of unemployment. A summary of our panel data sample is presented in

Table 2. Summary statistics: the Grand Sample (both IT Group and Non-IT Group).

Variables	Observations	Mean	Std Dev	Min.	Max.
YGAP: output gap	1001	−0.007	0.031	−0.155	0.145
ITR: inflation-targeting regime	1080	0.343	0.475	0.000	1.000
CPI: consumer price index (inflation)	1080	0.060	0.387	−0.045	9.517
SRIR: SR interest rate (the policy rate)	1029	0.047	0.054	−0.008	0.482
LSRIR (t − 1): lagged policy interest rate	1017	0.051	0.056	−0.008	0.482
UNEMP: unemployment rate	1043	0.078	0.043	0.005	0.275
YGROW: output growth rate	1035	0.026	0.029	−0.144	0.250
INFGAP: inflation gap (CPI minus target rate)	1080	0.030	0.385	−0.081	9.453

Data source: OECD (2020) Data extracted on 02 Aug 2020 16:41 UTC (GMT) from OECD.Stat.

below, where, in addition to thestated variables above, we also have the output gap, the inflation gap, and the lagged interest rate. These variables, although not used in the estimation of the reaction function due to the constraints stated in the previous section, are utilized when checking for the robustness of the results. The data were extracted from the OECD website, “OECD.Stat”.

6. Regression Results

6.1. The Grand Sample

We first estimate the reaction function for all the central banks in our sample using the Base Model, Equation (2), to see which variables are targeted specifically by these central banks, assuming that they act as a homogeneous group of countries.4 The results are presented in

Table 3. The reaction function results for the “Grand Sample” central banks.

The Base Model: it = α + βπt + γ yt + λut + εt
Control Variables	Coefficients β, γ, and λ	Std. Errors	t	p > |t|	[95% Conf. Interval]
Constant	0.012 ***	0.002	5.430	0.000	0.008	0.017
CPI	1.068 ***	0.025	42.760	0.000	1.019	1.117
YGROW	0.035	0.033	1.060	0.288	−0.030	0.100
UNEMP	0.005	0.022	0.220	0.827	−0.039	0.048
Source	SS	DF	MS		Observations	1014
Model	1.654	3	0.551		F (3, 1010)	615.49
Residual	0.905	1010	0.001		Prob > F	0.000
Total	2.559	1013	0.003		R-squared (R2)	0.646
					Adj R-squared	0.645
					Root MSE	0.030

Note: Short-run interest rate (SRIR), the central bank’s policy rate, is the dependent variable. The SRIR is used as a proxy for the estimation of the reaction function of all the central banks. The asterisks next to the coefficients (***) represent significance levels of 1%.

below. Note that the only variable that appears highly significant in Table 3, both statistically and economically, is the rate of inflation (CPI), which, according to the results, exerts a significant influence, more than one for one, on the monetary policy rate: a one-percentage-point change in the rate of inflation (CPI) would lead to a little over a one-percentage-point (1.07%) change in the policy rate. The results in Table 3 also show that other macroeconomic variables, such as the output growth (YGROW) and unemployment (UNEMP), have insignificant coefficients, both statistically and economically, which could be interpreted as having a very low influence on the monetary policy rate for the Grand Sample. In other words, the OECD countries behave as having a single-mandate monetary policy regime, where inflation plays the leading role, whereas the output growth and employment play somewhat secondary roles when it comes to setting the policy rate, even though 50% of the countries in this group are non-IT adopters who ought to be targeting more than the rate of inflation! Nonetheless, we cannot deduce any plausible conclusion from these results due to the clear heterogeneity among the group countries.

Next, we repeat the same regression exercise, but with a little tweak:

Table 4. The reaction function results for the “Grand Sample” central banks after dropping “output growth”.

