Journal Issue
Share
Article

Analysis of the Yield on Foreign Exchange Bearer Certificates: Rationality and Financial Behavior in Pakistan

Author(s):
Karim Abdel-Motaal
Published Date:
December 1994
Share
  • ShareShare
Show Summary Details

I. Review of Theory

Economic theory establishes a set of relations between the returns on different financial instruments that would obtain in an efficient and frictionless financial system. These conditions derive from portfolio optimizing behavior by rational agents. Arbitrage provides for nonpersistence of unexploited profit opportunities and an associated impossibility of forecasting them. To demonstrate this result, three strands of the macro/international literature are briefly reviewed.

The “term structure” literature, including contributions by Shiller (1979), Mankiw (1986), and Fama and Bliss (1987), examines the relation between the rates on different maturities of an instrument. The yield curve has received considerable attention due to its implication for a number of macroeconomic questions. It is widely believed, for example, that central banks have more direct control over short-term interest rates than over long-term rates. In so far as capital is costly to adjust and takes time to come on line, investment decisions may depend on the latter. The monetary transmission mechanism therefore which defines the link between policy and aggregate demand depends critically on the relationship between these rates.

The “Expectations Hypothesis” of the term structure, which is the main theoretical contribution of this literature, posits that long-term interest rates are a weighted average of present and expected future short rates. A positively sloped yield curve therefore reflects market expectations that short rates will rise. An approximately equivalent restatement is that expected one-period holding returns on bonds of all maturities should be equal. Since a long bond contract can be reproduced through a contract consisting of a series of short bonds, the two contracts should have the same price and expected return. In its purest form, the theory requires that the previous statements hold strictly, or that term premia are exactly zero. More recent work has shown that the only theoretically implied restriction is that term premia be time invariant. A particularly simple form of the proposition is equation (1.1) below:

where h(v,w:t) is the w period holding return at time t, on a bond maturing in v periods. Note η is not time subscripted. The empirical verdict on the hypothesis is mixed. There is some indication however that profit opportunities are not systematically forecastable, especially in the short-maturity end of the market. 1/

The “interest rate parity” literature, represented by Dornbusch (1983), Frenkel and Levich (1986), and Cumby (1987), investigates efficient arbitrage in international financial markets. Portfolio optimizing behavior in a context of perfect capital mobility implies the law of one price, i.e., the proposition that identical financial instruments ought to offer the identical returns, when expressed in the same currency. This work is essentially an extension of closed economy portfolio choice models. The efficiency proposition comes in ‘uncovered’, and ‘covered’ interest rate parity, forms. Namely,

where, r(x:t) is the return on an x period maturity bond at time t, and i(x:t) is the international equivalent. St is the exchange rate, defined as the local currency price of foreign exchange. Ft, t+1 is the forward exchange rate at time t for period t+1. Both (1.2) and (1.3) are simple rate arbitrage equations. The second assumes participation in forward currency markets, hence the term ‘covered’.

Empirical tests, under different assumptions concerning asset substitutability and capital mobility, generally sustain covered parity and reject uncovered parity. This calls into question the efficiency of foreign exchange markets, and has led to a battery of associated work on the relation between forward and spot exchange rates. The issue centers fundamentally around expectations formation. The so called ‘Peso problem’ of exchange rate credibility is capable of explaining the failure of uncovered parity. 1/ It amounts to a risk premium associated with the nonzero probability of exchange rate devaluation.

Finally, the ‘Asset Pricing’ literature, including portfolio choice models, and more recently, general equilibrium ones along the lines of Cox, Ingersoll, and Ross (1985), establishes indirectly a result of fundamental relevance to this exercise. In the process of optimal portfolio design, a rational agent will require compensation for risk associated with holding an asset. Portfolio choice models define risk as a set of second and higher moments of returns contingent on relevant market variables. The recent Capital Asset Pricing class of models (CAPM, CCAPM), require that risk be defined only in terms of covariance of returns with the market. In either case, the intuitive implication that identical instruments are driven through arbitrage to yield identical returns, is established.

