SBP Working Paper Series

No. 79 November, 2016

Assessing the Role of Money versus Interest

Rate in Pakistan

Zafar Hayat

Muhammad Nadim Hanif

STATE BANK OF PAKISTAN

 SBP Working Paper Series
Editor: Muhammad Nadim Hanif

The objective of the SBP Working Paper Series is to stimulate and generate discussions on
different aspects of macroeconomic issues among the staff members of the State Bank of
Pakistan. Papers published in this series are subject to intense internal review process. The
views expressed in these papers are those of the author(s) not State Bank of Pakistan.
© State Bank of Pakistan.

Price per Working Paper (print form)

Pakistan: Rs 50 (inclusive of postage)

Foreign: US$ 20 (inclusive of postage)

Purchase orders, accompanied with cheques/drafts drawn in favor of State Bank of Pakistan,
should be sent to:

Chief Spokesperson,
External Relations Department,
State Bank of Pakistan,
I.I. Chundrigar Road, P.O. Box No. 4456,
Karachi 74000. Pakistan.

Soft copy is downloadable for free from SBP website: http://www.sbp.org.pk

For all other correspondence:

Postal: Editor,
SBP Working Paper Series,
Research Department,
State Bank of Pakistan,
I.I. Chundrigar Road, P.O. Box No. 4456,
Karachi 74000. Pakistan.

Email: [email protected]

ISSN 1997-3802 (Print)

ISSN 1997-3810 (Online)

Published by State Bank of Pakistan, Karachi, Pakistan.

Printed at the SBP BSC (Bank) – Printing Press, Karachi, Pakistan.
 Assessing the Role of Money versus Interest Rate in Pakistan
Zafar Hayat and Muhammad Nadim Hanif

Research Department, State Bank of Pakistan

Abstract

We empirically examine the role of monetary aggregate(s) vis-à-vis short term interest rate as
monetary policy instruments, and the impact of State Bank of Pakistan’s transformation towards the
latter on their relative effectiveness in terms of inflation in Pakistan. Using indicators of ‘persistent
changes’ in the underlying behaviors of variables of interest, we found that broad money consistently
explains inflation in (i) monetary, (ii) transitory and (iii) interest rate regimes. Though its role has
receded whilst moving from the transition to the interest rate regime, the interest rate instrument
seems to be positively related to inflation, a phenomenon commonly known as price puzzle. In light
of these findings, we recommend that the role of money should not be completely de-emphasized
while moving towards flexible inflation targeting regime as planned.

Keywords: Monetary policy instruments, price puzzle, ARDL, Pakistan

JEL Classification: E31, E52

Acknowledgements

The authors are thankful to Ali Choudhary, Muhammad Rehman, Muhammad Jahanzeb Malik,
Fayyaz Hussain, Javaid Iqbal and two referees for their useful comments, discussions and suggestions
on earlier draft of this study.

Contact for correspondence:

Zafar Hayat
Joint Director, Research Department
State Bank of Pakistan
I.I. Chundrigar Road
Karachi 74000, Pakistan.
Email: [email protected]

-1-
Non-technical Summary

Pakistan’s monetary policy framework has evolved over time. Like other countries, the focus is tilting
towards price stability, which unequivocally is deemed essential for the achievement of a sustainable
economic growth. Along with the tilt in preference, as is evident from State Bank of Pakistan’s
Strategic Plan (for 2016-2020) to adopt flexible inflation targeting by 2020, several encouraging
developments at institutional level have occurred. One of interesting developments is the shift in focus
from quantity of money as an instrument of monetary policy to interest rate. Whether this shift of the
SBP focus from monetary aggregates towards the use of interest rate as an instrument of monetary
policy has been effective (to tame inflation); and should the former be completely deemphasized vis-
à-vis the latter are the major policy relevant questions, which this research has endeavored to
empirically explore, in a closed economy setting. The main findings of this research are as follows:

1. Broad money significantly and consistently explains inflation throughout monetary, transition
and interest rate regimes.
2. With the increased focus on interest rate, the role of monetary aggregates has receded. Yet
controlling money supply still has the capacity to effectively counter inflation.
3. Interest rate, on the other hand, has seemingly positive relationship with inflation, a
phenomenon commonly known as ‘price puzzle’ in the literature. There is a need to explore it
further. When taken in isolation, the positive magnitude of the interest rate has receded while
moving from transition regime to the interest rate regime.

In light of above findings the role of monetary aggregates should not be completely deemphasized in
the wake of the intended flexible inflation targeting framework.

