FORECASTING INTEREST RATES WITH SPOT MATURITY YIELD SPREADS: HONG KONG S ADMINISTERED DEPOSIT RATES

Transcription

1 FORECASTING INTEREST RATES WITH SPOT MATURITY YIELD SPREADS: HONG KONG S ADMINISTERED DEPOSIT RATES Robert Haney Scott, Richard PonArul, Suleman Moosa California State University, Chico ABSTRACT In this paper we examine the data on deposit rates set by the Hong Kong Association of Banks to see if there is any forecasting content in the spread between short-term and longer term interest rates on bank deposits in Hong Kong over the period from By forecasting content, we mean whether a future change in an interest rate can be forecast with any degree of confidence from a set of currently observable prices. We examine only within-sample data. The period chosen covers the life of the Interest Rate Agreement by the Hong Kong Association of Banks under which the association set rates of interest that members should pay to deposits with various maturities within one year. We find statistically significant coefficients in only two of the ten regressions that we estimated. Thus, our data do show that there was some information regarding forward changes in rates from yield differentials administered by the cartel. One possible explanation is that the cartel had better information than the market regarding its future course of action but chose not to reveal this information through administered rates. In concluding, we reflect briefly on this interesting line of research and its promise in the field of financial analysis in future years. INTRODUCTION We begin with reference to the literature in finance on the use of market interest rates to predict future events. Next we proceed directly to a description of our data source and follow this with a presentation of our findings. In conclusion we revert to brief commentary on this popular line of research within the field of finance. A POPULAR LINE OF RESEARCH IN FINANCE In a recent study of the relation of municipal bond yields and tax law changes by Fortune [3, p. 29] he writes, That security prices and yields reflect anticipation [by investors] of future events is an axiom of financial economics. An interesting question is: Can data that reflect market anticipations also be used with confidence to predict future values of some important variable? In Fortune s article the event the author is interested in predicting is not some future interest rate change, but rather a change in tax law under debate in Congress! The tax law in question related to municipal bonds. The interest income on those bonds is excluded from taxable income according to federal income tax regulations. Will the Congress abolish the exclusion? According to the author you might find the answer if you follow carefully the spread between the yield on Treasury securities and those on municipal securities. Since municipal securities yield less because income from them is exempt from federal taxation, investors who buy them pay an implicit tax by accepting a lower yield. The data seem to show that the size of this implicit tax can be useful as a variable in partially forecasting whether the Congress will change this law. In an earlier study, Page and Sirmans [8] examined yield differences between fixed and adjustable rate mortgages and used the spread between one-year and ten-year Treasury bond yields as a proxy for an interest rate premium. They were interested in determining whether the two types of mortgages had different default risk premiums, and they found such evidence. In a previous study, we undertook to see if there were any information contained in the spread between interest rates on fixed rate mortgages (FRM) and interest rates on adjustable rate mortgages (ARM). [9]. The dependent variable to be explained by this spread in rates was the change in the one-year Treasury bill rate over a three-month period. We also examined the change in the one-year rate over six-month and one-year periods. Regressions indicated that during the period from 1981 through 1995, some 15 per cent of the variation in one-year T-bill rates over a 3-month period was explained by the spread between the FRM rate and the ARM rate, that is, FRM rate minus ARM rate