The body of theory reviewed above gives some justification to the expectation stated in the introduction, that a ‘rational’ interest rate structure ought to prevail in a smoothly functioning financial market. After correcting for differences in maturity, holding returns, and currency of denomination, identical assets should command equal returns. 2/

II. The Financial System of Pakistan

In the period under analysis (1986-93), the financial system of Pakistan was characterized by a considerable degree of regulatory control. 1/ The State Bank of Pakistan (SBP) relied quantitative tools to channel credit to priority sectors. Under a “Priority Sector Lending Scheme”, commercial banks were assigned mandatory credit targets. Commercial banks, dominated by large public sector banks, are the primary financial intermediaries. They were subject to constraints on balance sheet optimization in the form of interest rate controls. A ten percent floor and twenty two percent ceiling exist on loans. Restrictions on deposit rates were also maintained on local and foreign currency deposits. Foreign currency accounts were paying spreads less than one percent above LIBOR.

A plethora of public debt instruments have traditionally been available, tapping various segments of the saving population. Federal Investment Bonds and Treasury bills are two local currency debt obligations of the central government that are currently auctioned at regular intervals. The returns on these instruments are partly reflective of market forces. In an attempt to access the foreign currency holdings of Pakistanis abroad, foreign currency accounts and FEBCs have been available since the mid-1980s. More specialized debt instruments are issued by other public institutions at the federal and provincial levels. The most important of these are Defense Savings Certificates, Mahana Admani (Islamic characteristics), and National Savings (NSS) Accounts.

A series of financial sector reforms were initiated in late 1989, with the goal of revising the relationship between the SBP and the fiscal activities of the government on the one hand, and the banking sector on the other. This involved a move towards indirect monetary control, the removal of quantitative restrictions, liberalization of interest rates, and the relaxation of mandatory credit schemes. These reforms resulted in a more competitive environment with the entry of new private institutions, the introduction of new financial instruments, and a perceptible reversal of a previous trend towards disintermediation. 2/

III. Salient Features of FEBCs

Foreign exchange bearer certificates are three year rupee denominated obligations of the Pakistan Government, that must be purchased with foreign exchange. They were introduced in August 1985 as a means of encouraging repatriation of foreign currency from overseas workers and assisting in bringing foreign exchange into the formal economy. Bonds can be encashed at any time in either rupees or foreign exchange at the prevailing official exchange rate. They are similar to zero coupon bonds in that a one-time interest payment is paid at maturity, or at the time of encasement for each completed one-year period. The amounts payable are currently 114.5, 131, and 152 percent of face value after one, two, and three years, respectively. There are no residency or source disclosure requirements. The instrument has been actively traded on the Karachi stock exchange, at a premium since time of inception. Until recently, it was the only instrument free of administrative control, and for which a well established secondary market existed. This makes it appropriate for an examination of financial behavior in Pakistan.

In principle, the FEBC has the characteristics of both a local and foreign currency asset. In so far as an agent must enter the foreign exchange market to purchase the instrument, it carries currency risk. Currency risk consists of uncertainty concerning official devaluation of the rupee, and depending on the source of foreign exchange, that of a change in the spread between the rates in the legal and parallel markets. Along with other domestic assets, it also carries the usual country risk.

Chart 1 illustrates the behavior over time of the nominal yield on FEBCs under the assumption of encashment one year after purchase. It also reports the “real” and “effective” one-year holding returns, defined as the nominal return minus the ex-post rates of change of the CPI and the official exchange rate, respectively, over the same month the preceding year. The series for one-year LIBOR is also shown. The instrument traded at a premium, hence the yield was below the coupon. Real and effective rates were low and often negative for extended periods of time. Holding the FEBC for one period would appear to be an unprofitable strategy according to both return concepts. 1/ Spreads over LIBOR also appeared consistently large and negative, suggesting incomplete arbitrage.

Chart 1FEBC Yield

(In percent per annum)

Taken at face value these results could be interpreted as indicating low sensitivity to local macroeconomic conditions and returns on alternative instruments. They would also imply the existence of a country premium as opposed to discount for Pakistan. There is evidence, however, of other factors at work. The instrument carries a premium over coupon in the local secondary market, and over the risk free rate in the international market, potentially as compensation for two functions that it performed. First, it was a source of legal foreign exchange in its own right, since it could be purchased on the stock market in rupees and cashed in foreign currency. Second, the absence of source disclosure requirements gave it a transfer of funds role from the informal market.

IV. A Reduced Form Model for FEBC Yield

The following model of FEBC yield is posited:

a reduced form for Rfebc is implied;

Definitions: R-nominal yield

  • superscripts rs, and $, denote rates on instruments

denominated in those respective currencies.