-2-
1. Introduction

Although the debate about the choice of an appropriate monetary policy instrument is well known, it
is far from being settled. The main instruments over which the contention has remained are the
price—interest rate—and quantity of money.1 While there is a consensus that both cannot be used
simultaneously to influence the target variables (Turnosky, 1975), studies concluded differently on
their relative effectiveness as monetary policy instruments. For example, Sargeant and Wallace (1975)
argued that reserve money is a better instrument as compared to the interest rate because the latter
suffers from the problem of equilibrium indeterminacy. Similarly, Bhattacharya and Singh (2007)
found that money maximizes welfare in the presence of real shocks. 2 Gordon (1979) on the other
hand concluded in favor of the superiority of the interest rate over monetary instrument for Canada.
Similarly Atkeson et al. (2007) found that interest rate has a natural advantage over quantity of money
as monetary policy instrument.

Research has also indicated that a combination policy (a certain mix between interest rates and
money), as given in Poole (1970) instead may be a better option. He chalked out a theoretical
framework for a combination policy; however his static unified framework only allows answering the
underlying question of the relative effectiveness of monetary policy instruments in terms of output;
rather than the inflation, taming which is the prime objective of most of the central banks today.3

Monetary policy practices at the State Bank of Pakistan (SBP), the country’s central bank have varied
over time (see Hanif, 2014 for details). Historically money played an important role as a monetary
policy instrument. The focus nevertheless has now shifted towards interest rate and currently an
interest rate corridor system is in place effective from August, 2009 (Hussain, 2009; Khan, 2010 and
Hanif, 2014). Whether this transition from focus on quantity of money towards the use of interest rate
as an instrument of monetary policy has been effective; and should the former be completely
deemphasized vis-à-vis the latter are crucial questions yet to find research-based answers. It is also
important to ascertain if the transformation of Pakistan’s monetary policy’s focus from targeting
monetary aggregates towards the active use of short term interest rate has any bearing on their relative
effectiveness in terms of inflation in the country.

To the best of our knowledge, this is an unexplored research area. In a related attempt, Ali and Ahmad
(2014) explored the relative performance of inflation, and price level targeting regimes under
alternative monetary policy instruments, and found money as a better performer relative to interest
rate for Pakistan. Their analysis, however, is based on calibrating their model while using parameters
from Din and Khan (2011), which used annual data from 1972–2009. Neither their focus was nor
could their study (by construct) observe the evolution of the relative role of money and interest rate—
especially in the context of SBP’s transition from the former to the latter—which was completed by
2009. It therefore did not take into account the full-fledged interest rate regime. Most recently Ahmad
et al. (2016) theoretically evaluated the role of money in propagating business cycle fluctuations in
Pakistan and found that cash based economy models under money growth rule perform well as
compared to the cashless economy models with Taylor type rule.

1
Atkeson et al. (2007) accounts for exchange rate as one of the instruments in addition to interest rates and
money.
2
It may be noted that they also found that in the presence of nominal and relatively small shocks instead of
money, interest rate instrument is better.
3
Woglom (1979) and Benaive and Richard (1983) are among others who have worked on similar lines.

-3-
In contrast to the approaches used in above cited literature, we use; i) the framework used by Hayat et
al. (2016) to extract indicators of persistent changes in the variables of interest to be able to closely
observe the most relevant underlying relationships and, ii) apply ARDL approach for estimation of
such relationships.

The results indicate that money remains a consistent performer vis-à-vis interest rate but its role has
been receding with an increased focus of SBP on interest rate as a monetary policy instrument.
Nevertheless, there is an evidence of a positive relationship between interest rate and inflation, which
is indicative of the possibility of a price puzzle. It needs more research. It may be advisable for the
SBP not to completely de-emphasize the use of money as an instrument of monetary policy (vis-à-vis
interest rate) as it has been significantly effective in taming inflation in the country.

We organize our paper as follows. Section 2 lays down methodological framework for generation of
indicators of persistent changes in variables of interest. Section 3 discusses testing and estimation
strategy, specifies the model, and highlights data. Section 4 brings forth the results and discussion
while Section 5 examines the soundness of the generated indicators for analysis and a sort of
robustness of the results found in this study. Section 6 concludes the paper.

2. Methodological Framework

In order to examine the relative importance of money versus interest rate as monetary policy tools as
well as their evolution over time, following Hayat et al (2016), we first generate indicators
representing persistent variations in variables of interest, and then use them to estimate their long-term
relationships through ARDL approach. Generating indicators of persistent variation is important
because only a fraction of variations in monetary policy instruments may tend to relate to a fraction of
variations in target variables (such as inflation and/or real economic growth) given that the central
bank may not necessarily exercise full control over the variations in monetary policy variables
especially broad money. This postulation, as a starting point is consistent with Bullard (1999); Uhlig
(2005) and Hayat et al. (2016). We derive indicators of persistent variations from a) growth in broad
money, b) market interest rate, c) inflation, and d) real GDP growth rate in two steps as follows.

