a macroeconometric analysis of the effectiveness of

Final Copy of Ph.D. Dissertation.docxPH.D. DISSERTATION
AUGUST 2014
75270 KARACHI, PAKISTAN
POLICY
PH.D. DISSERTATION
AUGUST 2014
Submitted to the Department of Economics, University of Karachi in partial fulfillment of the requirements of the Degree of Doctor of Philosophy in Economics
A MACROECONOMETRIC ANALYSIS OF THE
EFFECTIVENESS OF PAKISTAN’S MONETARY
POLICY
ii
Department of Economics University of Karachi 75270, University Road, Karachi, Pakistan Doctoral Dissertation in Economics
© 2014 Rubina Hassan
ALL RIGHTS RESERVED
iii
DECLARATION
I Rubina Hassan, daughter of Mohammad Hassan hereby declare that no part of the
work presented by me for this dissertation titled “A MACROECONOMETRIC
ANALYSIS OF THE EFFECTIVENESS OF PAKISTAN’S MONETARY POLICY”
and submitted to the Department of Economics, University of Karachi has been
plagiarized from anywhere. All sources have been duly acknowledged as and where
necessary. I further declare that this work has not been presented anywhere else, either
whole or in part, for the award of any other degree1.
Rubina Hassan Assistant Professor Department of Economics University of Karachi Karachi, Pakistan
1 Excerpts from the dissertation appear in the form of refereed published articles and in the form of
working papers at the MPRA, IDEAS and SSRN Repositories over the internet. Appropriate references to such works are also cited wherever necessary.
iv
CERTIFICATE
This is to certify that we have examined the work contained in the dissertation titled “A
MACROECONOMETRIC ANALYSIS OF THE EFFECTIVENESS OF PAKISTAN’S MONETARY
POLICY” submitted by MS. RUBINA HASSAN, Assistant Professor, Department of
Economics, University of Karachi, as conforming to the required standards for the
fulfillment of the degree of Doctor of Philosophy in Economics and as her original
contribution to the subject.
Supervisor DR. ABDUL WAHEED
University of Karachi
v
vi
ABSTRACT
This dissertation deals with investigation and evaluation of the design and conduct of Pakistan’s monetary policy by identifying, constructing, estimating and simulating some marginalized macroeconometric models for Pakistan’s monetary sector over the period 1974-75 to 2008-09. We seek to establish the quantitative strength of the various channels of transmission of changes in monetary instruments to the money market and to identify Pakistan’s monetary policy and its operationalization strategy with a view to analyze the effectiveness of Pakistan’s monetary policy. The first marginalized macroeconometric model of Pakistan’s monetary sector delineates the transmission mechanism of monetary policy by taking into consideration all structural money demand and money supply linkages along with the historically implied identifying assumption that the State Bank of Pakistan follows a Taylor-type policy rule. The model determines the interest rate as an equilibrium process akin to the way it is determined in the Loanable Funds Model. While the model performs significantly well both within and outside of the estimation sample, active denial by the State Bank of Pakistan regarding following a Taylor-type policy rule motivates us to identify Pakistan’s monetary policy in an alternative manner. To facilitate an alternative identification of monetary policy, we first make use of State Bank of Pakistan’s balance sheet data to construct Reserve Money and its various components in accordance with the Reserve Balance Equation. Empirical determination of the same then leads to the second and the third marginalized macroeconometric models of Pakistan’s monetary sector (the Lombra-Kaufman Model and the Free Reserve Targeting Model). While these models share the same money demand dynamics as in the Monetary Sector Model, they make use of the Reserve Balance Equation for determining money supply. These models not only identify Pakistan’s monetary policy operationalization strategy, but also ascertain the various monetary policy regimes followed by State Bank of Pakistan under alternative governors. The Lombra-Kaufman Model and the Free Reserve Targeting Model further show that the Taylor-type policy rule equation correctly identifies Pakistan’s monetary policy across various operationalization regimes. Among the various conclusions that the Monetary Sector Model, the Reserve Balance Equation, the Lombra-Kaufman Model and the Free Reserve Targeting Model facilitate, the most noteworthy are: (1) The Monetary Sector Model establishes a precise estimate of the interest elasticity of money supply. It predicts that a 100 basis points increase in
vii
the market rate of interest increases the supply of money by 2.265 percent (2.884 percent if the historically implied policy rule equation is known to all) in the impact period and by 3.14 percent in full equilibrium. (2) The Monetary Sector Model, the Lombra- Kaufman Model and the Free Reserve Targeting Model predict, using both the econometric and the narrative identification approaches, that State Bank of Pakistan’s policy behavior is mimicked by a Taylor-type policy rule equation. The atypical coefficients of output and inflation in this policy rule equation not only reflect the confusion that surrounds the design of monetary policy at the State Bank of Pakistan, they also reflect that State Bank of Pakistan accommodates inflation rather than combating it. (3) The Reserve Balance Equation shows that the data on Currency in Circulation, Reserve Money and Money Supply is incorrect. This inevitably leads to the conclusion that the discount rate is also incorrect, though by an infra-marginal unit. (4) The empirically determined Reserve Balance Equation for Pakistan leads to a number of monetary policy indicators for studying Pakistan’s monetary policy. These include the unborrowed, free and drainable reserve ratios; simple and reserve multipliers and their domestic and foreign components; endogenous and exogenous components of broad money supply; the exogenous component of reserve money and its growth rate; the ratio of broad money supply to unborrowed reserves; the difference between the unborrowed and drainable reserve ratios; and the liquidity definitions of budget deficit and balance of payments. (5) The Reserve Balance Equation, the Lombra-Kaufman Model and the Free Reserve Targeting Model corroborate that Pakistan’s monetary policy revolved around targeting unborrowed reserves before 1999, and around targeting free reserves later. Specifically, unborrowed reserves were targeted at 18.57%, 26.33% and 18.66% of money supply during 1981-86, 1987-93 and 1994-99 respectively, while free reserves were targeted at 3.8% during 2000-05 and at 4% and 2.4% during 2006-07 and 2008-09. Apart from the empirical findings relating to Pakistan’s monetary policy, the dissertation also makes an important contribution in developing an objective methodology for monetary policy identification by means of the narrative approach. We believe that this is an important advancement over the methodology of the Romers’. The dissertation also outlines a broad spectrum of empirical research for further analysis of Pakistan’s monetary policy.
ABBSTRACT (TRAN
xii
ACKNOWLEDGEMENTS
First and foremost, I am thankful to Almighty Allah who enabled me to accomplish the task of completing my Ph.D. through submission of this dissertation. During the course of writing this dissertation I have become indebted to a number of people worth mentioning. First of all, I am thankful to my supervisor, PROFESSOR DR. ABDUL WAHEED, Chairman, Department of Economics, University of Karachi for his cooperation and support and guidance towards the completion of this work. Thanks are also due to PROFESSOR DR. M. NISHAT, Associate Dean, Institute of Business Administration under whose supervision I commenced work on my dissertation. I am also grateful to PROFESSOR DR. ABID AZHAR, Co-Director, A. Q. Khan Institute of Biotechnology and Genetic Engineering, University of Karachi for his help, co-operation and effort towards realization of this work in the form of a dissertation. Thanks are also due to the Publications Department, FEDERAL RESERVE BANK (BOG) for sending an important piece of out-of-print reference at a very early stage of this work. I am also indebted to DR. ATHER ELAHI, a former student and now a faculty member at the Institute of Business Administration for his help extended through provision of some important data without which this work could not have got underway. The encouragement and assistance provided by MR. MIRZA M. SHAHZAD, Assistant Professor, Department of Economics, University of Karachi towards the task of completion of my dissertation is also of great value to me. I also wish to express my gratefulness towards the DEPARTMENT OF ECONOMICS, UNIVERSITY OF KARACHI which facilitated the completion of this work. Lastly I wish to express my deep and sincere thanks to my family and especially to my daughter. Her continued cooperation and moral support in all respects during all these years, while granting all time which was rightly hers, provided the freedom to accomplish this task.