The Base Model (Without Output Growth): it = α + βπt + λut + εt
Control Variables	Coefficients β and λ	Std. Errors	t	p > |t|	[95% Conf. Interval]
Constant	0.0134 ***	0.002	6.42	0.000	0.009	0.017
CPI	1.059 ***	0.025	42.91	0.000	1.010	1.107
UNEMP	0.007	0.022	0.320	0.746	−0.036	0.050
Source	SS	DF	MS		Observations	1020
Model	1.675	2	0.837		F (2, 1017)	921.22
Residual	0.925	1017	0.000		Prob > F	0.000
Total	2.600	1019	0.003		R-squared (R2)	0.644
					Adj R-squared	0.643
					Root MSE	0.030

Note: The asterisks next to the coefficients *** represent significance levels of 1%.

below presents the results after we drop the output growth from our Base Model, whereas

Table 5. The reaction function results for the “Grand Sample” central banks after dropping “unemployment”.

The Base Model (Without Unemployment): it = α + βπt + γ yt + εt
Control Variables	Coefficients β and γ	Std. Errors	t	> |t|	[95% Conf. Interval]
Constant	0.013 ***	0.001	9.27	0.000	0.010	0.016
CPI	1.062 ***	0.023	45.90	0.000	1.016	1.107
YGROW	0.036	0.033	1.110	0.267	−0.028	0.100
Source	SS	DF	MS		Observations	1016
Model	1.909	2	0.954		F (2, 1013)	1062.5
Residual	0.910	1013	0.000		Prob > F	0.000
Total	2.819	1015	0.003		R-squared (R2)	0.677
					Adj R-squared	0.676
					Root MSE	0.030

Note: The asterisks next to the coefficients *** represent significance levels of 1%.

presents the results when unemployment is substituted with output growth. In both cases, we observe that the only significant variable, both statistically and economically, appears to be the rate of inflation. The other two variables, unemployment and output growth, having insignificant coefficients, could be thought as having very little influence over the monetary policy rate. In light of the findings in Table 3, Table 4 and Table 5, we can postulate that the central banks in the OECD countries, when grouped together, generally behave as obeying a single-mandate monetary policy regime, mainly emphasizing price-level stability and having relatively lower consideration for other macroeconomic variables, particularly employment and output growth.

6.2. The Non-IT Group

We now turn to our sub-sample of countries that may be classified as having a dual-mandate or a multiple-mandate monetary policy framework, perhaps due to the fact that they do not explicitly target inflation or any other specific variable when designing their monetary policy.

As stated earlier, we will be utilizing three models, namely Equations (2)–(4), to estimate the reaction function of this group of central banks, labeled as the Non-IT Group. We expect, based on theoretical understanding from the literature, that these central banks consider at least one more variable (growth or employment) in addition to price-level stability when designing their monetary policy. This should translate, empirically, into at least one more control variable in addition to the rate of inflation to appear as statistically and economically significant. The results are presented in

Table 6. Reaction function results for the non-IT central banks: the Base Model.

The Base Model: it = α + βπt + γ yt + λut + εt
Control Variables	Coefficients β, γ, and λ	Std. Errors	t	p > |t|	[95% Conf. Interval]
Constant	0.009 ***	0.003	3.04	0.003	0.003	0.015
CPI	0.968 ***	0.055	17.59	0.000	0.860	1.076
YGROW	−0.022	0.041	−0.54	0.591	−0.103	0.059
UNEMP	0.022	0.028	0.77	0.441	−0.034	0.078
Source	SS	DF	MS		Observations	531
Model	0.283	3	0.094		F (3, 527)	107.04
Residual	0.465	527	0.000		Prob > F	0.000
Total	0.749	530	0.001		R-squared (R2)	0.379
					Adj R-squared	0.375
					Root MSE	0.030

Note: Short-run interest rate (SRIR), the central bank’s policy rate, is the dependent variable. The SRIR is used as a proxy for the estimation of the reaction function of all the central banks. The asterisks next to the coefficients *** represent significance levels of 1%.

below.