In the first step the Hodrick and Prescott (1997) filter—henceforth HP filter—is applied to decompose
the observed series ( ), which we call base variable in this study, into its permanent (long term) path
( ) and the transitory fluctuations ( ). The are obtained by subtracting the long-term path from
the observed time series such that . In the second step, we apply HP filter to to
extract its permanent part ( ), which we consider as an indicator of persistent variations in . One
may ask about the justification of use of the HP filter. First, our choice of this filter is driven by the
fact that HP filter allows the trend to vary over time and hence the magnitudes of deviations, which
may better represent policy responses (variations) in the underlying policy as well as goal variables
(Hayat, et al. 2016). Second, double HP filter outperforms other detrending and smoothing methods in
turning point signal stability i.e. identifying turning points quickly (Nilsson and Gyomai, 2011),
which are reflective of structural changes and hence regimes. This feature is important because our
purpose is to observe the evolution of money and interest rate instruments across different regimes: (i)
monetary, (ii) transitory and (iii) interest rate regimes.

For all the four (base) variables—broad money growth, interest rate, inflation and real GDP growth
rate—we therefore apply the two step procedure to obtain our desired indicators of persistent
variations as follows.

-4-
2.1 Indicator of Persistent Variations in Broad Money Growth

In the first step the HP filter is applied to decompose the observed series of growth in (denoted by
) over time into its long-term growth path and the fluctuations around it , such that:

for

In the second step, the HP filter is applied to to obtain its long-term trend path, which
corresponds to persistent variations, denoted by such that:

for

for

Where is the desired series representing persistent variations in broad money growth.

2.2 Indicator of Persistent Variations in Interest Rate

Similarly, the HP filter is applied to decompose the interest rate ( ) over time into its long-term path
and the fluctuations around it. In the first step:

for

In the second the HP filter is applied again to to obtain its long-term path of our interest as
follows:

for

for

2.3 Indicator of Persistent Variations in Inflation

The two-step strategy of application of HP filter is also employed to generate indicators of inflation,
and real GDP growth as follows:

for

where is inflation rate in time . The is its long-term path in time and represents the
fluctuations around over time. In the first step, the HP filter is applied to to obtain and .
In the second step, the HP filter is applied to to obtain its long-term path such that:

for

for

Where, is the desired inflation indicator.

2.4 Indicator of Persistent Variations in Real GDP Growth Rate

Likewise, the strategy of the application of the two-step HP filter is used to obtain the real GDP
growth indicator. Firstly, the time series of the growth in real GDP ( ) is decomposed into its long-
term growth path and the fluctuations around it i.e. such that:

-5-
 for

Secondly, the HP filter is applied to to obtain its long-term path as:

for

for

Where, is the desired real growth indicator.

3. Estimation Approach, Model Specification and Data

We use auto regressive distributed lag (ARDL) bounds testing and estimation approach to
cointegration proposed by Pesaran and Shin (1999) and Pesaran et al. (2001) to obtain long-run
parameter estimates. The estimators of the ARDL are super-consistent for long-run coefficients and it
performs particularly well in small samples without losing long-run information. The ARDL approach
allows the selection of optimal dynamic model. Since, Pesaran and Pesaran (1997) and Pesaran and
Shin (1999) reported that Schwartz Bayesian Criterion (SBC) is a consistent model selection criterion
in small samples and that it selects a relatively more parsimonious model (Enders, 1995), we use
SBC. The ARDL works even in the presence of endogenous regressors irrespective of the order of
integration (1 or 0) of explanatory variables (Pesaran and Pesaran, 1997; Pesaran and Shin 1999).

Operationally, the ARDL is a two-stage procedure. The first stage is to test for the existence of
cointegration by computing the F-statistic. Since the asymptotic distribution of this F-statistic is non-
standard, Pesaran et al. (2001) tabulated two sets of appropriate critical values both for I(0) and I(1),
for different numbers of regressors ( ) with and without intercept and trend. If the computed F-
statistic falls outside the band for respective critical values of I(0) and I(1), cointegration exists. If it
falls within that band then the result of the inference is inconclusive. In the second stage, long-run
coefficients are obtained, provided the cointegration is established in the first stage. In general form,
the error correction version of our ARDL model may be given as:

(1)

Where , , , and are inflation, broad money growth, interest rate and real GDP growth rate
indicators, respectively. The denotes the first difference operator and is the error term.