  • q-quantity traded of FEBCs on secondary market, in Rs
  • Π-inflation rate
  • S-Rupee/$ exchange rate
  • FX-excess demand for foreign exchange
  • T-demand for transfer services
  • t-time in months

FEBC demand derives primarily from a rupee investor community. The relevant price arguments of the demand function are therefore own-yield, and the yield on rupee denominated substitutes. In so far as real, not nominal returns matter, the inflation rate is explicitly included. The regulatory environment governing FEBCs and the financial system, indicate two quantity variables to be relevant determinants of demand behavior. Since the FEBC can be purchased in rupees on the secondary market and cashed in dollars, makes it an avenue for acquiring foreign currency and the relaxation of the private sector’s constraint on the acquisition of foreign exchange from the formal banking system. A term for excess demand for foreign exchange is therefore included. The absence of a source disclosure requirement in an otherwise regulated banking system, makes the instrument an avenue for transferring funds to the formal sector. A ‘transfer’ term is therefore also included. These regulation induced roles played by the FEBC have considerable anecdotal support.

FEBC supply is determined essentially by agents holding foreign currency, since it can only be purchased using foreign exchange. 1/ Supply of the instrument is perfectly elastic at the primary level, in that the government stands ready to satisfy all requests for purchase (hence the term “on tap”). At the secondary level supply depends clearly on own-yield (the secondary market premium), and the yield on dollar substitutes. The expected rate of official devaluation is included to render the rupee return on FEBCs ‘effective’ in the base foreign currency, which is assumed here to be US dollars. 1/

The model is specified in nominal terms in order to capture the different concepts of deflated returns which figure in demand and supply side optimization. The supply side is fundamentally a dollar economy, while the demand side is essentially a rupee one. The reduced form is a rate parity condition. In a regulation free and otherwise frictionless environment, it would reduce to a relation between the yield on FEBCs and the rate on a substitute asset. 2/

A priori expectations concerning sign of coefficients in the specified reduced form, are immediate. Noting that an increase in demand (supply) lowers (raises) the required yield, any variable which enters the demand (supply) curve positively, must enter the reduced form negatively (positively). 3/ Reduced form coefficients on the dollar return, expected devaluation, and foreign exchange and transfer demands, are therefore negative. The coefficient on the rupee return is positive. Because it enters the demand side in two ways, inflationary expectations have an ambiguous sign. In so far as expectations of high inflation imply lower anticipated real returns, it lowers demand and raises the yield. They also have a direct positive effect however, in that high inflation leads to substitution towards FEBCs and away from local assets. It should be noted that as a reduced form the model is general, and is unfortunately consistent with alternative theories. While it is clearly posited that the model is an appropriate description of the FEBC market, the empirical work below cannot be considered a test of this model against the null of alternatives. In fact the reduced form can be appropriately described as an uncovered interest parity condition, controlling for differences in asset characteristics between the FEBC, and the local and foreign substitutes. For those unimpressed by the theory above however, the empirical exercise and results below should still be of interest.

V. Empirical Results

1. Data

Monthly yields on FEBCs, from February 1986 to March 1993, were computed from prices reported for new issues on the Karachi stock exchange. These are secondary market prices inclusive of the, previously mentioned, premium above face value. 1/ For a rate on a rupee substitute, the money market, and government bond rates, reported in the International Financial Statistics (IFS) are used. The former is an interbank lending rate reflecting short-term liquidity movements. The latter is a composite yield series based on average prices quoted on the last Wednesday of the month for the 11 3/4 percent government bond due in 2002, and prior to 1989 for the 11 1/2 percent bond maturing in 1999. 2/ While these rates display considerable variability, they are still subject to the general regime of financial controls in the period under consideration. As the only freely functioning financial series over the relevant period, the stock market index was therefore also used as a proxy for rupee substitute assets. The series, reported in the IFS, refers to midday quotes for the last Friday of the month for 242 common shares on the Karachi stock exchange. For the rate on dollar substitutes, both the LIBOR and US T-bill rates were used for relevant maturities, as well as the rate on local foreign currency accounts reported in the State Bank of Pakistan Bulletin. Inflation and exchange rate changes were computed from the IFS series for the CPI, and the official rupee/dollar rate of exchange, respectively. 3/

A critical aspect of the exercise is finding appropriate proxies for the demands for excess foreign exchange, and transfer services. Both are inherently unobservable. For the former, a natural measure suggests itself in the information contained in foreign currency encashments of FEBCs. Cashing the instrument in foreign exchange may reflect an optimal currency choice on the part of market agents, and not necessarily, an attempt to relax a foreign exchange constraint. There is substantial evidence however, that these encashments often occur long before maturity, and often before the conclusion of the minimum one-year period required for receipt of Interest. This behavior is inconsistent with an original purchase of the instrument for its inherent investment value. There is also, as mentioned above, considerable anecdotal support to the contention that the instrument is used as a venue for foreign currency acquisition. The ratio of foreign currency encashments to total encashments, in percent, is taken therefore as a proxy for movements in FEBC demand associated with excess demand for foreign currency.