Rubina Hassan March, 2014
Chapter 1 INTRODUCTION . . . . . . . . . . . . . . . . 2
1.1.1 The Quest for Econometric Modeling of Monetary Policy 2
1.1.2 An Overview of Pakistan’s Monetary Policy Framework 3
1.1.2.1 Monetary Policy Framework: 1974-75 – 2008-09 4
1.1.2.2 The Targets, Indicators and Instruments of Monetary Policy 7
1.1.2.3 The Operationalization Strategy of Monetary Policy 8
1.1.2.4 SBP’s Conception of Money Supply 9
1.2 Purpose and Objectives 10
1.3 An Overview of the Thesis 13
xiv
2.1 The Identification of Monetary Policy 18
2.2 The Econometric Modeling of Monetary Policy 25
2.3 The Study of Monetary Policy in Pakistan 29
2.4 Concluding Remarks 37
MONETARY POLICY . . . . . . . . . . . . . . . . 39
3.1.1 Forecasting the Monetary Survey and Monetary Statistics 40
3.1.2 Model Specification 41
3.2 Empirical Results 47
3.2.2.2 The Economic Structure of the Model 53
3.2.2.3 The Policy Rule Equation 57
3.2.3 Simulation Results 60
3.3 Conclusions 67
ANALYTICS OF MONETARY POLICY . . . . . . . . . . 104
4.1 The Reserve Equation and Its Empirical Determination 106
4.1.1 Constructing Reserve Money from Balance Sheet Data 108
4.1.2 A Digression into the Measurement of Currency in Circulation 114
4.1.2.1 The Economic Costs of the Central Bank’s Mistake 116
4.1.3 The Empirical Determination of the Reserve Equation 120
4.1.4 Reserve Equation, Money Supply and Monetary Policy 123
xv
4.2.1 The Strategy of Open Market Operations at SBP 130
4.2.2 An Examination of SBP’s Operational Targets 134
4.2.3 Some Monetary Policy Models for Testing Operational Targets 139
4.3 Estimation and Simulation Results 143
4.3.1 Estimation and Simulation of the Lombra-Kaufman Model 143
4.3.2 Estimation and Simulation of the Free-Reserve Targeting Model 145
4.4 Epilogue – From Reserves to Interest Rates 148
4.5 Conclusions 154
5.2 Some Important Findings 241
5.3 Some Guidelines for Future Research 245
APPENDIX – A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
3.1 Complete Description of the Monetary Sector Model 70-72
3.2 Least Square Estimates of Behavioral Equations of Monetary Sector Model 73-74
3.3 Regression Statistics Accompanying the Estimated Equations 75
3.4 One Period Ahead Static Forecast Evaluation (1991 – 2007) 77
3.5 Many Periods Ahead Dynamic Forecast Evaluation with and without the Policy Rule (2000 –2007) 78
3.6 Impact Period and Long Run Multipliers of a 100 Basis Point Reduction in Discount Rate with and without the Taylor Rule
79
3.7 Impact Period and Long Run Multipliers of a 140 Basis Point Increase in T Z and a 100 Basis Point Increase in D Z with and without the Taylor Rule
80
4.1 Schematic Balance Sheet: SBP (Issue Department) 159
4.2 Schematic Balance Sheet: SBP (Banking Department) 160
4.3 Schematic Balance Sheet Of Banking Sector 161
4.4 Reserve Money and Its Components (The Supply of Reserves) 162
4.5 Reserve Money and Its Components (The Uses of Reserves) 163
4.6 Reserve Money and Its Components (Sources of Reserves) 164
4.7 Components of Money Supply (The Monetary Survey) 165
4.8 The Misalignment of the Discount Rate 166
4.9 Unborrowed Free and Autonomous Reserve Ratios 168
4.10 Various Types of Money Multipliers 170
xvii
4.13 Policy Targets & Shifts at State Bank of Pakistan 178
4.14 Liquidity Effects of Budget Deficit and Balance of Payments 180
4.15 Estimates of Behavioral Equations of the Lombra-Kaufman Model 183
4.16 Regression Statistics for the Estimated Lombra-Kaufman Model 184
4.17 Static and Dynamic Deterministic Forecast Evaluation Statistics - LKM Model (Version I) 194
4.18 Dynamic Deterministic Forecast Evaluation Statistics LKM Model (Version II) 194
4.19 Dynamic Stochastic Forecast Evaluation Statistics of LKM (Version I) 201
4.20 Dynamic Stochastic Forecast Evaluation Statistics LKM (Version II) 201
4.21 Estimates of Behavioral Equations of the Free Reserve Targeting Model 202-203
4.22 Regression Statistics for the Estimated Free Reserve Targeting Model 204
4.23 Static Deterministic Forecast Evaluation Statistics FRTM (1991-2009) 219
4.24 Dynamic Deterministic Forecast Evaluation Statistics FRTM (2000-2009) 220
4.25 Dynamic Stochastic Forecast Evaluation Statistics FRTM (2000-2009) 228
4.26 SBP’s Policy Targets and Changing SBP Governorship 230
4.27 Pakistan’s Monetary Regimes in Perspective 231
4.28 Estimates of ML-ARCH Equation for Growth Rate of M2 232
xviii
Number Title Page No.
2.1 The Effects of a Monetary Expansion when the Central Bank Misperceives the Money Supply Function 34
3.1 The Causal Nexus of Monetary Policy 46
xix
3.1 Actual and Fitted Values of the T-Bill Rate From the Policy Rule Equation 76
3.2 – 3.15 Many Periods Ahead Forecasts of MSM with and without Monetary Policy Rule (2000-07) 81-87
3.16 – 3.29 Dynamic Stochastic Simulation of MSM (2008-2015) 88-94
3.30 – 3.43 Many Period Ahead Forecasts of MSM (1991-2020) The Steady State Solution 95-101
3.44 The Interest Elasticity of Money Supply 102
4.2 Unborrowed, Free and Autonomous Reserve Ratios 169
4.3 – 4.4 Various Types of Money Multipliers 171
4.5 Endogenous & Exogenous Components of Reserve Money 173
4.6 – 4.9 Open Market Operations, T-Bill Rate & Inflation 175-177
4.10 Policy Targets & Shifts at State Bank of Pakistan 179
4.11 – 4.13 Liquidity Effects of Budget Deficit and Balance of Payments 181-182
4.14 – 4.18 Static Deterministic Simulation of the LKM: 1981-2000 185-187
4.19 – 4.23 Dynamic Deterministic Simulation of the LKM: 1991-2000 188-190
4.24 – 4.29 Dynamic Deterministic Simulation of the LKM: 1991-2009 191-193
4.30 – 4.34 Dynamic Stochastic Simulation of the LKM: 1991-2000 (LKM Model Solved with 4.31, 4.33) 195-197
4.35 – 4.40 Dynamic Stochastic Simulation of the LKM: 1991-2009 (LKM Model Solved with 4.30, 4.32) 198-200
4.41 – 4.53 Static Deterministic Simulation of the FRTM: 1991-2009 205-211
xx
LIST OF FIGURES (CONTD.)