2 The t-value of the coefficient of the spread was 5.7. The percentage of variation explained was successively lower as the period extended from 3-months to 6-months, and the t-values fell, but remained significant at 3.0 even for the 6-month period. The theory is that a large spread between FRM and ARM rates will reflect the expectation of higher interest rate levels in the coming months. Why? If interest rates are expected to rise, then the demand for fixed rate mortgages will be high because home buyers will want to lock in a fixed rate for a long period and will not be attracted by ARMs in fear of rates on those mortgages rising. (The opposite analysis applies if the spread is narrow.) Thus, the expectation of rising rates, as reflected in the spread, presumably provides relevant information about the future path of interest rates over time. The spread becomes a forecasting variable with some, albeit small, forecasting power. At least that seems to be what the data have revealed. It is suggested that forecasting models of interest rates might usefully include the spread as an argument in a forecasting equation. In a very interesting study Kenneth Kasa [7] focused on the political event of the turnover of Hong Kong to China at midnight June 30, He used the differential between the interest rate on three-year government bonds in the U.S. and that on the same maturity of government bond in Hong Kong. Since 1984 this differential averaged about 50 basis points. The pegged exchange rate between the HK dollar and the US dollar requires that interest rates remain the same in the two places in order to prevent arbitrage. He states that such bond prices implicitly embody forecasts of the future. And since Hong Kong was to revert to Chinese sovereignty, any fear about Hong Kong not honoring its debt obligations in US dollar equivalent money after the handover would be reflected in the yield differences. No significant change in the yield differences was noticed, and the conclusion was that investors think things will go on pretty much as usual on July 1, In the event, Kasa s forecast was correct and continued to be correct for four months until the slump in the stock market in October DATA ON DEPOSIT RATES ADMINISTERED BY THE HONG KONG ASSOCIATION OF BANKS This study concerns the administration of a cartel and its decisions to change prices of certain financial products. In this sense, our study is also a study of events and whether yield spreads can predict them. The cartel in question is the Hong Kong Association of Banks, and the product price is the interest that the administrators permit members of the cartel to pay to depositors on deposits of various maturities. Laws established in the former British Colony of Hong Kong are ordinances, and the Hong Kong Association of Banks Ordinance, Section 12, provides that the association s committee,... after consultation with the Financial Secretary, shall determine from time to time... the maximum rate of Interest payable, or minimum rates of Deposit Charge that must be levied, by all Members on those Deposits to which the Rules apply in respect of money held on:... current account, savings account; deposits fixed for seven days or less; deposits fixed for one week, two weeks; one, two, three, six, nine, and twelve months or more but less than 15 months. The interested reader might find it strange that the committee, besides being given authority to determine the rate payable by member banks, was also given authority to establish a minimum deposit charge that member banks would have to levy on deposits. A charge on deposits could effectively mean not only a negative real return on deposits but a negative nominal after tax return as well. This provision was part of an amendment to the Ordinance made on recommendation of the Exchange Fund when the pegged exchange rate system was under pressure. In 1989, because of low interest rates in the United States and weakness of the US dollar on foreign exchange markets, interest rates on Hong Kong dollar deposits fell to below one percent. The purpose of this negative interest rate scheme was to discourage people from selling US dollars and buying Hong Kong dollars to put on deposit in Hong Kong banks. Such selling might otherwise put upward pressure on the value of the Hong Kong dollar in the foreign exchange markets and jeopardize the pegged rate system. Some economists recommended that the Hong Kong dollar be allowed to appreciate. For a review of this matter, see Greenwood, Ho, and Law, [4]. The authority has not been used, but remains on the books. In the 1970s the Swiss imposed a similar tax provision to be levied on foreign deposits if the franc were under pressure to depreciate. [5, p. 9]