We use quarterly data series from Q1-1974 to Q2-2015. We also analyzed the regime wise
subsamples: monetary (1974-1995), transitory (1995-2009) and interest rate regimes (2009-2015)4.
The main variables are expressed in terms of a year on year (YoY) change in CPI, real GDP and
broad money—which allows us to control possible seasonality—while the call money rate (as a proxy
of policy rate) is in levels. Data for broad money growth and call money rate is taken from the SBP.
Since the policy rate remained constant from 1977 to 1990 and also for a short span in early 2000s

4
More specifically the respective data spans are as follows: Q1-1974 to Q1-1995, Q1-1995 to Q1-2009, and Q1-
2009 to Q2-2015.

-6-
(Figure 1), we use call money rate as an alternative. The call money rate closely mimics the behavior
of the policy rate. The correlation coefficient between policy rate and call money rate is 0.95.

Figure 1: Call Money Rate (MIR) and Discount Rate (DR )

MIR DR (RHS)
16 21

13 17

10 13

6 8

3 4

0 0
Feb-74

Feb-81

Feb-09
Feb-88

Feb-95

Feb-02
Aug-77

Aug-84

Aug-12
Aug-91

Aug-98

Aug-05
Apr-82

Apr-89

Apr-96
Apr-75

Apr-03

Apr-10
Jun-97

Jun-04

Jun-11
Jun-76

Jun-83

Jun-90
Dec-72

Dec-93

Dec-00

Dec-07
Dec-79

Dec-86

Dec-14
Oct-85

Oct-92
Oct-78

Oct-99

Oct-06

Oct-13
Inflation data is obtained from the country’s national statistical agency, the Pakistan Bureau of
Statistics (PBS). National income accounts are compiled by the PBS only on annual basis; we
therefore use the quarterly GDP data for Pakistan for the fiscal years 1973–2012 estimated by Hanif et
al. (2013). Since they quarterized the data only until 2012, we extended their data set up to 2015 while
using the proportions therein for 2012, on the basis of the latest available annual data from the PBS
for the period 2013–2015. It may be noted that for the entire series to be consistent, we transformed
their series from 1974 to 2012 on the new base year i.e. 2005–2006.

In order to be confident whether our generated ‘indicators’ represent the behaviors in the respective
base variables, and to make sure that it might not have led to a considerable loss of information, we
checked their correlations with their respective base variables and cycle series. Figure 2 shows that
by and large, double filtering has not resulted in loss of significant information as exhibited by
correlation coefficients with respective base and cycle variables, especially in case of inflation and
interest rate indicators. Instead it seems that rather the noise part has been purged, which may not
necessarily be representing policy induced actions and responses in variables (see Hayat et al. 2016). 5
For example, the correlation between the base variable CPI inflation and its cycle series (CPIT) is
0.71 and that of CPI inflation and our generated indicator is 0.63 (see Figure 2).

Figure 2: Correlation of the Base Variable with Respective Cycle Series and Generated Indicators
CPI GM2 GDP MIR
1.0

0.8

0.6

0.4

0.2

0.0
MIRTP
GM2TP
CPITP

MIRT
GDPTP
CPIT

GM2T

GDPT

5
A true test to this effect nevertheless would be for the indicators to yield intuitive results as against the base and
cycle series (see Section 4 and 5).

-7-
4. Results

4.1 Model Selection, Diagnostics, and Cointegration Tests

Given the lack of theoretical guidance as to what should be the appropriate maximum lag length in a
particular situation, we relied on a general-to-specific approach for selection of optimal lag lengths.
We started with 10 quarters as the maximum lag length in case of full sample and used the SBC
model selection criterion and selected a model with no issues like those pertaining to serial
correlation, heteroskedasticity, and estimated coefficients’ stability. This allowed lags for up to two
and half years. It is a reasonable transmission time for the effects of monetary policy instruments at
least in the case of Pakistan (considering past research in this area like Khan, 2008). In case of
subsamples as well, any maximum allowable lag length (lower than 10) was tried during the selection
process. We used the SBC model selection criterion as it selects the most parsimonious model.

For the model in equation (1), the null hypotheses of non-existence of a long-run relationship is given
by against the alternative . The F–statistics
computed for the joint significance of , and for the full (1974–2015) and subsamples (1974–
1995, 1995–2009 and 2009–2015) are 10.99, 17.51, 9.35 and 162.67, respectively. All these
computed statistics exceed the corresponding critical value bands of Pesaran et al. (2001) for
unrestricted intercept and no trend at 1% level, leading us to reject the null of non-existence of a long-
run relationship. This implies that the decision to proceed with computing long-run coefficients is
conclusive and there is no need to know the cointegration rank (Pesaran and Pesaran, 1997). The
long-run parameter estimates are obtained subsequently.