It is posited that the demand for transfer services is positively related to activity in the informal sector of the economy. The task is then to proxy for informal sector activity. A readily available measure used in other contexts is currency outside the formal banking sector. In so far as it describes disintermediation, whether due to a buoyant parallel market, shifts in confidence in the formal financial system, or other causal factors, it captures recycling of funds through the informal sector. The ratio of currency outside banks to M2 seasonally adjusted, In percent, In the form of deviations from trend, is taken therefore, as the relevant proxy. 1/ Both variables are reported in the IFS.

2. Estimation

The reduced form model estimated is represented by equation (3.3). The dependent variable is the one-year holding period return on FEBCs. In the interest of parsimony, only one local and one foreign currency interest rate will be included in the regression. None of the local rates of return except for the stock market index were found to be significant. This is not surprising given the administered nature of the underlying instruments. As a relatively free variable, the stock market index ought to capture adequately the notion of a local currency substitute to FEBCs, For a dollar rate of return, the one-year LIBOR is used. Inflation and devaluation expectations are assumed rational, and are taken as the output of an ARIMA forecasting model. 2/

The time series properties of model variables are examined in Table 1. Dickey-Fuller (DF) unit root tests are implemented in basic and augmented form to allow for the possibility of high order serial correlation. 3/ The null of nonstationarity is rejected by all statistics for the one-year FEBC yield. The null for the stock market index is accepted by the DF statistic against a trended alternative, and by both ADF statistics. This is evidence of nonstationarity. One-year LIBOR is unambiguously nonstationary. The nulls for inflation and devaluation expectations are rejected by all statistics except the ADF against trendless alternatives. This is taken as evidence of stationarity. Finally the nulls are strongly rejected by all statistics for the FX and transfer proxies. In short there is enough evidence of nonstationarity to require close attention to asymptotic degeneracy of the standard test statistics, and to the potential for a spurious regressions problem. To err on the side of caution all ambiguous time series processes will be treated as potentially nonstationary.

Table 1:Unit-Root Tests: Monthly Data 1986:1-1993:1
FebclIndexLiborInflationDevaluationFxTransfer
DF-18.39-16.721.12-15.66-15.12-31.14-23.72
DF(trend)-55.12-12.87-1.97-22.57-21.37-64.15-48.24
ADF-19.72(2)-3.58(1)0.57(2)-10.21(2)-9.85(1)-29.35(1)-31.12(1)
ADF(trend)-56.68(2)-4.62(1)-3.11(2)-24.62(2)-28.25(1)-85.36(1)-72.54(1)

Fx denoted the ratio of foreign currency encasement to total encasement of FEBCs in percent. Transfer denotes deviations from trend of the ratio of currency outside of banks to M2 seasonally adjusted. Inflation and devaluation are one month ahead forecasts from an ARIMA model.

DF is the Dickey-Fuller statistic. ADF is the augmented DF statistic. Parentheses indicate order of serial correlation allowed in the ADF test. The critical value for the DF/ADF statistic N(rho-1) is -13.7 and -20.7 for the constant/no trend, and constant/trend alternatives respectively, for the 5 percent significance level and N = 100 observations.

Fx denoted the ratio of foreign currency encasement to total encasement of FEBCs in percent. Transfer denotes deviations from trend of the ratio of currency outside of banks to M2 seasonally adjusted. Inflation and devaluation are one month ahead forecasts from an ARIMA model.

DF is the Dickey-Fuller statistic. ADF is the augmented DF statistic. Parentheses indicate order of serial correlation allowed in the ADF test. The critical value for the DF/ADF statistic N(rho-1) is -13.7 and -20.7 for the constant/no trend, and constant/trend alternatives respectively, for the 5 percent significance level and N = 100 observations.