4.54 – 4.66 Dynamic Deterministic Simulation of the FRTM: 2001- 2009 212-218
4.67 – 4.79 Dynamic Stochastic Simulation of the FRTM (2001-2009) 221-227
4.80 Simulated Value of Free Reserves without the Add Factor 229
4.81 Conditional Variance of Growth Rate of Money Supply 233
xxi
AG Loans and Advances to Government
ANBFI Loans & Advances NBFI ’s
ARBI Other Assets with Reserve Bank of India
AssetOther Other Fixed Assets
BBS Borrowing for Budgetary Support
BC Bank Credit
BDC Consolidated Budget Deficit
BoEInternal Internal Bills of Exchange and Commercial Papers
BoPak Balances Held outside Pakistan
BOT Balance of Trade
BPay Bills Payable
CBanking Currency Held in SBP (Banking Department)
CC Currency in Circulation
CDR Currency Deposit Ratio
CGN Government Consumption (Nominal)
CGR Real Government Consumption
Variable Name Description
CIssue Rupee Coin/ Notes
CMR Call Money Rate
CN Total Nominal Consumption
CPN Private Consumption (Nominal)
CPR Real Private Consumption
CR Total Real Consumption
CRF Capital and Reserves
D## Dummy variable for the ## year
DB Deposits of Banks
DEBT Total Outstanding Public Debt
DEBT_FLT Floating Debt
δG Total Deposits of Government by SBP
xxiii
DISCR Discount Rate
DNFA$ Change in NFA in Dollars (BOP Account)
DO Other Deposits at SBP
DoB Deposits of Banks
DYR Public Debt to YNN Ratio
DYR_F Floating Debt as Percent of Total Debt
ER Exchange Rate
G Gold Coins and Bullion
GDB Government Debtor Balance
GYNR Growth Rate of YNR
Γ Total Gold Reserves of SBP
xxiv
ID Weighted Average Interest on Demand Deposits
IG Investment in Government Securities
IGB Interest on Government Bond (10 Year Compound DSC Rate)
IGN Government Investment (Nominal)
IGR Real Government Investment
INBFI Investment in NBFI ’s
INFL Inflation Rate (from GDP Deflator)
INTA Weighted Average Interest on Long Term Advances
IOther Other Investments
ITR_N Indirect Taxes at Current Prices
ITR_R Indirect Taxes at Constant Prices
ITYR Indirect Taxes as Percent of YNN
LiabOther Other Liabilities
M Broad Money
m Money Multiplier
MM Money Multiplier (Derived as M2/R)
MN Imports (Nominal)
MR Real Imports
NCO Net Credit (Other)
NCOSBP Net Credit (Other) of SBP
NCPS Net Credit to Private Sector
NDA Net Domestic Assets
NFA Net Foreign Assets
NFI_N Net Factor Income from Abroad at Current Prices
NFI_R Net Factor Income from Abroad at Constant Prices
OD Other Deposits at SBP (Government Sector)
ODR Ratio of OD to Total Deposits
ORF$ Official Reserve Financing in Dollar terms
PD GDP Deflator (Market Prices)
PN GNP Deflator (Market Prices)
xxvi
RB Borrowed Reserves
RE Excess Reserves
Reval Revaluation Account
RF Free Reserves
RNCPS Real NCPS
RRD Required Reserves Against Demand Liabilities
RRT Required Reserves Against Time Liabilities
RU Unborrowed Reserves
S SBP’s Holdings of Government Securities
SBanking Bills Purchased & Discounted (Internal Bills of Exchange and Government T Bills)
SBPHANFI Dummy Variable for the HANFI Governorship Period
SBPKAZI Dummy Variable for the KAZI Governorship Period
SBPPAREKH Dummy Variable for the PAREKH Governorship Period
SBPYAQUB Dummy Variable for the YAQUB Governorship Period
SDRAllocation Allocation of SDR ’s by IMF
SDRBanking SDR Held with IMF
SDRIssue Unutilized Allocation of SDR ’s
SIssue Govt. of Pakistan Securities
SUBS_N Subsidies at Current Prices
SUBS_R Subsidies at Constant Prices
xxvii
T Total Time Liabilities of Commercial Banks
TBR T-Bill Rate
TD Time Deposits
Θ Total Foreign Assets of SBP
TREND Time Trend
UPP_DEBT Utilization of Private Proceeds
UPP_HBL Utilization of Private Proceeds
UPP_WAPDA Utilization of Private Proceeds
USTBR US T-Bill Rate
VCR Ratio of Vault Cash to Total Deposits
x Balancing Items (Net Liabilities) on SBP’s Consolidated Balance Sheet
XN Exports (Nominal)
XR Real Exports
y Balancing Items (Net Assets) on Consolidated Balance Sheet of Banks
YNN GNP at Market Prices (Nominal)
YNR GNP at Market Prices(Real)
YSAN Agricultural Sector Output at Current Prices
YSAR Real Agricultural Sector Output
YSIN Industrial Sector Output at Current Prices
YSIR Real Industrial Sector Output
xxviii
YSSR Real Services Sector Output
ZAK Zakat and Utilization of Privatization Proceeds
ZAKAT Zakat Fund Utilization
ZNBFI or Z Ratio of RR_NBFI to Total Deposits
ZT Required Reserve Ratio Against (TD+RFCD)
A Macroeconometric Analysis of The Effectiveness of Pakistan’s Monetary Policy
Chapter 1
Chapter 1: Introduction
A Macroeconometric Analysis of The Effectiveness of Pakistan’s Monetary Policy 2
Chapter 1
1.1.1 The Quest for Econometric Modeling of Monetary Policy
Decisions about monetary policy usually originate in the theoretical understanding
regarding what monetary policy is capable of doing and in the empirical appreciation of
technical and structural relationships signifying both the structure of the economy and the
behavior of economic agents. This theoretical, empirical and institutional understanding
governs the process of setting monetary policy objectives, choosing quantitative targets
and instruments and setting guideposts for monitoring the working of monetary policy.
This entire system is termed, as per conventional wisdom, a monetary policy framework
and it sets the environment within which monetary policy is designed (see e.g. McNees,
1987; Fry et al, 2000; Blinder, 1997, 1998; Bindseil, 2004 etc.).
Within a monetary policy framework, policymakers need to choose the way monetary
policy is to be operationalized. The conduct (or implementation or operationalization) of
monetary policy consists of ‘choosing an operational target, establishment of an
operational framework and the use of an instrument on a daily (weekly) basis to achieve
the operational target’ (Bindseil, 2004). This operationalization strategy, i.e. the set of
procedures to be followed for desk/window operations, governs the day-to-day (or the
week-to-week) working of the monetary authority such that deviations of the policy
variable from its short term and long term targets, conditional upon the economic and
institutional structure of the economy, are minimized and the central bank is able to fine-
tune the economy in the face of shocks over the immediate short-run.
In order for monetary policy to be able to deliver, the processes of the design and conduct
of monetary policy need to rest on strong empirical foundations. Central banks usually
perform this task by gathering a large volume of information about the economy and then
impose structure on this information to derive positive conclusions regarding the design
and conduct of monetary policy. The most suitable historically known method of
imposing such structure on a bulk of information is by means of specifying and
estimating an econometric model. While economists across academia and the central
banks disagree about the types of models they should choose for a particular purpose,
none disagrees to the demerits of disposing them altogether. The design and conduct of
an effective anti-inflationary monetary policy, as pointed out, e.g., by Brunner (1973), is
thus strongly tied with economic and econometric modeling of the causes and effects of
monetary policy.
1.1.2 An Overview of Pakistan’s Monetary Policy Framework
Pakistan’s monetary policy has largely remained unable to deliver. The State Bank of
Pakistan (henceforth SBP) declares its ultimate objectives to be (a) the supervision and
regulation of the financial system so as to ensure its soundness and broadening of its
services, (b) the maintenance of price stability with economic growth, (c) the
management of exchange rate and foreign currency reserves, and (d) the strengthening of
the payments system (State Bank of Pakistan, 2007a). A persistently high rate of inflation
that the SBP, and many others, endeavor to explain away by declaring it a non-monetary
phenomenon and / or outside of the domain of control of the central bank; a persistently
large differential between the average borrowing and lending rates; a relatively narrow
market for deposits and loans; a one-week long reserve maintenance period; a letting go
of the nominal exchange rate; and a foreign exchange reserve base well below the peer
average (State Bank of Pakistan, 2007b) constitute sufficient evidence that the SBP
misses on every target it sets itself1. Pakistan’s monetary policy, thus, stands unsuccessful
on every account.
This policy failure cannot be and is not intermittent. These outcomes, we conjecture,
come to sustained realization because Pakistan’s monetary policy lacks an empirically
well-founded design, is vague and imprecise about its conduct (implementation) strategy
and has therefore lost effectiveness. A brief elaboration and substantiation of each of
these claims follows.
1.1.2.1 Monetary Policy Framework: 1974-75 – 2008-09
The span of time ranging from 1974-75 – 2008-09 (which happens to be the sample
period chosen for this dissertation) may broadly be subdivided into two periods. The
period before 1990 when Credit Planning characterized monetary policy and interest
rates were fixed; and the period after 1990, when financial deregulation made it possible
for money and capital markets to function and monetary policy to work through the
interest rates (State Bank of Pakistan (2001).
1 See Appendix I.
The Credit Planning system worked by setting quantitative targets for the various
components of the (flows of) net credits extended to the public and private sectors and
the (flow of) net foreign assets, keeping in view the expected levels of components of
aggregate demand, the budget deficit and the balance of payments surplus / deficit. If all
credit expansions / contractions stayed within targets, the growth in money supply would
equal its pre-determined value. Since interest rates were held constant and the secondary
market for t-bills did not exist, monetary policy worked on the basis of direct controls e.g.
required reserve ratios, CRR and SLR, assigning of quantitative limits to credit
disbursements (both bank-wise and sector-wise), mandating banks to hold government
securities in proportion to their deposits base etc. (State Bank of Pakistan, 2001). The
monetary framework implicitly contained in this methodology is none else than Milton
Friedman’s famous constant money growth rule.
Following the financial deregulation of 1990’s, the management of monetary policy
underwent significant change. Direct controls were substituted away for indirect
measures and money market based instruments; credit ceilings were abandoned; the role
of multiple regulatory authorities was eliminated; money and capital markets were
allowed to function; and hence, ample room was created for ‘monetary policy to work
through the efficacy of interest rates’ (State Bank of Pakistan, 2001). However, the SBP
maintained its adherence to following the fixed money growth strategy as its monetary
policy framework (see e.g. State Bank of Pakistan, 2008), whereby the money growth
target was determined on the basis of estimates of net foreign assets, estimates of
government’s budgetary borrowing, and estimates of aggregate demand pressure
reflected in the saving-investment gap (State Bank of Pakistan, 2009c).
Within this framework, the SBP identified its monetary policy as consisting of a set of
discretionary measures that it implements as and when it deems necessary and which
derive from ‘a detailed review of the state of the economy, the practices of the banking
system and the statement of objectives of monetary policy’ (State Bank of Pakistan,
2009a, 2009c). It is indeed very clear from this description that, over the entire span of
time ranging from 1974-75 – 2008-09, the monetary policy framework that the SBP
subscribed to remained to be the fixed money growth rule.