Although we are mindful of the possibility of endogeneity, we have confidence in our estimates as
the ARDL methodology works well even in the presence of endogenous regressors and irrespective of
the order of integration [I(1) or I(0)] of explanatory variables (Pesaran and Pesaran, 1997; Pesaran and
Shin 1999)6. Alternative methodologies to minimize the extent of possible endogeneity are 2SLS and
GMM, which however require identification of ‘instrumental variables’. Generally, it is hard to find
‘instrumental variables’ for the variables in the equation to be estimated. In such cases, suggestion in
the empirical literature is the use of lagged variables. In this study, the way we have developed each
indicator, it in itself is like an instrument for the underlying variable. And that the use of the lagged
values of these indicators in ARDL modeling reduces the chances of endogeneity in our estimation.

4.2 Money versus Interest Rate Effectiveness—Regime-Wise Results

Since our objective is to assess if SBP should exert an increased emphasis on interest rate compared to
the broad money and whether the transformation in the focus of Pakistan’s monetary policy from the
latter to the former has had any bearing on their relative effectiveness in terms of inflation, we
conduct analysis of sub-periods that correspond to i) monetary targeting regime (1974–1995), ii)
transition period from monetary targeting to interest rates as monetary policy instruments (1995–
2009), and iii) an interest rate regime (2009–2015).7 As a cross check, we used Bai Perron’s (1998,
2003) test for identification of multiple breakpoints and found supporting evidence that breaks
occurred in 1995 and 2009 (Appendix).

6
We applied ADF test and found that none of the variables used in this study for ARDL model estimation is
integrated of order higher than 1.
7
See Hanif (2014) for a discussion on key developments in these phases.

-8-
The results do not reveal a clear picture when the estimations are carried out for the full sample from
1974 to 2015 as neither money nor interest rate has a significant role in explaining inflation in
Pakistan (Table 1, column (a)). This may be due to the fact that during the entire sample period,
SBP’s monetary policy preferences in terms of use of instruments have varied, which has obscured
the results for the overall sample. This, however, is not the case when we subsequently observe the
results for the specific regimes.

Table 1: Long-Run Estimates

Full Sample Monetary Regime Transition Period Interest Rate Regime

Variables 1974-2015 1974-1995 1995-2009 2009-2015
(a) (b) (c) (d)
0.39 0.16 1.06 0.61
[0.12] [0.67] [0.00]*** [0.00]***
0.08 0.95 1.26 0.6
[0.57] [0.02]** [0.00]*** [0.00]***
-0.81 -5.93 -1.37 -1.7
[0.09]* [0.04]** [0.00]*** [0.00]***
0.01 0.14 0.05 -0.07
[0.38] [0.01] [0.05] [0.00]
-0.003 -0.004 -0.02 -0.35
[0.00]*** [0.00]*** [0.00]*** [0.00]***
ARDL (8,3,6,4) (8,1,5,8) (5,2,5,0) (2,0,2,2)
COIN 1% 1% 1% 1%
DW 1.86 2.1 2.03 2.18
R2 0.99 0.99 0.99 0.99
This table reports the long-run coefficients and the P-values. The latter are reported in square brackets. ARDL shows the
order of the lags of the selected models whereas COIN stands for cointegration. ***, ** and * indicate significance at
1%, 5% and 10% level, respectively.

Supply side effects of real growth seem rather visible, which tend to reduce inflation. This result is
consistent with Hayat et al. (2016), which found an inverse relationship between real growth and
inflation indicators using annual data from 1961 to 2010. Although the question may remain that
whether the real activity indicator used is proxying supply or demand side of the economy. We
advocate the former because real growth in GDP is used rather than nominal. Furthermore as is visible
from column (b) through (d) in Table 1, the magnitude of the effect of the real growth on inflation
decreases, which make sense only when real growth represent an increased supply of goods and
services—as the average real growth witnessed in the sample period used in columns (b) (c) and (d)
are 5.60, 4.47 and 3.41, respectively.

During monetary targeting regime, the role of money in explaining inflation is both significant and
quantitatively large as against interest rate (Table 1, column b). This result is consistent with a range
of studies that has found brad money an important determinant of inflation in Pakistan such as
Chaudhary and Ahmad (1996), Price and Nasim (1999), Khan and Schimmelpfennig (2006), Serfraz
and Anwar (2009), Hayat et al. (2016), and Ahmed et al (2016). A straight forward policy implication
for SBP from these results is that money plays a significant role in explaining inflation and therefore it
may be used as an effective monetary policy instrument to tame it. It is however interesting to note
that interest rate whilst being insignificant during the monetary targeting regime, grew in significance
during the transition period to the interest rate regime (Table 1, column (c)).