The model is estimated on monthly data, for the period February 1986 to July 1991. Table 2 reports results. Regression 1 is the well specified model, without attention to nonstationarity. All coefficients have the a priori expected signs, and all are significant at the 1 percent level except for those on the stock market index, and expected devaluation. A one percentage point increase in the ratio of foreign currency FEBC encasement to total encasement, ceteris paribus, lowers the yield on FEBCs 3 basis points. 1/ A one percentage point increase in the ratio of currency outside banks to total money, above trend, lowers the yield on FEBCs 24 basis points. The coefficient on LIBOR is also large, indicating significant sensitivity to international rate developments. A one percent increase in LIBOR lowers the yield on FEBCs 35 basis points. Inflation expectations are highly significant and enter negatively. Demand for FEBCs appears to increases with inflation. The constant however, is significantly different from the theoretically implied value of zero, and is large. At any point in time 9.7 percent of the yield on FEBCs is explained by a time invariant constant. This must be viewed as a potential mis-specification problem.

Table 2:Dependent Variable is One-Year Holding Return on FEBCs
1.2.3.4.5.6.7.
Constant9.661

(6.030)
9.056

(5.770)
8.921

(6.109)
10.442

(6.774)
9.625

(5.874)
6.107

(6.219)
6.415

(5.214)
Index0.010

(1.662)
0.012

(1.954)
0.012

(2.008)
..

..
0.018

(1.687)
0.012

(2.001)
..

..
Libor-0.347

(-2.060)
-0.251

(-1.505)
-0.303

(-1.890)
-0.297

(-1.780)
-0.388

(-2.001)
..

..
-0.151

(-0.925)
Inflation-0.238

(-3.024)
-0.209

(-2.636)
-0.218

(-2.862)
-0.214

(-2.743)
-0.216

(-2.811)
-0.227

(-2.777)
..

..
Devaluation0.089

(1.645)
..

..
0.090

(1.696)
0.096

(1.771)
0.089

(1.641)
..

..
0.080

(1.418)
Fx-0.025

(-2.654)
-0.025

(-2.673)
-0.094

(-2.590)
-0.026

(-2.773)
-0.025

(-2.512)
..

..
..

..
Transfer-0.238

(-2.362)
-0.175

(-1.789)
..

..
-0.175

(1,758)
-0.189

(-1.857)
..

..
..

..
Indext-1..

..
..

..
..

..
..

..
-0.009

(-1.221)
....
Libort-1..

..
..

..
..

..
..

..
0.114

(1.514)
..

..
..

..
Adj R20.6010.5780.6000.5920.6310.3800.324
Durbin Watson1.8201.8281.9221.8771.9871,8921.928
Sample86:1-93:186:1-93:186:1-93:186:1-93:186:2-93:186:1-93:186:1-93:1
Period
NOBS85858585848585

T-statistics in parentheses.

Index is the Karachi stock exchange stock market index, LIBOR is the one year London Interbank Offer Rate, inflation and devaluation expectations are one period ahead forecasts from an ARIMA model, fx is the ratio of foreign currency to total encashments, transfer is deviations from trend in the ratio of currency outside of banks to M2 seasonally adjusted.

Data are end of month values from January 1986 to January 1993, from the State Bank of Pakistan monthly bulletin, and IFS.

T-statistics in parentheses.

Index is the Karachi stock exchange stock market index, LIBOR is the one year London Interbank Offer Rate, inflation and devaluation expectations are one period ahead forecasts from an ARIMA model, fx is the ratio of foreign currency to total encashments, transfer is deviations from trend in the ratio of currency outside of banks to M2 seasonally adjusted.

Data are end of month values from January 1986 to January 1993, from the State Bank of Pakistan monthly bulletin, and IFS.

Given the less than parsimonious nature of the model, considerable attention was paid to issues of specification. The cross-correlation matrix for the independent variables is computed as an indicator of multicollinearity. Table 3 reports results. All correlation coefficients, except those between transfer and index, transfer and libor, and index and libor, are below 0.5 in absolute value. The coefficient for inflation and devaluation is low at 0.13. This is due to the controlled nature of exchange rates, and the short time frame of the sample.