Since the ability of a central bank to deliver upon its promise of keeping inflation low in
a fixed money growth environment critically rests on the stability of the money demand
function, therefore, much of the research in the area of monetary economics in Pakistan
remained confined to estimating money demand equations (or the income velocity
functions or the money multiplier) and testing their stability2. The SBP also relied only
on an estimated money demand equation to determine the impact of its monetary
operations (State Bank of Pakistan, 2008). The SBP, thus, ignored the fact that while the
stability of the demand for money is a necessary pre-requisite for targeting the rate of
monetary growth, it is not all that is needed to spell out its monetary policy framework3.
The disregard of the corroborated significance of the credit channel of monetary
transmission (State Bank of Pakistan, 2001; Agha, Ahmad, Mubarik and Shah, 2005;
2 See Section 2.4, Chapter 2 for a detailed review of this literature. 3 Specifically, a fixed money growth targeting regime requires that (a) the monetary aggregate be
perfectly controllable (i.e. it should be exogenous), and that (b) there be a stable and predictable relationship between the monetary aggregate and the price level/the rate of inflation (see e.g. Munoz, 2001).
State Bank of Pakistan, 2008b), as well as the endogeneity of money supply (Ahmad and
Ahmed, 2006; Omer and Saqib, 2008) forced SBP to remain ignorant of the effect of
interest rates on the supply of money. The failure to impose a quantitatively sound
theoretical structure to macroeconomic data, even after conjectural determination of the
causal nexus that explains the transmission mechanism of monetary policy both before
and after the financial reforms of the 1990’s (State Bank of Pakistan, 2001), thus served
as the key to the policy outcomes mentioned above.
1.1.2.2 The Targets, Indicators and Instruments of Monetary Policy
The relevance of a certain variable as being a target or an indicator or an instrument of
monetary policy derives; in conjunction with the institutional structure in which monetary
policy is to be conducted; from knowledge about the theoretical model of monetary
policy; from the lag structure with which a variable affects the objectives of monetary
policy; and from the exogeneity status of the variable itself within the empirical model
(see e.g. Tinbergen, 1952; Brunner and Meltzer 1967, 1969; Saving, 1967; Hamburger,
1970; Poole, 1970; Davis, 1971; Friedman, 1976; Pindyck and Roberts, 1976; Cagan,
1982; McCallum, 1990; Woodford, 1994). The absence of an econometric model for
monetary policy, thus, forces the SBP into an amalgamation of the target-indicator-
instrument problems. The choice, if any, remains but trivial.
The SBP thus remains confused as to whether it is the discount rate, or the t-bill rate or
the interbank market rate which monetary policy ought to be concerned with. The SBP
documents empirical facts about the economy and its operations using a large number of
trivially chosen indicators in its annual and quarterly reports, but leaves it as per its staff’s
discretion as to what inference may be drawn about its policy from each one of them4.
1.1.2.3 The Operationalization Strategy of Monetary Policy
The operationalization strategy of SBP’s monetary policy, i.e. what and exactly how SBP
intends to do over the desk / window in the period between its policy meetings, also
remained obscured until the publication of State Bank of Pakistan (2009c). According to
the Statement, before 2009, the SBP endeavored ‘to keep the weighted average overnight
repo rate stable and close to the policy discount rate … [using] … reverse .. open market
operations’ (State Bank of Pakistan, 2009c)5. However, the SBP, did not attempt to
formally spell out how such an operationalization strategy might link, within an
econometrically sound structure, with its overall monetary policy framework6. The
concern with keeping the weighted average overnight repo rate close to the discount rate
coupled with a passive adjustment of t-bill rate (through conducting a fixed number of
4 The evaluation that a certain change in the value of a policy indicator or a policy target is due, all or in
part, to a change in the policy instrument, or the evaluation that a certain change in the value of a target variable is not due to change(s) in the policy instrument(s), is an assessment of the same sort and order as done by the Romers’ (see e.g. Romer and Romer, 1989, 1993, 1997) to identify their dummy variable. Such evaluations should therefore be subject to the same criticisms (e.g. by Hoover and Perez, 1994a, 1994b; Christiano et al, 1996; Leeper, 1997) as against the Romers’ approach. (See Chapter 2 for details).
5 The Statement considered this operationalization strategy inadequate because while the discount rate served as a ceiling for the weighted average overnight repo rate, the same was free to move in the downward direction. This downward movement, signaling an excess supply of reserves, translated into a pressure on the exchange rate. The SBP, hence, moved towards the ‘corridor operational framework’ in 2009 whereby the weighted average overnight repo rate is supposed to stay within the limits set by the discount rate and the deposit facility rate (i.e. interest rate paid on excess reserves) which is set 300 basis points below the discount rate (see State Bank of Pakistan, 2009c for details).
6 From a theoretical perspective, the gap between the overnight repo rate and the discount rate is a variable which, according to the fundamental equation of monetary policy, is closely associated with borrowed / free reserves (see e.g. Tobin, 1968; Poole, 1968). Targeting borrowed reserves / net free reserves of the banking system is a strategy which has adequately been demonstrated to be a misleading guide for open market operations (Meigs, 1962 was the first to explain this inadequacy; several other accounts highlighting the inaptness of this strategy exist. See Chapter 4 for details). The question as to whether SBP’s operationalization strategy before 2009 refers to free reserve targeting or not is empirically taken up in Chapter 4.
open market operations each year7) led SBP only to announce, and quite incorrectly, that
changes in t-bill rate are not a signal of changes in the monetary policy stance (State
Bank of Pakistan, 2008). The trivialization of the t-bill rate and the assignment of a
passive role to open market operations, coupled with an operationalization strategy
subservient to money market conditions (phrase due to Guttentag, 1966) and lacking both
theoretical rigor and empirical content hence left monetary policy doomed to failure.
1.1.2.3 The SBP’s Conception of Money Supply
A further problem that inhibits SBP’s appraisal of the facts documented above is its
inability to appreciate what constitutes its reserve base and the money supply. As we
explain in Chapter 4, this leads not only to a miscalculation (though not of a very
significant order) of the outstanding quantities of currency in circulation and reserve
money, but also to using text-book definitions of the various monetary indicators when
the data implied indicators are different from them.
To cite but just one example, the SBP calculates the Currency-Deposit Ratio as Currency
in Circulation divided by the total volume of Demand, Time and Foreign Currency
Deposits (see e.g. State Bank of Pakistan; 2007b). The same variable is calculated in the
Economic Survey as Currency in Circulation divided by the sum total of Demand, Time,
Foreign Currency and Other Deposits with SBP (Government of Pakistan, 2008). We
show in Chapter 4 that both such approaches are naïve and that ‘Other Deposits with
SBP’ are essentially the same as ‘Currency in Circulation’. The appropriate method of
7 The SBP conducted 26 open market operations each year during the period 2000-2009, with the
exception of years 2002 and 2005 when it conducted, respectively, 25 and 27 open market operations.
calculating the currency deposit ratio should therefore consist of taking the ratio of
Currency in Circulation plus Other Deposits with SBP to the sum total of Demand, Time
and Foreign Currency Deposits. The SBP also overlooks computing some important
indicators that its policy targets and responds to.
1.2 PURPOSE AND OBJECTIVES
In consideration of the discussion regarding the elements an effective monetary policy
should be based upon, and the critical account of Pakistan’s monetary policy explicated
in the preceding section, it is indeed very clear that an objective empirical analysis of the
effectiveness of monetary policy must begin with economic and econometric modeling of
the causes and effects of monetary policy. Accordingly, this dissertation focuses on
investigating and evaluating the design and conduct of Pakistan’s monetary policy by
identifying, constructing, estimating and simulating some appropriate marginalized
macroeconometric model(s) of Pakistan’s monetary sector with a view to analyze the
effectiveness of Pakistan’s monetary policy.
In recent times, macroeconometric modeling has taken the form of estimating
cointegrating VAR models that derive from explicit dynamic optimization based behavior
of representative agents in a general equilibrium environment. Such analyses (e.g.
Clarida, Gali and Gertler, 1999; McCallum and Nelson, 1999; Smets ad Wouters, 2003,
Woodford, 2009 etc.) usually describe monetary policy in terms of an interest rate
regression that signifies central bank behavior. Much of this literature studies monetary
policy without explicitly incorporating money or the money market and hence leaves the
question of the effectiveness of monetary policy aside8. We do not take such an approach
in this dissertation. Instead, we focus on analyzing the quantitative strength of the impact
of central bank operations on the money market within the context of a
macroeconometric model that focuses exclusively on the workings of the money market
(hence the term marginalized). Thus our focus is entirely on studying the dynamic effects
of monetary policy within the money market – i.e. how changes in monetary policy
instruments transmit to market interest rates.