3 The data readily available from the Hong Kong Association of Banks cover the period from 1978 through However, prior to 1982, maximum rates were set on savings accounts, 7 Days Call paper, and 3, 6, and 12 month deposits. It was not until 1982 that rates were also set on 24 Hr. Call deposits and 1 week, 2 week, 1 month, and 2 months deposits as well. After 12 years of sometimes controversial setting of deposit rates there began a phasing-out of the administered rate program because the Legislative Council was under pressure from the Consumer Council to restrict the monopoly power of banking institutions in Hong Kong. Data indicate that the cartel stopped setting rates on deposits with maturities of more that one month in September In January 1995, it stopped setting rates on two week and one month deposits. In February 1995, it ceased setting deposit rates on one week and 7 days Call deposits but continued to set them on time deposits of 6 days or less, 24 hour Call deposits, savings deposits, and on current accounts. The phasing out of deposit rate setting was not fully completed at that time because banks lobbied the Legislative Council for authority to continue to set rates on savings deposits and 24 hour Call deposits. Lobbyists raised fears that smaller banks might be driven into financial difficulty if their cost of funds were to rise after the controls were removed. This is reminiscent of the anachronistic fear of cut-throat competition for deposits by large financial institutions prevalent in the U.S. until the 1980s under the Federal Reserve s Regulation Q. Thus, the Hong Kong Association of Banks continues to set the shortest deposit rates. On August 30, 1999 reporters for The Wall Street Journal noted that the association... raised its short-term deposit rate 0.25 percentage point to 3.75% to match a U.S. rate increase earlier in the week. That had all the major banks raising their prime lending rates a quarter percentage point, to 8.5%. [12] The association keeps its presence known. However, the end may be near for deposit rate regulations. A consultancy study commissioned in 1998 for the Hong Kong Monetary Authority (HKMA) has recommended that the remaining interest rates be deregulated. The August 1999 issue of the Quarterly Bulletin published by the HKMA reports that, The HKMA, along with the banking industry, considers that deregulation of the remaining [interest rate regulations] is inevitable... [6, p. 33] However, the HKMA recommends that the remaining deregulation be phased in so that banks and their customers can become better informed and prepared. Periodic regulatory changes caused discontinuities in the stream of observations for our empirical study. Thus, to have a complete and consistent set of observations the data used for paper include only those available beginning in May 1982 and ending in August We used rates set on 1, 2, 3, 6, 9, and 12-month deposits. FORMULATION AND RESULTS OF EMPIRICAL TESTS Let 1 r t be the one-month rate today, and 2 r t be the two-month rate, so that today s spread between one-month and two-month deposits rates equals ( 2 r t - 1 r t ). The spread is the independent variable in our regression equations. It is a predetermined variable because it was set by the cartel at the last meeting of the cartel s committee. Our question is: Does this spread predict the change in 1 r t from the last meeting of the cartel s committee to its current meeting? Thus, the regression is: 1 r = ( 1 r t+1-1r t ) = α + β( 2 r t - 1 r t ) + ε (1) The above formulation is typically useful for rates that are set in a competitive market with the zero arbitrage condition. It is well known that the debt-capital markets are not well-developed in Hong Kong and that bank financing is a principal source of capital. Here the well-heeled (bank) investors with sufficient funds to move markets get together in their cartel to decide on the structure of rates on deposit maturities. If these dominant investors formulate expectations about future values of short-term rates and set the current set of rates for different maturities working as a cartel, the spread between the short and longer rates should imply an expectation about a movement in future short-term rate. And if the meetings occurred at one month intervals, the form of the above regression would reflect the pure expectations approach to the term structure for one and two month maturities. Of course, the level of interest rates set by a cartel is not expected to have a relation to rates set in a competitive market. However, we claim that the relation between interest rates -- one-month and two-month and so on -- will be similar to what is seen in a competitive market

4 First we assume that the monopsonist (cartel) knows (or has some information on) what one-month rate it will pay next month. We conclude that when the cartel sets the two-month rate, it would take into account its knowledge of the expected one-month rate one month later. We ignore term premiums. Let us consider the scenario when the one-month rate is expected to go up next month. If the cartel sets the two month rate to be the same as (or less than) the one-month rate it will be offering the two-month depositors a second month deposit at a rate below the monopsony price. Hence the cartel s two-month rate will be set higher than the one-month rate. Expected increases in interest rates should thus make the yield curve upward sloping. Similarly, when the cartel expects the one-month rate to decline the yield curve is downward sloping and the interest paid on two-month deposits will be below the one-month rate. Thus, we believe that even when the level of monopsony interest rates differs from competitive rates, the relation rates on different maturities will be similar for monopsony and competition. Since cartel meetings routinely occurred more frequently than at one-month intervals, the data do not match their theoretical definition precisely. Results in Table 1. The results of our first regression are as follows: 1 r = ( 1 r t+1-1r t ) = α + β( 2 r t - 1 r t ) + ε (1a) α = β = R 2 =.0035 (0.107) ( ) Within this same structure, the independent variable was changed in each of five regressions. As shown, the first regression used the spread between the 2 mo. and 1 mo. rates as the independent variable. The second used the spread between the 3 mo. and 1 mo. rates. Following regressions used the spread between the 6 mo. and 1 mo. rates, and the 9 mo. and 1 mo. rates, and finally the spread between the 12 mo. and 1 mo. rates. Results are provided in Table 1. The expectations approach to the term structure of interest rates suggests that changes in the one-month rate could be forecast by the spread between the one and two- month rates. However, note that t values for the coefficients increase as the horizon over which the spread increases is lengthened. Changes in the 1 mo. rate can be forecast with some degree of confidence by the spread between the 9 mo. and 1 mo. rates and also by the spread between the 12 mo. and 1 mo. rates according to the significance of the t-values of the coefficients shown in the table. However, the very low R 2 coefficient suggests that a very low percentage of variation is explained