Ironically, the results do not support the critics of inflation targeting that a dual-mandate monetary policy framework gives equal importance to the other macroeconomic variables, employment and growth. The only significant variable in the control matrix of our sub-sample of the Non-IT Group, both statistically and economically, appears to be the rate of inflation (CPI), which accounts for an almost one-for-one influence on the monetary policy instrument: a one-percentage-point change in the CPI inflation leads to a 0.97% change in the policy rate of interest. The coefficients for unemployment and growth are statistically insignificant. So, inflation enjoys primary consideration, whereas growth and unemployment receive secondary consideration when the policymakers are designing the monetary policy, as per our results in Table 6.

Let us now look at our findings of the reaction function for the same sub-sample after we drop one control variable, namely the output growth, and keep unemployment along with our main control variable, the rate of inflation.

Table 7. The reaction function results for the non-IT central banks after dropping output growth from the Base Model.

The Base Model Without Output Growth: it = α + βπt + λut + εt
Control Variables	Coefficients β and λ	Std. Errors	t	p > |t|	[95% Conf. Interval]
Constant	0.008 ***	0.003	2.87	0.004	0.003	0.014
CPI	0.978 ***	0.054	18.10	0.000	0.872	1.084
UNEMP	0.027	0.028	0.94	0.348	−0.029	0.083
Source	SS	DF	MS		Observations	533
Model	0.297	2	0.148		F (2, 530)	166.06
Residual	0.474	530	0.001		Prob > F	0.000
Total	0.772	532	0.001		R-squared (R2)	0.385
					Adj R-squared	0.382
					Root MSE	0.030

Note: The asterisks next to the coefficients *** represent significance levels of 1%.

above presents the estimation results for Equation (3): once again, we observe that the CPI inflation is highly significant, both statistically and economically (0.978), meaning that a one-percentage-point change in the CPI inflation leads to almost one-percentage-point (0.98%) change in the policy rate, whereas the unemployment control variable appears to be totally insignificant. We can, therefore, deduce from these empirical results, for the sub-sample of the Non-IT Group, that the central banks in this group assign the key role to inflation and a secondary role to unemployment. We repeat the same exercise and substitute the output growth for unemployment in our control matrix. The results are presented in

Table 8. The reaction function results for the non-IT central banks after dropping unemployment from the Base Model.

The Base Model Without Unemployment: it = α + βπt + γ yt + εt
Control Variables	Coefficients β and γ	Std. Errors	t	p > |t|	[95% Conf. Interval]
Constant	0.011 ***	0.002	5.76	0.000	0.007	0.015
CPI	0.971 ***	0.055	17.72	0.000	0.864	1.079
YGROW	−0.025	0.041	−0.62	0.537	−0.106	0.055
Source	SS	DF	MS		Observations	531
Model	0.283	2	0.141		F (2, 528)	160.39
Residual	0.465	528	0.000		Prob > F	0.000
Total	0.749	530	0.001		R-squared (R2)	0.378
					Adj R-squared	0.376
					Root MSE	0.029

Note: The asterisks next to the coefficients *** represent significance levels of 1%.

below, telling the same story: the only significant control variable happens to be the rate of inflation, which is highly significant, both statistically and economically: a 1-percentage-point change in the CPI inflation leads to a 0.98-percentage-point change in the policy rate, whereas the output growth receives less attention from the policymakers in this group of central banks.

6.3. The IT Group

Finally, we turn to our second sub-sample, the IT Group, which includes the countries that have opted for the inflation-targeting regime as their monetary policy framework. As with the previous samples, we will be estimating our three empirical models separately. Given the criticism this group of central banks has been subjected to in the literature, namely the narrow approach of targeting inflation while putting the other equally important macroeconomic variables on the back burner, we expect to see inflation only appearing as highly significant in the three models’ estimation results.

The results for the Base Model, Equation (2), are presented in

Table 9. The reaction function results for the inflation-targeting central banks the Base Model.