The analysis of the dynamic effects of monetary policy consists of three important
relationships in the money market: the demand for money, the supply of money and the
central bank reaction function which reflects the identification assumption. Much of the
research in the area of monetary economics in Pakistan has focused entirely on the
demand for money relationship. Even when the supply of money is relationship is
modeled, it does not take into account the feedback of interest rates on money supply9.
Finally, we find that virtually all studies of the effects or effectiveness of Pakistan’s
monetary policy assume, in line with SBP’s claim, that Pakistan’s monetary policy
framework is described by a fixed rule for monetary growth. The question of policy
identification is thus completely neglected. This dissertation, therefore, advances the
current understanding of the dynamic effects of Pakistan’s monetary policy as well as its
8 In fact, given the consensus view (called the New Neo-classical Synthesis) that monetary policy is
effective towards controlling inflation but irrelevant for explaining business cycles, and that systematic monetary policy matters only towards determining the extent to which technology-shock induced business cycles may propagate (see Goodfriend and King, 1999; Mankiw, 2006; Woodford, 2009, 2011), there is no need to question the effectiveness of monetary policy. In contrast, the analysis of monetary policy in this literature of the new vintage emphasis modeling only institutional aspects of the monetary and financial sector (see e.g. Goodfriend and McCallum, 2007; Gertler and Karadi, 2011, 2013; Gertler and Kiyotaki, 2010; Gerali, Neri, Sessa and Signoretti, 2010; Curdia and Woodford, 2011 etc).
9 See Section 2.4, Chapter 2 for a detailed review of this literature.
effectiveness in an important manner.
Since the demand for money relationship has already been studied extensively over the
past three decades in Pakistan, we take advantage of it and do not focus on re-
determining such relationship in this dissertation. However, the supply of money as well
the identification of monetary policy have received little or no attention at all so far.
Hence, this dissertation focuses on finding an empirical answer to the question of the
identification of Pakistan’s monetary policy as well as its operationalization strategy
within the context of some appropriate marginalized macroeconometric model(s) of
Pakistan’s monetary sector that take into account both the supply and demand
relationship of money within the specific institutional structure of the Pakistan economy.
This exercise is important precisely because such answer(s) would contribute to
improvement in both the design and conduct of Pakistan’s monetary policy and would
enable a stronger control over inflation.
In particular, we attempt:
• to reconstruct data on the various money and reserve aggregates (that are either not
published by the SBP or are miscalculated in official publications) that cast important
information about Pakistan’s monetary policy.
• to identify Pakistan’s monetary policy and its operationalization strategy in two
alternative but coherent theoretical scenarios.
• to construct and estimate marginalized monetary sector models that incorporate the
correct identifying assumption with a view to establish the quantitative importance of
the various channels of monetary transmission.
• to identify the channels of transmission of changes in various monetary instruments to
changes in the intermediate targets of monetary policy using the estimated and
simulated macroeconometric model(s).
• to simulate the constructed models to analyze and evaluate some recent monetary
policy decisions made by the State Bank of Pakistan.
1.3 AN OVERVIEW OF THE THESIS
The dissertation concentrates on two theoretical models of money supply, tailored to
represent Pakistan’s institutional structure, to develop the structure of macroeconometric
models for analyzing monetary policy. These models describe Pakistan’s monetary policy
framework as well as its operationalization strategy. While the monetary policy
identification assumption in Chapter 3 derives from econometric analysis of the links
between the policy instruments and final objectives, Chapter 4 makes use of the narrative
approach to identify operational targets of Pakistan’s monetary policy. Chapter 2 reviews
the relevant literature and critically evaluates earlier studies on Pakistan’s monetary
policy. The dissertation concludes in Chapter 5. Below we give a brief overview of each
of the chapters.
First of all, chapter 2 places the dissertation in the appropriate theoretical and empirical
context by recounting upon earlier theoretical works related to the identification of
monetary policy and to the construction and estimation of marginalized
macroeconometric models of monetary policy. The chapter also includes a critical review
of the literature on the analysis of monetary policy in Pakistan.
Chapter 3 consists of the first of the monetary policy models we develop for analyzing
Pakistan’s monetary policy. The model focuses on forecasting the Monetary Survey and
Monetary Statistics while obtaining the short term rate of interest as an equilibrium
process generated through the interaction of real and monetary variables. We identify
Pakistan’s monetary policy as a Taylor-type rule and corroborate the aptness of this
finding using the estimated and simulated marginalized macroeconometric model of
Pakistan’s monetary sector. We use this model to analyze three of State Bank of
Pakistan’s recent policy decisions as well as spell out the model implied transmission
mechanism of these policies in detail. The particularly important concrete conclusion
from this chapter consists of an estimate of the interest elasticity of the money supply
function which takes into account all structural linkages of the monetary sector.
Chapter 4 presents a non-orthodox analysis of Pakistan’s monetary policy. Owing to
SBP’s official position that its policy may not be described as a rule based regime, and to
certain theoretical criticisms regarding the interpretation of interest rate regressions as
policy reaction functions, we attempt a re-identification of SBP’s policy. The analysis
builds up on a reorganization of monetary sector data in accordance with our empirically
determined reserve balance equation for Pakistan to obtain consistent estimates of
borrowed, unborrowed, free and autonomous reserves for the Pakistan economy over the
course of the sample period covered in the dissertation. The chapter analyses various
indicators of monetary policy developed using our empirically determined reserve
balance equation for Pakistan which reveal important information regarding Pakistan’s
monetary policy.
The chapter develops two marginalized macroeconometric models of Pakistan’s
monetary sector; the Lombra-Kaufman Model (originally developed by Lombra and
Kaufman, 1984, and Bradley and Jansen, 1986) and the Free Reserve Targeting Model10,
that evaluate the hypotheses regarding SBP’s monetary policy operationalization strategy.
The chapter concludes with an explicit identification of Pakistan’s monetary policy
regimes across alternative governorships of SBP. An important contribution of this
chapter is the development of a new and objective methodology within the narrative
approach to policy identification.
Chapter 5 summarizes the various findings from the three marginalized
macroeconometric models that we construct, estimate, simulate and analyze in earlier
chapters and concludes the dissertation while pointing out some important guidelines for
future research.
10 This is our own construct. We derive this model from Poole (1982) and base it on similar logic as that
of the Lombra Kaufman model.
A Macroeconometric Analysis of The Effectiveness of Pakistan’s Monetary Policy
Chapter 2
LITERATURE REVIEW
Chapter 2: Literature Review
Chapter 2
LITERATURE REVIEW
This chapter seeks to place this dissertation in the appropriate theoretical and empirical
context. We concisely recount upon earlier theoretical works related to the identification
of monetary policy and to the construction and estimation of marginalized
macroeconometric models of monetary policy. A critical review of the literature on the
analysis of monetary policy in Pakistan follows.
As pointed out in chapter 1, an important part of our analysis is based on modeling the
supply of money. For this purpose, we make use of the Brunner-Meltzer Framework (see
Brunner and Meltzer, 1963, 1964, 1968; Meltzer, 1967; Burger, 1969; Brunner, 1973 all
excellently reviewed in Brunner and Meltzer, 1990; Papademos and Modigliani, 1990)
and the Free Reserves Model of money supply (see Meigs, 1962; Dewald, 1963; White,
1963; Davis, 1965; Guttentag, 1966; Poole, 1968 etc). Since this literature is now quite
well known, we do not add its review here. Adaptations of these models, tailored to
represent Pakistan’s monetary sector, appear in the relevant chapters.
The chapter proceeds as follows: Section 2.1 discusses the nature and significance of the
identification problem and covers the debate over the issue. Section 2.2 discusses issues
in econometric modeling of monetary policy. A critical review of the literature on the
empirical analysis of monetary policy in Pakistan is presented in Section 2.3. The chapter
concludes in section 2.4.
2.1 THE IDENTIFICATION OF MONETARY POLICY
The problem of the identification of monetary policy refers to the understanding as to
how much of the change in a monetary variable is the central bank’s reaction to economic
conditions and how much of it is due to private sector’s response to policy actions (Zha,
1997). The sorting out of these changes, respectively termed as monetary policy actions
and monetary policy shocks (Christiano, Eichenbaum and Evans, 1996), is vital because it
bears importantly on answers to central questions in the empirical analysis of monetary
policy (see Zha, 1997; Mankiw, 1999 for a discussion).
The identification of monetary policy consists of two elements: First, one needs to
ascertain a variable that the central bank defines (explicitly or implicitly) as its monetary
policy instrument. Secondly, changes in the monetary policy instrument need to be
examined for possible feedback effects from other macroeconomic variables (monetary
actions) so that endogenous changes in the realized value of the instrument are separated
from the purely exogenous innovations (monetary shocks). Thus, monetary policy
identification may be regarded as a method that decomposes total change in money
supply (or the monetary policy instrument) into these two components so that
unanticipated monetary policy changes are synchronous with shocks (see also Sims,
1992; Bernanke and Blinder, 1992; Leeper, Sims and Zha, 1996; Bernanke and Mihov,
1998; Höoppner, 2000 etc. for more details).