Table 3:Cross-Correlation Matrix
IndexLiborInflationDevaluationFxTransfer
Index1.00.6070.3790.2030.0080.743
Libor0.6071.0-0.0530.231-0,2300.564
Inflation0.379-0.0531.00.1330.2940.315
Devaluation0.2030.2310.1331.00.1720.239
Fx0.0088-0.2300.2940.1711.00.091
Transfer0.743-0.564-0.315-0.2390.0911.0

The model is estimated therefore with ‘transfer’ dropped, in order to estimate the index and libor coefficients efficiently. Results of this exercise are reported in regression 3 of Table 2. The index coefficient becomes highly significant, and increases in magnitude. Regression 4 shows output with the index variable dropped. Results are basically unaffected. For completeness, regression 2 shows the model estimated with devaluation dropped. Coefficient estimates appear therefore robust to specification changes. Equation 5 examines robustness of these results to the presence of potentially nonstationary right hand side variables. The first lags of the stock market index and LIBOR are added to the regression. This is technically equivalent to differencing without the imposition of unit root restrictions. The coefficients on both lagged variables are insignificantly different from zero and none of the other coefficient estimates are altered. To further rule out the possibility of spurious results, the residuals of equations 1 to 5 are themselves subjected to unit root tests. The output of the latter rejects the null of I(1) residuals with a high degree of confidence for all regressions. These findings are taken as evidence of the soundness of the time series exercise.

In order to interpret this output further, interest rate parity relations are estimated in equations 6 and 7 of Table 2. Regression 6 is a Fisher equation. The negative sign on the inflation expectations coefficient constitutes a failure of this relation, and confirms the FEBCs role as an ‘inflation tax’ haven. Equation 7 is an uncovered interest parity relation. Sensitivity of FEBC yield to the returns on LIBOR assets and expectations of official exchange rate devaluation is not sustained. There is therefore clear indication that correction for ‘regulation induced’ effects is critical. Controlling for asset characteristics, the one-year holding period yield on FEBCs is highly sensitive to innovations in returns on rupee and dollar assets, as well as inflation expectations. The well specified model also exhibits almost twice the explanatory power of these simple arbitrage relations. Priors concerning signs set out in the model are all confirmed. The coefficient on expected devaluation, however, is consistently insignificant at the 5 percent level, and inconsistent with priors concerning sign. 1/ In view of the high relative importance of the speculative effects represented by the regulation proxies, this result is not surprising. It is likely that the instrument is not held long enough for devaluation to figure in agents’ optimization. If the objective of FEBC suppliers is to simply reap the stock market premium, the instrument may be sold immediately after purchase at very low exchange risk. There is considerable anecdotal evidence, as mentioned above, of this type of behavior.

Structural stability of is examined by re-estimating the model over the subperiods 1986:1-1990:1, and 1990:1-1993:1. 2/Table 4 shows regression output. Priors are roughly confirmed, in that the process of reform appears to have increased FEBC yield sensitivity to the rates on alternative instruments, and reduced the importance of regulatory effects. The output of regression 1 represents the phase, where financial reform in progress, can be argued to have not yet fully filtered through to rates and prices. The regulatory premia are both highly significant, and have sizable impacts on FEBC yield. Equation 2 is estimated for the period in which financial reform Intensified. 1/ index and LIBOR which have insignificant partial effects in the early period, become highly significant in the second. The fx coefficient becomes statistically insignificant in the later period. The transfer coefficient however rises in magnitude and significance. This result is unexpected, A possible explanation is that informal market activity itself responded positively to reform, raising demand for transfer services. Its proxy therefore increases in importance.

Table 4:Dependent Variable is One-Year Holding Return on FEBCs
1.2.
Constant9.465

(3.250)
7.329

(5.813)
Index-0.006

(-0.289)
0.017(3.243)
Libor0.055

(0.175)
-0.066

(-1.870)
Inflation-0.432

(-3.131)
-0.221

(-2.872)
Devaluation0.127

(1.710)
-0.012

(-0.270)
Fx-0.030

(2.270)
-0.019

(-1.473)
Transfer-0.297

(2.197)
-0.313

(-2.277)
adj R20.6760.518
Durbin Watson1.5491.922
Sample86:1-90:190:1-93:1
Period
NOBS4937

T-statistics in parentheses.

T-statistics in parentheses.

These results, strictly interpreted, constitute a failure to adequately explain FEBC yield behavior. The latter is explained mainly by a time invariant constant. They remain suggestive however, in that all variables enter significantly with signs that conform to priors. Sensitivity to innovations in local and international rates of return, and to regulation induced asset characteristics is unambiguously sustained. There is reason to expect improvements in model performance with the financial reform program, and with better proxies.