The Base Model: it = α + βπt + γ yt + λut + εt
Control Variables	Coefficients β, γ, and λ	Std. Errors	t	p > |t|	[95% Conf. Interval]
Constant	0.010 ***	0.004	2.50	0.013	0.002	0.017
CPI	1.039 ***	0.028	37.31	0.000	0.984	1.093
YGROW	0.169 ***	0.058	2.92	0.004	0.055	0.282
UNEMP	0.106 **	0.048	2.22	0.027	0.012	0.200
Source	SS	DF	MS		Observations	423
Model	1.172	3	0.391		F (3, 419)	485.19
Residual	0.337	419	0.000		Prob > F	0.000
Total	1.509	422	0.003		R-squared (R2)	0.776
					Adj R-squared	0.774
					Root MSE	0.028

Note: Short-run interest rate (SRIR), the central bank’s policy rate, is the dependent variable. The SRIR is used as a proxy for the estimation of the reaction function of all the central banks. The asterisks next to the coefficients ** and *** represent significance levels of 5% and 1%.

below, and they look surprisingly different than our expectations. All three control variables in our empirical model appear highly significant, both statistically and economically. First, the rate of inflation (CPI) is very highly significant: a one-percentage-point change in the CPI inflation leads to a more than one-percentage-point (1.04%) change in the monetary policy rate. Second, the output growth is also highly significant: a one-percentage-point change in the output growth leads to about one-sixth of a percentage-point (0.17%) change in the monetary policy rate. Third, and lastly, the unemployment is also significant: a one-percentage-point change in the unemployment leads to a little over one-tenth of a percentage-point (0.106%) change in the monetary policy rate. All of these results are highly significant, both statistically and economically, and they question the validity of the ongoing criticism directed in the literature at the inflation-targeting regime. These results appear to be supporting the claim of the proponents of inflation targeting that a single target does not necessarily imply a 100% emphasis on one variable, the price-level stability; instead, the underlying factors receive equal emphasis as well, such as the output and unemployment.

We now perform the last set of regressions on the remaining two models, namely Equations (3) and (4), that is, after we drop one of the two control variables at a time. The results for these modified models are presented in Table 9 and

Table 10. The reaction function results for the inflation-targeting central banks after dropping output growth from the Base Model.

The Base Model Without Output Growth: it = α + βπt + λut + εt
Control Variables	Coefficients β and λ	Std. Errors	t	p > |t|	[95% Conf. Interval]
Constant	0.015 ***	0.003	4.35	0.000	0.008	0.021
CPI	1.019 ***	0.028	36.54	0.000	0.964	1.073
UNEMP	0.108 **	0.048	2.23	0.026	0.013	0.204
Source	SS	DF	MS		Observations	427
Model	1.168	2	0.584		F (2, 424)	695.69
Residual	0.356	424	0.000		Prob > F	0.000
Total	1.524	426	0.003		R-squared (R2)	0.766
					Adj R-squared	0.775
					Root MSE	0.029

Note: The asterisks next to the coefficients ** and *** represent significance levels of 5% and 1%.

.

These results are in line with the ones we obtained when estimating the Base Model, Equation (2), and presented in Table 8 above. Notice that the unemployment control variable in Table 10 remains highly significant in the second model estimation as well, where a one-percentage-point change in the unemployment leads to a 0.11% change in the monetary policy rate after dropping the output growth. The comments by the Chair of the Federal Reserve, Jerome Powell, stated in Section 2 seem to be supported by the results in Table 10. As for the central banks that target output growth along with the rate of inflation,

Table 11. The reaction function results for the inflation-targeting central banks after dropping unemployment from the Base Model.

The Base Model Without Unemployment: it = α + βπt + γ yt + εt
Control Variables	Coefficients β and γ	Std. Errors	t	p > |t|	[95% Conf. Interval]
Constant	0.017 ***	0.002	6.97	0.000	0.012	0.021
CPI	1.040 ***	0.025	41.31	0.000	0.991	1.090
YGROW	0.162 ***	0.058	2.80	0.005	0.048	0.276
Source	SS	DF	MS		Observations	425
Model	1.400	2	0.700		F (2, 422)	854.31
Residual	0.346	422	0.000		Prob > F	0.000
Total	1.746	424	0.004		R-squared (R2)	0.802
					Adj R-squared	0.801
					Root MSE	0.029

Note: The asterisks next to the coefficients *** represent significance levels 1%.

presents the evidence where the output growth remains highly significant along with the inflation after we drop the unemployment in our third and last empirical model: a one-percentage-point change in the output growth leads to a 0.162% change in the monetary policy rate after dropping the unemployment.