Identifying monetary policy, however, is difficult, because, in practice, economic
realizations are contemporaneous with policy actions and determining whether policy
precedes macroeconomic changes or whether macroeconomic changes induce a change in
policy variables is difficult11. Monetary economists have tried to solve this problem in
two ways. One is to take a narrative approach to monetary history. This approach dates
back at least to Friedman and Schwartz’s classic study, which attempts to outline the
causes and effects of monetary changes over a century for the United States while
emphasizing historical and institutional factors (Friedman and Schwartz, 1963).
More recently, Romer and Romer (1989, 1994) have extended this narrative approach.
The fundamental claim this approach makes is that by combining narrative information
about monetary policy decisions with modern statistics, the direction of causation
between monetary factors and real developments can be identified. The Romers make
statistical room to sound scientific by casting their results in the framework of a time
series model for industrial production and the unemployment rate. To see whether
monetary policy matters or not, they construct a dummy variable which takes the value
one for all periods in which there is a hint in the Records of Monetary Actions of the
Federal Open Market Committee (FOMC) about a change in the monetary stance to
control inflation, and include this dummy variable in an otherwise simple time series
model. Statistical significance of the difference between the errors when the dummy is
included and excluded is then taken as evidence of the significance of monetary policy
(see Romer and Romer, 1989, 1994).
11 In Mankiw’s words: “Since central banks do not conduct controlled experiments in practice, we
cannot observe what might happen to the economy after these actions have been taken and see their effects. In practice we only observe the equilibrium outcome of central bank actions and real economic activity on monetary variables. The study of monetary policy thus deals with the difficult task of sorting out the effects of central bank actions from the causes of those actions” (Mankiw, 1999).
A second way of trying to disentangle the causes and effects of monetary policy is to take
the econometric approach. That is, rather than relying on a careful reading of history, one
can study the effects of monetary policy by applying time-series analysis to
macroeconomic data. To make this approach work, some identifying assumption, i.e. the
behavioral rule the central bank sticks to for conducting monetary policy, is necessary to
sort out cause and effect. One might assume, for instance, that changes in short-term
interest rates not explicable by other macroeconomic variables reflect changes in the
preferences of the central bank towards inflation. Broadly, such identifying assumptions
may range over anything from Friedman’s constant monetary growth rule to inflation
targeting to fiscal theory of price level, or to Taylor’s rule or Calvo’s Principle etc.
One drawback of this econometric approach is that it relies on identifying assumptions
that are usually open to dispute12. On the other hand, compared to the narrative approach,
the econometric approach asks for less subjective judgment on the part of the researcher,
which increases the verifiability of the results and reduces the possibility of involuntary
bias. Moreover, as Hoover and Perez (1994a, 1994b) and Leeper (1997) show, the
descriptive approach is no better than the econometric approach towards understanding
monetary policy in that the identification assumption has to be determined exogenously
for both. While Romer and Romer (1997) attempt to re-establish the significance of their
methodology, Christiano, et al (1996) show conclusively that the Romers’ contention of
considering every particularly big policy move as completely exogenous implies that the
12 Bernanke and Mihov (1998) discuss this dispute regarding the choice of appropriate identification
assumption for studying US monetary policy. However, Sims (2007) and Gali and Gertler (2007) document the fact that virtually all monetary policy models have verified that monetary policy can best be identified as an interest rate setting rule. We corroborate the same for Pakistan economy.
feedback component of monetary policy is zero. This in itself constitutes an identification
assumption which has no justification at all.
The dominance of the econometric approach – i.e. applying econometric analysis to
macroeconomic data to determine the monetary policy reaction function – over the
narrative approach coupled with the belief that most central banks conduct monetary
policy through adjusting interest rates based on analyses like Taylor (1993, 1998),
Rotemberg and Woodford (1997) and Clarida, Gali and Gertler (1998), has subsequently
led monetary policy analysts across central banks and academia to identify monetary
policy by running interest rate regressions13. The predictive accuracy of these interest rate
equations is then taken as evidence that the central bank is adhering to a certain policy
rule.
The econometric approach, however, has recently been criticized by e.g. Minford,
Perugini and Srinivasan (2002, 2003), Carare and Tchaidze (2005), Auray and Feve
(2003, 2008) among others, on account of its strict adherence only to econometric
goodness of fit criteria, which are but insufficient to identify monetary policy. Thus,
Minford et al (2002) argues that interest rate regressions signifying a relationship
between interest rate, output and inflation do not conclusively identify the monetary
policy reaction function unless other supplementary qualitative information about central
bank behavior is also known. The argument draws up on Taylor (1999) which observes
that actual interest rate behavior resulting from use of a Taylor rule would be the same as
13 Carare and Tchaidze (2005) report that the search for the keyword “monetary policy rules” for 2000–
03 in the EconLit database returns 361 published articles, or an average of 90 a year.
the interest rate behavior which might result if the central bank were following, e.g., a
fixed money growth rule. Carare and Tchaidze (2005) document this observational
equivalence by pointing to a number of different interest rate regressions fit to actual US
data in the 80’s and 90’s each trying to explain the same interest rate behavior with a
somewhat different policy rule. Minford et al (2003) shows further that this observational
equivalence leads to substantial differences in the stochastic behavior of the economy and
hence no conclusion about which policy rule a central bank is following may be obtained
from simple regressions. Similarly, Chowdhury and Schabert (2008) consider interest rate
rules and money supply rules as two ways of viewing / implementing the same monetary
policy.
On the other hand, Auray and Feve (2003, 2008) demonstrate that Friedman’s money
growth rule and Taylor’s interest rate rule may turn out observationally equivalent only
under very implausible macroeconomic conditions (i.e. the existence and influence of
extrinsic random variables (the so called sunspots) on macroeconomic equilibria).
McCallum (2002) also shows that interest rate behavior derived from a Taylor rule is not
the same as one derived from McCallum’s base money rule14. While the literature on
observational equivalence (or otherwise) is still evolving, one conclusion out of these
studies that Minford et al (2002) points to is indispensable: that while the econometric
approach to the identification of monetary policy is superior as compared to the approach
of the Romers’ (see Romer and Romer, 1989; 1994; 1997; 2004), it must be
14 This relatively less well-known rule was originally proposed by McCallum (1988) and repeatedly
referred to in McCallum (1993, 2000 and 2002). It has recently become the subject of policy discussions following Plosser (2008).
complemented with other supplementary information before one can categorically
determine the policy framework a central bank adheres to.
A Note on Our Methodology
While debate over the adequacy of the two approaches to monetary policy identification
still continues, a wide range of macroeconometric models for monetary policy analysis
have emerged that are robust to Lucas’ (Lucas, 1976) and Sims’ (Sims, 1980) criticisms
(see e.g. McCallum, 2001; Gali and Gertler, 2007 and Sims, 2007 for a review of this
literature). These models impose structure on macroeconomic data either through
explicitly incorporating expectations and identification assumptions or through dynamic
optimization based microfoundations.
However, the entirety of this literature studies monetary policy without explicitly
incorporating money or the money market. The phenomenon of vanishing of the LM
curve from IS-LM analysis as explicated in Romer (2000) serves as the hallmark of these
studies. Although this approach has shown to work (even though McCallum (2001)
considers it loosely quantitatively misspecified), we still need marginalized models of
sub-sectors of the economy that can explain more than these aggregate models. Notable
example of such sub-sector models include Boughton, Brau, Naylor and Yohe (1969),
Christ (1971), Palanivel and Klien (1999) and International Monetary Fund (2001).
Although these models do not incorporate optimizing households and firms, they provide
a very rich framework for including all relevant structural details about the monetary
sector of the economy while analyzing monetary policy.
Accordingly, we base the theoretical structure of our marginalized monetary sector
macroeconometric models in accordance with the structure implied by these studies.
Methodically, our approach differs from the existing literature on empirical monetary
policy analysis in two essential ways. First, contemporary monetary policy models,
whether they are of the SVAR or DSGE vintage, are economy-wide models and focus on
ascertaining the effect of monetary policy on output inflation and/or the exchange rate.
We do not engage ourselves with such issues. Instead, we focus on tracing the effects of
changes in monetary policy instruments on monetary equilibrium within a marginalized
sub-sector macroeconometric model. This exercise, as pointed out e.g. by Modigliani,
Rasche and Cooper (1970), is a necessary pre-requisite for determining the effectiveness
of monetary policy as it establishes the dynamics of the various monetary policy actions.