-9-
Under the interest rate regime, both money and interest rate played a significant role in explaining
inflation; however the quantitative effect of the latter is more pronounced in this regime as compared
to the transition regime. On the other hand, the quantitative effect of broad money receded vis-à-vis
the interest rate instrument during the transition period. These results imply that the shift in focus
from monetary aggregates towards interest rate as monetary policy instrument has had implications
both for the relative importance and significance of the two monetary policy instruments.

When taken in isolation, although the interest rate instrument grew in significance during transition
and interest rate regimes, it may not effectively guide the monetary policy as it brings forth an
important monetary policy issue for the SBP, commonly known in the literature as ‘price puzzle’,
wherein interest rate and inflation are positively related. Considering price puzzle issue we suggest
that money should not be completely deemphasized. The phenomenon of price puzzle is well
documented in related literature on Pakistan (for example Khan, 2008) as well as some other countries
(see for example Sims, 1992).8 Although requiring further in-depth research, in case of Pakistan, this
phenomenon can possibly be explained by being gradual and sometimes accommodative monetary
policy actions. For example, negative real (market) interest rates for most of the period under this
study, with few exceptions, especially the period since September, 2011, where the real interest rate
remained positive.

5. Soundness of Indicators and Robustness Check

To check whether our generated double filtered indicators has given intuitively consistent
approximations of the underlying phenomena; we did the estimations using both the base data and
cycle series. The results obtained using base data (Table 2)—although not as intuitive as the results
obtained from our double filtered indicators—by and large provide support to our main results as
compared to the results obtained using cycle series (Table 3).

Table 2: Long-Run Estimates Using Base Data

Full Sample Monetary Regime Transition Period Interest Rate Regime
Variables 1974-2015 1974-1995 1995-2009 2009-2015
(a) (b) (c) (d)
0.62 0.33 1.53 1.41
[0.01]** [0.34] [0.00]*** [0.00]***
0.44 0.26 0.2 0.24
[0.01]** [0.12] [0.51] [0.46]
0.01 -1.07 1.79 -0.39
[0.98] [0.02]** [0.15] [0.54]
-3.71 8.04 -16.49 -7.14
[0.25] [0.08] [0.02] [0.23]
-0.22 -0.29 -0. 29 -0.67
[0.00]*** [0.00]*** [0.11] [0.05]*
ARDL (7,0,7,1) (8,0,3,3) (7,5,8,5) (2,0,0,0)
COIN 1% 2.50% Nil Nil
DW 1.99 2.01 1.98 2.26
R2 0.9 0.86 0.98 0.83
This table reports the long-run coefficients and the P-values. The latter are reported in square brackets. Nil means no
cointegration. ARDL shows the order of the lags of the selected models whereas COIN stands for cointegration. ***, **
and * indicate significance at 1%, 5% and 10% level, respectively.

8
Javid and Munir (2010) also found similar results. Felipe (2009) and Naqvi and Rizvi (2010) are yet other
studies that pointed towards this issue while examining the suitability of adoption of inflation targeting for
Pakistan.

-10-
 Table 3: Long-run Estimates Using Cycle Data
Full Sample Monetary Regime Transition Period Interest Rate Regime
Variables 1974-2015 1974-1995 1995-2009 2009-2015
(a) (b) (c) (d)
-0.31 0.09 0.33 -0.68
[0.22] [0.82] [0.32] [0.49]
-0.15 -0.20 0.23 -0.29
[0.18] [0.18] [0.35] [0.46]
-0.01 -0.57 0.22 1.85
[0.93] [0.08]* [0.64] [0.18]
-0.05 -0.10 0.03 -0.34
[0.81] [0.83] [0.95] [0.52]
-0.46 -0.34 -0. 28 -0.55
[0.00]*** [0.00]*** [0.04]** [0.01]**
ARDL (8,3,6,1) (5,0,3,0) (5,0,0,1) (2,1,0,1)
COIN 0.01 Nil Nil Nil
DW 1.98 1.95 1.70 2.30
R2 0.84 0.80 0.88 0.67
This table reports the long-run coefficients and the P-values. The latter are reported in square brackets. Nil means non
existence of a cointegrating relationship. ARDL shows the order of the lags of the selected models whereas COIN stands
for cointegration. ***, ** and * indicate significance at 1%, 5% and 10% level, respectively.