7. Concluding the Debate

There are many economists who are skeptical of the inflation-targeting regime, perhaps due to its narrowly defined single mandate of targeting price-level stability and paying less attention to the other equally important variables, employment and output growth. This biased approach by the policymakers could lead to real business cycle fluctuations. Indeed, this was the case, according to these skeptical economists, when the global economy was subjected to the unprecedented fluctuations that bred crises, such as the great financial crisis of 2008 over the past few decades. One potential solution to mitigating these fluctuations, as per these economists, would be to reconsider the mandate of central banks that adhere to a single-mandate monetary policy regime, namely inflation targeting: these central banks ought to be targeting employment and output growth as well, in addition to their core target of price-level stability. This argument has led to the so-called single-mandate versus dual-mandate debate in the literature. This mandate debate thus far has mainly relied on theoretical foundations. The literature offers no empirical evidence on whether the single-mandate central banks indeed target inflation only, and whether the non-inflation-targeters, in fact, have multiple targets when designing monetary policies.

This study aimed to find empirical evidence to conclude the dual-mandate debate. Using a panel dataset from 33 OCED member countries, we estimated the reaction functions of central banks in these countries. When treating them as one large group of countries, the results show that they all have one common target and that is the rate of inflation (price-level stability). However, when these countries are assigned to their appropriate classification, namely the inflation-targeting adopters and the non-IT adopters, we obtain surprising results: For the non-IT group of countries that are thought of as having one or two more targets in addition to inflation, we observe that their central banks are mostly targeting CPI inflation only, thus behaving as single-mandate central banks. The inflation-targeting central banks, on the other hand, are presumed to have a single target of inflation; instead, they appear to be targeting employment and output growth as well, in addition to their core target of inflation. These results are shockingly surprising because the empirical results were expected to be the opposite as per the critics of the inflation-targeting regime: the non-IT central banks should have had dual or multiple targets, whereas the IT central banks should have had a single target of inflation. These results are all significant, both statistically and economically, and they are also robust to various modifications of the empirical model.

In light of our empirical findings, the dual-mandate debate seems to be holding no ground at all.

Inflation targeting, although explicitly a single-mandate monetary policy regime, appears to be assigning equal importance to employment and output growth when targeting price-level stability, whereas the so-called dual- or multiple-mandate monetary policy regimes, in fact, have a single target and behave as a narrowly single-mandate regime!

These findings should suffice to put an end to the unnecessary criticism of inflation targeting as a single-mandate monetary policy regime, which, as per our findings, does not necessarily mean that the central banks only care about inflation when designing their monetary policy; they do consider other important variables, such as the output growth and unemployment. We, therefore, cannot label them as adhering to a single narrow mandate. As for the central banks that do not explicitly adopt the inflation-targeting regime and are assumed to have dual or multiple targets, our findings appear to suggest that they devote much attention to the rate of inflation when setting their policy rate. Noteworthy to state here is that these results must not be interpreted as implying the negligence by these non-IT central banks towards other important variables, such as the output growth and employment. No central bank can afford to do so. But these results do tell us that their consideration of inflation when designing their monetary policy outweighs the consideration that they assign to output growth and unemployment. After all, this study uses a panel dataset, which is often considered as a major shortcoming in the empirical analysis. A better way to verify the findings of this study and draw a concrete conclusion, perhaps, would be to employ a case-study approach, where each central bank is studied alone, using short-period observations, such as monthly or quarterly.

More interestingly, future research should look into not only employing the case-study approach, as stated here, but other aspects of this debate also, particularly the claim of the efficiency of a single-mandate versus a dual/multiple monetary policy regime.