Secondly, the economic specification of our monetary sector models is much more
detailed as compared to the existing models (whether of the SVAR or DSGE vintage)
used for this purpose. While the existing models offer a very rich environment for
analyzing monetary policy effects, they typically ignore a lot many details about the
monetary sector itself. Whereas DSGE models of this new vintage are in early phases of
development, (see e.g. McCallum, 2007; Goodfriend and McCallum, 2007; Gertler and
Karadi, 2009, 2011, 2013; Gertler and Kiyotaki, 2010; Gerali, A., Neri, S., Sessa, L., &
Signoretti, F. M. (2010); Curdia and Woodford, 2011 etc.), they still leave a lot that needs
explanation.
Finally, we make one important departure from the mainstream narrative approach (in
Chapter 4) before applying econometric methods to establish policy identification. Thus,
rather than relying on the historical records of monetary policy committee to obtain
information about the changing monetary policy stance, we make use of the consolidated
balance sheet of the central bank to extract such information. The motivation behind this
happens to draw on Bindseil (2004) which notes that monetary policy always works by
changing the price or quantity of some item on the balance sheet of the central bank.
2.2 THE ECONOMETRIC MODELING OF MONETARY POLICY
Macroeconometric modeling came into fashion following World War II at the Cowles
Commission for Economic Research (which was later renamed as The Cowles
Foundation) by predominantly Keynesian economists15 and served as the apparatus that
provided for economic engineering of society through use of policy16. The foundations of
these large scale simultaneous equation macroeconometric models (developed in line
with research at the Cowles Commission) towards economic measurements dwelled in
(a) modeling simultaneity in economic outcomes by specifying and estimating linear (or
logarithmically linear), systematic, behavioral relations among observable variables that
15 Some of the most notable associates of the Cowles Foundation have included, inter alia, Jacob
Marschak, Tjalling C. Koopmans, James Tobin, Kenneth J. Arrow, Trygve Haavelmo, Theodore W. Anderson, Lawrence R. Klein, Gerard Debreu, Leonid Hurwitz, Herbert A. Simon, Harry Markowitz and Franco Modigliani. Many also received Nobel prizes for their contributions. See http://cowles.econ.yale.edu/archive/people/directors/ and http://cowles.econ.yale.edu/archive/people/nobel.htm for complete and detailed lists.
16 Some of the notable macroeconometric models developed in this tradition included the Klein- Goldberger model; the Wharton model (which evolved into the Data Resources, Inc. (DRI) model); the Brookings model; the MPS (MIT-PENN-Social Science Research Council) model which later developed into the FRB/US (US Federal Reserve Board) model; the BEA model; the St. Louis model; the Michigan model; the Hickman-Coe model and the Fair-MC model. Macroeconometric models for countries other than the US included the DRI-Canada and MFTM models for the Canadian economy; the NIESR model, the Oxford model and Bank of England models for the UK; the Dutch macroeconometric model and the Scandinavian model (see e.g. Bodkin, Klein and Marwah, 1991; Welfe, 2013 for more details). The most ambitious of such models happens to be Project LINK model which seeks to develop a world-wide macroeconometric forecasting model.
derive from economic theory; (b) resolving the exogeneity-endogeneity status of model
variables (i.e. imposing statistical identifying restrictions) on the basis of some relevant
economic theory and in line with the question(s) confronted; (c) estimation of the
behavioral relationships in such a manner that the structural residuals are normally
distributed, zero mean, white noise processes with finite variances; having a non -
singular covariance matrix, and exhibiting both serial independence and dynamic
stability; and (d) establishing the independence of the explanatory (i.e. policy) variables
thus ascertaining the robustness of policy simulations17.
However, the inability of many models constructed on such principles in forecasting the
sequence of atypical happenings in the 1970’s and 1980’s coupled with theoretical
developments like e.g. (a) Milton Friedman’s and Robert Lucas’ criticisms against the
existence, first, of a long-run and eventually, under rational expectations, even of a short
run, trade-off between inflation and unemployment (Friedman, 1968; Phelps, 1968;
Lucas, 1972); (b) the domination of macroeconomics by the rational expectations
hypothesis; (c) Lucas’ and Sims’ criticisms (Lucas, 1976; Sims, 1980) against
econometric policy evaluations; (d) popularization of structural cointegrating VAR
models following Sims (1980) etc., made the Cowles Commission approach appear less
and less appealing to a large number of newly trained professional economists (see
Favero, 2007 for a good discussion). Consequently, the Cowles Commission approach,
17 Following Bautista (1988) and Capros, Karadeloglou and Mentzas (1990), these models are classified
under two broad headings: the macroeconometric models and computable general equilibrium models. The former are further subdivided by Challen and Hagger (1983) into five broad types on the basis of their structure. These correspond to the Keynes-Klein type models, the Phillips-Bergstrom type models, the Walras-Johansen type models, the Walras-Leontief type models, and the Muth-Sargent type models. See Valadkhani (2004) for more details.
being named “econometric failure at the grand scale” (Lucas and Sargent, 1979) or being
called “non-representative of data as well as theory” (Pesaran and Smith, 1995), became
quite unpopular among academia and macroeconomics became dominated first by
structural cointegrating VAR’s and later by the DSGE approach towards structural
modeling of ‘deep parameters’ for quantitative policy evaluations.
One major advance in construction of macroeconometric models of this new vintage has
been the recent demonstration that the solution to a log-linearized DSGE model can be
approximately reasonably represented by a cointegrating-VAR model (see e.g. Favero,
2007; Canova, 2007). The traditional criticism against structural cointegrating-VAR
models regarding their being purely statistical representations of data lacking economic
content, their reliance on a set of unsustainable assumptions, and being fundamentally
flawed on account of the Lucas’s critique (Canova, 1995), thus also looses significance.
However, since a theoretical DSGE model may not be considered as a model for the data
but (only) as a generator of prior distribution both for parameter estimation and model
evaluation (Sims, 2007), therefore the structural cointegrating-VAR models, which were
traditionally considered “a tool to summarize data interdependences, to test generically
formulated theories, to conduct policy analyses, and, more recently, .. a way to compare
actual data with the time series generated by artificial economies with calibrated
parameters” (Canova, 1995), gain a new interpretation and face a new dilemma: they are
models suited “not … to yield advice on the best policy but rather to provide empirical
evidence on the response of macroeconomic variables to policy impulses” (Favero,
2007). It is exactly in this sense that Mankiw (2006) entitles the New Neoclassical
Synthesis (see Goodfriend and King, 1999; Mankiw, 2006; Woodford, 2009 and 2011,
for details) and the associated DSGE models (or their log-linearized cointegrating-VAR
representations) as being ‘a research program … too abstract and insufficiently
practical’. Practical policy evaluations and policy advice, thus, continue their reliance on
models based on the Cowles Commission approach18.
Pragmatic macroeconometric modeling in the Cowles Commission tradition, however,
has identified, in response to past failures and criticisms, a number of important issues
that need to be adequately addressed. These include: (a) an adequate attention to data
quality as well as use of appropriate and validated economic theory (Sowey &
Hargreaves, 1991); (b) appropriate handling of proxies, measurement errors, additional
variables, functional form, dynamic structure and stochastic specification of the model
(Pesaran & Smith, 1985); (c) a careful sorting out of causality, parameter interpretation,
identification and time series properties of the data (Kloek, 1988); (d) adequate
determination of the micro foundations of the model – However, in this regard Schlicht
(1985) suggests that even without having a comprehensive knowledge of microeconomic
foundations, one can still construct a macroeconometric model and obtain fairly
reasonable results, while Fisher (1983), Houthakker (1956), and Stiglitz (1969) mention
that due to the differences in the qualitative relationships between micro and macro laws,
18 Mankiw (2006) cites particular examples of successful monetary and fiscal policies from recent US
history that pay little attention to the DSGE based NNS wisdom and explicitly derive from some macroeconometric model (like e.g. the FRB/US model) of the predominantly Keynesian type. The Bank of England’s macroeconomic model (The MTMM) also belongs among the same class of models (Pagan, 2003). Project LINK World Model (http://projects.chass.utoronto.ca/link/) is another recent example of the practical usefulness of such models.
the generalization of microeconomic rules to a macro level does not generally yield
consistent generalization (see Valadkhani, 2004 for further details).
Thus, in what follows, we shall take the received pragmatic wisdom as granted and will
follow the Cowles Commission approach towards contemplation of the tasks set in
section 1.2. Since various accounts of the history and theory of such methodology exist in
literature, we do not attempt to review the same in this dissertation (see e.g. Challen and
Hagger, 1983; Bodkin, Klein and Marwah, 1986a, 1986b, 1991; Fair, 1987; Valadkhani,
2004 etc.). While the use of such approach may be considered a limitation of the present
research by some, we consider pragmatic explanation of observed macroeconomic facts
within a theoretically coherent framework as the necessary prior (as e.g. pointed out by
Goodfriend and King, 1999 and Woodford, 2009, 2011) to the use of sophisticated
modeling methodologies.