VI. Financial Market Efficiency Revisited

The seemingly perverse picture of comparative returns exhibited by Chart 1 can now be re-examined. Taking the model at face value despite its shortcomings, a quantitative measure of the discount on the yield required to hold FEBCs, obtains from the coefficient estimates of equation 1 in Table 2. Applying the latter to the foreign exchange and transfer proxies, and adjusting the FEBC yield, a true yield series emerges. Chart 2 illustrates comparative rates of return after this correction. The conclusions drawn from this picture are more consistent with a priori expectations than those derived from Chart 1. The real and effective returns are on average positive, and often sizably so, episodes of exceptionally high inflation and devaluation notwithstanding. While only the nominal series appears consistently above LIBOR, there is some indication as noted previously, that devaluation risk is low, and that the latter is the appropriate indicator of the return that agents arbitrage against international interest rates. A visible upward trend is apparent since mid 1989 in the real and effective series reflecting the relaxation of financial controls. The country premium, a measure of which is given by the spread between FEBC yield and LIBOR has been on the rise since mid-1990. This is again indication of the repressed premium emerging. The 400 basis point spread exhibited in January 1993 is consistent with Pakistan’s comparative credit risk position among LDCs at that time. 2/

Chart 2FEBC Yield Adjusted

(In percent per annum)

VII. Conclusion

An equilibrium reduced form model for FEBC yield behavior is specified and estimated. Proxies for excess foreign exchange demand and transfer services are used to correct for regulation effects. The interest rate and proxy variables prove significant and conform in sign to theoretical priors. Positive changes in the stock market index raise the yield required to hold FEBCs. Positive changes in LIBOR, inflation expectations, excess foreign exchange demand, and demand for transfer services, lower the yield. Controlling for the FEBC's role in circumventing financial regulation, the yield is sensitive to the returns on substitute local and international assets. The model is structurally unstable. The process of financial liberalization increased the significance of the interest rate variables and decreased that of the foreign exchange proxy. A new FEBC yield series corrected for regulation discounts is constructed and shown to be competitive with LIBOR. The empirical exercise depends critically on the interpretation of the proxies. In view of the dearth of available measures for excess foreign exchange demand and informal sector activity, there are no easy solutions to this problem. Results are encouraging however in that they suggest the rationality of the rate structure underlying regulated interest rates in Pakistan. They indicate indirectly the sensitivity of the Pakistan public to arbitrage opportunities on local and international debt instruments. This sensitivity seems to have increased with reform.

Bibliography

    CoxJ.C.J.E.Ingersoll and S.A.Ross“An intertemporal General Equilibrium Model of Asset Prices”Econometrica. (1985a) vol.53 pp.363384.

    CumbyRobert E.“Is it Risk ? Explaining Deviations from Uncovered Interest Rate Parity”National Bureau of Economic Research Working Paper No. 2380:1 (September1987).

    • Search Google Scholar
    • Export Citation

    DickeyDavid and WayneFuller“Distribution of The Estimators for an Autoregressive Time Series With A Unit-Root”Journal of The American Statistical Association. June1979 pp.427431.

    • Search Google Scholar
    • Export Citation

    DornbuschR“Exchange Rate Risk and the Macroeconomics of Exchange Rate Determination”National Bureau of Economic Research Working Paper. No. 493 (June1980).

    • Search Google Scholar
    • Export Citation

    EdwardsS. and MohsinKhan“Interest Rate Determination in Developing Countries”IMF Staff Papers (September1985) vol.32 pp.377403.

    El-ErianMohamed A.“Currency Substitution in Egypt and the Yemen Arab Republic: a comparative quantitative analysis”IMF Staff Papers (March1988) vol.35 pp.85103.

    • Crossref
    • Search Google Scholar
    • Export Citation

    FamaEugene and RobertR. Bliss“The Information in Long Maturity Forward Rates”American Economic Review No. 77 (1987) pp.680692

    • Search Google Scholar
    • Export Citation

    FrenkelJacob and RichardLevich“Covered Interest Arbitrage:Unexploited Profits”Journal of Political Economy. Vol 48 (August1986) pp.213228.

    • Search Google Scholar
    • Export Citation

    KhorE. Hoe and LilianaRojas-Suarez“Interest Rates in Mexico: the role of exchange rate expectations and international credit worthiness”IMF Staff Papers (December1991) vol.38 pp.850871.

    • Crossref
    • Search Google Scholar
    • Export Citation

    MankiwN.G. and J.A.Miron“The Changing Behavior of The Term Structure of Interest Rates”Quarterly Journal of Economics. No.101 (1986) pp.221228.

    • Search Google Scholar
    • Export Citation

    ShabsighGiath“The Economic Implications of The Underground Economy”unpublished1992.