All the mainstream models obtained using double filtered indicators not only fit the data well but also
approximate cointegrating relationships as compared to the models that instead use the variables in
base and cycle form. The results obtained using cycle series are highly inconsistent, largely depict
incorrect signs and do not pick the breaks, thus failing to track the evolution in the relative role of
money versus interest rates. On the contrary the results obtained using the model with generated
indicators better identifies the breaks in a cointegrating relationship (see Appendix).

As far as the robustness is concerned, our results are largely robust to alternative specifications for
almost all the sample sizes. First we dropped the real growth indicator and estimated the model for all
the regimes. Second we ran the regressions for interest rate and broad money indicators both
individually and in combination which led us to conclude in favor of our main findings.9 Since we
also found evidence of structural break in 1982–83 (Appendix), we controlled for it through dummy
variable and re-estimated our models (a) and (b) in Table 1 just in case the results turn out to be
different than without controlling for the structural break. Our inference from the new results obtained
however remained unaltered.10

6. Conclusion

Pakistan’s monetary policy has evolved over time. The evolution of the relative role of money and
interest rate is examined across three phases of monetary policy experience in Pakistan i.e. regime of
targeting monetary aggregates, period of transition towards interest rate and interest rate regime. A
framework has been chalked out that allowed generation of indicators capturing persistent variations

9
For brevity purposes, these results however are not reported and may be obtained from the corresponding
author if needed.
10
These results are not reported and may be obtained from corresponding author upon request.

-11-
in underlying variables. Broad money is found to consistently perform vis-à-vis interest rate
throughout the entire spectrum in controlling inflation in the country. Its quantum effect, however,
started receding during the transition period and almost equalized the interest rate instrument during
the interest rate regime. The role of interest rate is found to be puzzling as it is positively and
significantly related to the inflation. The use of the interest rate by SBP therefore may not be as
effective as desired unless this puzzle is explored and addressed. Since broad money is still effective,
its role should not completely be deemphasized.

-12-
References

Ahmad, S., F. Pasha, and M. Rehman, (2016), “The Role of Money in Explaining Business Cycles for
a Developing Economy: The Case of Pakistan”, SBP Working Paper Series No. 74.
Ali, H., and E. Ahmad, (2014), “Choice of Monetary Policy Instrument under Targeting Regimes in a
Simple Stochastic Macro Model”, PIDE Working Paper No. 102.

Atkeson, A., V.V. Chari, and P.J. Kehoe, (2007),“On the Optimal Choice of a Monetary Policy
Instrument”, NBER Working Paper Series No. 13398.

Bai, J., and P. Perron, (1998), “Estimating and Testing Linear Models with Multiple Structural
Changes”, Econometrica, 66 (1), pp. 47–78.

(2003), “Computation and Analysis of Multiple Structural Change Models”,

Journal of Econometrica, 18, pp. 1–22.

Benavie, A., and T.F. Richard, (1983), “Combination Monetary Policies to Stabilize Price and Output
under Rational Expectations”, Journal of Money, Credit, and Banking, pp. 186–98.

Bhattacharya, J., and R. Singh, (2007),“Optimal Choice of Monetary Instruments in an Economy with
Real and Liquidity Shocks”, Iowa State University.

Bullard, J. (1999), “Testing the Long-run Neutrality of Money: Lessons from the Recent Research”,
Federal Reserve Bank of St. Louis Review, 81, pp. 57–77.

Chaudhary, M.A., and N. Ahmad, (1996), “Sources and Impacts of Inflation in Pakistan”,
Pakistan Economic and Social Review, 34(1), pp. 21–39.
Din, M. U., and M.A. Khan, (2011),“A Dynamic Macroeconometric Model of Pakistan’s Economy”,
PIDE Working Papers No.69.

Enders, W. (1995), “Applied Econometric Time Series”, John Wiley and Sons, USA.

Felipe, J. (2009), “Does Pakistan Need to Adopt Inflation Targeting? Some Questions”, SBP Research
Bulletin, 5(1), pp. 113-161.

Gordon, S. R. (1979), “The Choice of Monetary Policy Instruments in Canada”, The Canadian Journal
of Economics, 12, pp. 15–24.

Hanif, M.N. (2014), “Monetary Policy Experience of Pakistan”, MPRA Paper No. 60855.

Hanif, M.N., J. Iqbal, and M.J. Malik, (2013), “Quarterisation of National Income Accounts of
Pakistan”, SBP Research Bulletin, 9(1), pp. 1–61.