2.3 THE STUDY OF MONETARY POLICY IN PAKISTAN
While there has been a significant amount of research effort devoted towards
understanding money demand, money supply and money multiplier processes in Pakistan
independently over the past two decades, the building up of a complete
macroeconometric model that specifically focuses on delineating the causes and effects
of monetary policy has not been in place. Very recently, some effort in this direction has
been initiated by the SBP but the same fails to be any impressive at all19.
19 See Hanif et al (1998). This unpublished manuscript from the State Bank of Pakistan’s research staff
constructs a model of the economy and estimates it using nominal macroeconomic data.
Monetary economists in Pakistan have, in general, tried to explain the impact of
monetary policy using only estimated demand for money functions. This practice began
in mid 1970’s20 and forms the core of research in the area of monetary economics in
Pakistan. Some important studies, amongst others, that benchmark this line of
investigation include Akhtar (1974), Abe et al (1975), Khan (1980, 1982a, 1982b, 1989),
Ahmad and Khan (1990), Hossain (1994), Khan (1997), Ahmad and Munir (2001),
Qayyum (2005) and Hsing (2007))21. The various approaches that this investigation
scheme has employed from the 1970’s up to 2009 derive from one fundamental
observation: that the unit root from the error term in the demand for money relationship
cannot be disposed off even after accounting for the impact of income, interest rate(s), the
general price level, the inflation rate and any dynamical short run inertia influences on
the demand for money. Accounting for this unexplained trend component gives rise to
various forms of the money demand relationship.
While some authors believe that changing the definition of money supply, income and or
interest rates (or broadening the spectrum of interest rates to be included in the equation)
should be sufficient to get rid of this additional trend (e.g. Khan, 1981; Tariq, 1997),
some others believe that employing new time series techniques and/or including variables
like financial innovation, financial liberalization and deregulation, exchange rate (or
deposits denominated in foreign currencies), monetized fiscal deficits and/or variables 20 Empirical estimation of money demand relationships dates back to the works of Feige (1967). Judd and
Scadding (1982), Goldfeld (1976) and Laidler (1977) set important benchmarks in this area of investigation. Much of the empirical money demand estimates are based on the guidelines suggested by these authors. Some principal money demand studies for Pakistan include Akhtar (1974), Mangla (1979), Khan (1982), Hasan (1987), Hossain (1994) among many others.
21 The preoccupation with studying the demand for money seems to arise out of the belief that the State Bank of Pakistan is correctly describing its monetary policy framework as a fixed money growth targeting regime.
linking to stock market activity etc. should explain this additional trend (see e.g. Khan,
1994; Hossain, 1994; Qayyum, 1994; Agénor, 1996; Khan, 1997; Ahmad and Munir,
2001; Caruso, 2001; Khan, 2005; Oskooee, 2006; Hye, 2009; Rao, 2009). Whereas some
authors take a disaggregated approach to study money demand to account for this missing
trend (e.g. Khan and Arby, 2000; Qayyum, 2001), a few believe that money demand is
inherently unstable and instead concentrate on studying the income velocity relationship
or the dynamics of the money multiplier (see e.g. Abbas and Qasim, 1993; Chaudhry and
Khan, 1997; Qayyum, 2006; Omer and Saqib, 2008). Nonetheless, this broad spectrum of
studies is not only insufficient to study the entire monetary sector and the impact of
monetary policy on the macroeconomy but also misleading.
In 2006, when the first and only successful round of combating inflation using monetary
policy had just finished, SBP announced a conference titled “Monetary-Cum-Exchange
Rate Regime: What Works Best for Emerging Market Economies?”, which sought a
theoretical and empirical answer as to whether the SBP adopt inflation targeting as the
new monetary policy framework or not. The proceedings were published in the newly
born SBP Research Bulletin and quite ardently summarized in Ahmad (2006). The
conference, apart from defending the managed-float exchange rate strategy and
identifying the need for prudent econometric modeling of monetary policy, rejected
adopting inflation targeting as the monetary policy strategy.
The SBP, even though after describing the conference as successful, continued to practice
the pseudo inflation targeting regime. The inflation model that the SBP developed in
response to the issues identified at the conference was based on the assumption that
inflation has a life of its own22. The time series models that the SBP uses in conjunction
with other narrative information about the macroeconomy are equally inappropriate,
insufficient and misleading as the demand oriented approach described earlier.
A yet another stream of research about Pakistan’s monetary sector can be found in studies
by e.g. Haider and Khan (2008), Pasha et al (1996), Hasan (1987), Saqib and Ahmed
(1986) and Imam (1970). These studies attempt to delineate the effects of monetary
policy using contemporary state-of-the-art macroeconometric modeling approaches but
fail to appreciate the underlying structure of the economy they are analyzing by making
their analyses rest on naïve assumptions.
The Financial Sector Assessment (FSA) Report (State Bank of Pakistan, 2001) gives
details of the monetary transmission mechanism in Pakistan both before and after the
financial sector reforms of the early 1990’s. According to FSA, the transmission of
monetary policy in the pre-reform period occurred “through the credit channel but
without the efficacy of interest rates”, whereas, in the post reform period, “monetary
transmission is taking place through the interest rate channel”. In any case, the FSA
maintains that monetary policy affects aggregate demand and the price level through
affecting direct and portfolio investment, demand for durables and net exports (see State
Bank of Pakistan, 2001). The SBP’s strategy of ignoring this feedback in calculating the
22 This resulted in a number of univariate time series models of the various measures of the rate of
inflation. Inflation forecasts happened to be the average of the forecast from alternative models. At best, this strategy carried over to simultaneous modeling of the various sectoral components of inflation thereby giving rise to VAR models of inflation components.
impact of monetary policy implicitly amounts to assuming that the interest elasticity of
money supply is zero23. This means, on the one hand that forecasts of future activity do
not take into consideration the credit channel of monetary transmission, while on the
other, implies that actual money supply changes associated with given nominal interest
rate changes will be smaller than expected and controlling inflation would be more
difficult as it may otherwise appear.
To elaborate, consider the case where the central bank’s estimate of supply elasticity is
smaller as compared to its actual value. The central bank would then misperceive the
money supply function to be SM1 instead of SM . If the central bank wants to cut the
nominal interest rate by BA ′′ , it would increase the supply of money by amount AE ,
which would at best be insufficient for this purpose (keeping in view the correct money
supply curve the required increase in money supply will be AC ). Since ACAE < ,
therefore, there would always be excess demand for liquidity in the money market (the
actual money supply curve will shift to ''SM while the interest rate is at B′ ; hence excess
demand for liquidity would equal the distance BF ). Constrained with a smaller supply
of reserves, the banking sector would face an excess demand for credit by the private
sector and thereby the divergence between lending and borrowing rates would persist.
Attempts to cut the t-bill rate below D would only intensify the excess demand
for liquidity and so long as the misperceived elasticity of supply is not corrected
for, the divergence between borrowing and lending rates would persist. Indeed,
23 Ahmad and Ahmed (2006) observe that money supply is exogenous neither in the short nor in the long
run. This observation alone makes SBP’s strategy of monetary policy scientifically invalid.
SupplyMoneyDemandMoney &
BANK MISPERCEIVES THE MONEY SUPPLY FUNCTION
SOURCE: AUTHOR’S CONSTRUCTION
this situation will only be more intensified if private credit demand for new investment is
ridden with indivisibilities. Monetary equilibrium would only be possible if money
demand shifts down. Thus, a misperceived supply elasticity of money might lead an
expansionary monetary policy headlong into recession while asset prices (both real and
financial) would continue to rise24. Assessing monetary policy actions solely on the basis
of an estimated money demand function, therefore, may prove catastrophic.
Pasha et al (1996) rests on a very simplistic but completely supply oriented description of
the monetary sector of the economy25. The model thus remains completely ignorant and
independent of the demand for real money balances. As such, the structure of the model
implies that the money market always remains in equilibrium whatever state the rest of
the economy is in. This implicitly means that economic agents are always willing to hold
as much money as is thrown by the central bank in to the economy. If this is held true,
then one is forced to believe, on the one hand, that monetary growth is completely
absorbed by changes in real income in each time period, and on the other that changes in
interest rate cannot be accounted for in any of the markets and they remain purely
exogenous. Thus inflation can never even be a fractionally monetary phenomenon.
24 Pakistan has recently experienced this entire phenomenon on its macroeconomic front: a marked
divergence between lending and borrowing rates, an almost stagnant real investment with both unemployment and stock prices rising to unprecedented levels, an upsurge in exchange rate and real estate prices etc. We offer only one possible interpretation of all of these observations i.e. the misperception of policy makers about monetary transmission. The validity or otherwise of any such interpretation, however, is not the subject of this paper.
25 This model explains changes in m

a macroeconometric analysis of the effectiveness of

Documents