    ShillerRJ“The Volatility of Long Term Interest Rates and Expectations Models of The Term Structure”Journal of Political Economy. No.87 (1979) pp.11901219.

    • Crossref
    • Search Google Scholar
    • Export Citation

    VillanuevaDelano and AbbasMirakhor“Strategies for Financial Reforms: Interest Rate Policies, Stabilization, and Bank Supervision in Developing Countries”IMF Staff Papers (September1990).

    • Crossref
    • Search Google Scholar
    • Export Citation
1/The paper was written while the author was a summer intern in the Middle Eastern Department. I am grateful to members of the Middle Eastern Department and the participants of the Development Economics workshop at Harvard University. Special thanks to Kal Wajid, Mohamed El-Erian, and Antonio Furtado for critical comments and suggestions. Any remaining errors are my own.
1/The liquidity preference/preferred habitat theory is a hypothesis concerning the nature of the risk premium above. It relates the premium to quantities of different maturities of assets.
1/see Krasker (1980).
2/The theoretically implied model for a local interest rate is a rate arbitrage equation. The relevant prior is a significant positive relation to the rates on substitute assets, and high explanatory power.
1/The financial sector reform measures recently implemented involve a move towards indirect monetary control. This included removal of quantitative credit controls, liberalization of rates of return, and gradual removal of directed credit programs.
2/This discussion relies on various internal IMF literature.
1/The outliers in all figures are due to unusually large devaluations in relevant periods.
1/It is technically possible for the FEBC to be purchased in rupees at the primary level if agents enter the parallel (i.e., curb) currency market. There is some indication that arbitrage opportunities of this sort exist ex-post, but that they have not been large enough on average to justify transaction costs.
1/The model assumes that the exchange rate and inflation expectations are not jointly determined. The official exchange rate is tightly managed. While domestic inflation is a factor in setting the official exchange rate, there is less than complete correspondence between the two. In any event, results of estimation with one of the variables dropped are reported.
2/Direct data on quantity demanded and supplied proved unavailable. This rules out separate identification of supply and demand parameters.
3/As noted earlier, FEBCs have been trading at a premium over face value on the secondary market. An increase in the premium implies a lower yield by definition. The model and consequent analysis could have been expressed entirely in termas of the premium. Note also that supply of FEBCs is really “on tap” demand by the expatriate community for new government issues.
1/A 100 rupee face value FEBC has traded at an average 106 rupees, in the period under analysis.
2/The IFS recently switched to reporting Federal Investment Bond rates.
3/Both inflation and devaluation are rates of change over the same month the preceding year.
1/This specification is appropriate in that it only captures unusual movements of currency outside banks, around its long run downward trend. Work by Shabsigh on measurement of informal sector activity in Pakistan, broadly justifies the use of this proxy.
2/The residuals of the regression are corrected for the presence of forecasted independent variables.
3/The usual caveats about finite sample unit root tests apply. For any unit root process there exists a stationary process that will be indistinguishable from the unit root representation for any given sample size N. (for example rho close to but not equal to unity). In addition, Dickey-Fuller tests rely on the absence of moving average components in the tested time series. Because an MA(1) process has a finite order autoregressive representation, this problem can be alleviated by adding a higher order of AR parameters than would be implied by a pure AR process.
1/Alternatively, it raises the secondary market premium, above face value, by the same amount.
1/Devaluation expectations were also taken as the parallel market exchange rate, quoted in Dubai. Results are robust to this change. The official exchange rate is the only exchange rate series that was obtainable for the full period under analysis.
2/The split point was chosen in order to capture the period where financial reform was most intensive, namely from 1990 onwards. During this period government debt auctions were implemented, credit ceilings were abolished in favor of credit/deposit ratios, and residency requirements and other foreign exchange regulations were reviewed.
1/A Chow test rejects the null of structural stability at the 1% level. The unrestricted model is equation 1 in Table 2.
2/See Salomon Bros. “Emergin Markets”. Pakistan’s Institutional Investor creditworthiness rating was 27.7 in 1985 and 30.0 in 1990 out of a maximum possible 100 points. Pakistan ranked 66 and 60 in those years respectively in a sample of 112 countries. This rating index is based on polls of international bankers. For comparison, the Philippines with a comparable credit rating of 25.9 in 1990, had average foreign currency deposit rates 600 basis points above LIBOR in the early 1990s.

Other Resources Citing This Publication