Hayat, Z., F.Balli, J. Obben, and S. Shakur, (2016), “An Empirical Assessment of Monetary
Discretion: The Case of Pakistan”, Journal of Policy Modeling, 38, pp. 954–970.

Hodrick, R. J., and E.C. Prescott, (1997), “Postwar U.S. Business Cycles: An Empirical
Investigation”, Journal of Money Credit and Baking, 29, pp. 1-16.

Hussain, F. (2009), “Framework for SBP’s Operations in the Money Market: Current Practices and
Recommendations”, Internally Circulated Policy Note of Monetary Policy Department, State Bank of
Pakistan.

Javaid, M., and K. Munir, (2010), “The Price Puzzle and Monetary Policy Transmission Mechanism
in Pakistan: Structural Vector Autoregressive Approach”, The Pakistan Development Review, 49(4),
pp. 449–460.

-13-
Khan, H. (2010), “Overnight Repo Rate Instability: Current Practice and solutions”, Internally
Circulated Policy Note of Monetary Policy Department, State Bank of Pakistan.

Khan, M. H. (2008), “Short Run Effects of an Unanticipated Change in Monetary Policy: Interpreting
Macroeconomic Dynamics in Pakistan”, SBP Working Paper No. 22.

Khan, M.S., and A. Schimmelpfennig, (2006), “Inflation in Pakistan”, The Pakistan Development
Review, 45, pp. 185–202.

Naqvi, B., and S.K.A. Rizvi, (2010), “What does Pakistan have to Join Inflation Targeters Club, a
Royal Flush or a Seven-deuce Offsuit?”, MPRA Paper No. 19575.

Nilsson, R., and G. Gyomai, (2011), “Cycle Extraction: a Comparison of the Phase-Average Trend
Method, the Hodrick-Prescott and Christiano-Fitzgerald Filters”, https://www.oecd.org/std/leading-
indicators/41520591.pdf

Pesaran, M. H., and Y. Shin, (1999), “An Auto Regressive Distributed Lag Modelling Approach to
Cointegration Analysis Econometrics and Economic Theory in the 20th Century”, The Ragner Frisch
Centennial Symposium (S.Strom ed.), Cambridge: Cambridge University Press.

Pesaran, M. H., Y. Shin, and R.J. Smith, (2001),“Bounds Testing Approaches to the Analysis of Level
Relationships”, Journal of Applied Econometrics, 16, pp. 289–326.

Pesaran, H. M., and B. Pesaran, (1997), “Working with Microfit 4.0: Interactive Econometric
Analysis”, Oxford: Oxford University Press.

Poole, W. (1970), “Optimal Choice of Monetary Policy Instruments in a Simple Stochastic Macro
Model”, Quarterly Journal of Economics, 84, pp. 197–216.

Price, S., and A. Nasim, (1999), “Modelling Inflation and the Demand for money in Pakistan:
Cointegration and the Causal Structure”, Economic Modeling, 16, pp. 87–103.

Sargent, T.J., and N. Wallace, (1975), “Rational Expectations, the Optimal Monetary Instrument and
the Optimal Money Supply Rule”, Journal of Political Economy, 83, pp. 241–255.

Serfraz, A., and M. Anwar, (2009), “Fiscal Imbalances and Inflation: A Case Study of Pakistan”,
Pakistan Journal of Social Sciences, 29(1), pp. 39–50.

Sims, C.A. (1992), “Interpreting the Macroeconomic Time Series Facts: the Effects of Monetary
Policy”, European Economic Review. 36, pp. 1001–1011.

Turnovsky, S. J. (1975), “Optimal Choice of Monetary Instrument in a Linear Economic Model with
Stochastic Coefficients”, Journal of Money, Credit and Banking, 7, pp. 51-80.

Uhlig, H. (2005), “What are the Effects of Monetary Policy on Output? Results from an Agnostic
Identification Procedure”, Journal of Monetary Economics, 52, pp. 381–419.

Woglom, G. R. H. (1979), “Rational Expectations and Monetary Policy in a Simple Macroeconomic

Model”, Quarterly Journal of Economics, 93, pp. 91–105.

-14-
Appendix

Table 1A Bai-Perron Multiple Break Points Test

Breaks F-Statistic Critical Value Break year(s)

1* 41.8 18.26 1982

2* 55.89 14.45 1983, 1991
3* 56.67 12.16 1983, 1991, 2005
4* 61.99 10.56 1983, 1989, 1995, 2006
5* 56.68 8.71 1983,1989,1995,2002, 2008

* Denotes significance level at 1%. A trimming level of 15% was used and the maximum breaks allowed were 5. The
critical values are that of Bai and Perron (2003).

-15-