5 Table 1. Regressions to Forecast the Change in the One-month Deposit Rate Intercept Slope Spread R 2 (1) 1m = (2m - 1m).0035 (-1.072) (.538) (2) 1m = (3m - 1m).0139 (1.386) (1.075) (3) 1m = (6m - 1m).0287 (-.184) (1.556) (4) 1m = (9m - 1m).0547 (-2.342) (2.177) (5) 1m = (12m - 1m).0585 (-2.405) (2.256) Key: 1m is the rate of interest set on deposits of one month, 2m is the rate set on deposits of two months, etc. 1m is (1m t+1-1m t ). The spread between one and two month rates is (2m - 1m) etc. (t-values in parentheses.) That 1 mo. ahead changes in interest rate are not predicted by the spread between 2 mo. and 1 mo. rates, but are predicted by the 12 mo. to 1 mo. spread, may indicate either that the cartel is systematically slow in responding to expected changes in interest rates or slow in forming expectations about future changes in interest rates. We have no model to explain either scenario. Results in Table 2. We also have five regressions with the same independent variables but with the change in the 12 mo. rate as the dependent variable in each case. It this case the t statistic was highest using the 12 mo. - 1 mo. spread. The results are shown in Table 2. Table 2. Regressions to Forecast the Change in the 12-month Deposit Rate Intercept Slope Spread R 2 (1) 12m = (2m - 1m).0001 ( -.645) (.080) (2) 12m = (3m - 1m).0018 (-.646) + (.384) (3) 12m = (6m - 1m).0003 (-.224) + (-.152) (4) 12m = (9m - 1m).0051 (1.911) (.647) (5) 12m = (12m - 1m).0125 (-1.213) (1.011) Key: 1m is the rate of interest set on deposits of one month, 2m is the rate set on deposits of two months, etc. 12m is (12m t+1-12m t ). The spread between one and two month rates is (2m - 1m) etc. (t-values in parentheses.) In the regressions of Table 2 the change in the 12-mo. deposit rate is the dependent variable. Spreads between the 1 mo. and the 3 mo., 6 mo., and 9 mo. rates do not seem predict changes in the 12-mo. rate. However, a positive 12 mo.-1 mo. spread leads to an increase in the 12 mo. deposit rate. The t-values for the 12 mo. -1 mo. spread take a value of and this suggests an 85 per cent confidence that the beta coefficient is greater than zero under a one

6 tail test. This very mildly supports the recent use of the term structure as a leading indicator of business fluctuations because upward steepness predicts higher interest rates and tighter monetary conditions. Nevertheless, all the relations reported here have very low correlations. COMMENTARY ON FORECASTING WITH CHANGES IN INTEREST RATE SPREADS In this paper we examined an historical set of data to estimate forecasting content in the spread between short-term and longer term interest rates on bank deposits of various maturities set by the Hong Kong Association of Banks from Whereas we are dealing with within-one-year securities in this study, other studies take a similar approach using rate spreads on capital market instruments and mortgages. Now we glance briefly at unresolved issues. Unresolved issues. We mentioned the study of the political event of the turnover of Hong Kong to China by Kenneth Kasa [7] earlier in this paper. An issue not analyzed in his study concerns the inflation rate in Hong Kong. It ranged around seven per cent over the period covered by his chart while the rate in the US was only three per cent or even less. According to the International Fisher Effect, interest rate differentials should equal inflation differentials. So why would not the expected differential in the respective three-year rates be more like 400 basis points? Because it is only 50 basis points, Kasa contended that the differential indicated overwhelming confidence in the outlook for Hong Kong under China s sovereignty. Can we really use such a differential to infer something about political confidence? The differential is a descriptive variable. It may be a measure of political confidence. But there is something more important afoot. The scientific question is: Should we expect to find this differential to be useful in predicting the future, especially since political confidence is such an ethereal concept? What is the model that the study is testing? What implications can we draw from it and what variable is being predicted? These are important questions, too. Recent research on the information content of yield structures and the structure of term premia along the maturity continuum reveals that economists have broadened their outlook over many years of statistical effort to obtain some promising and many frustrating empirical results. (See Abken, [1] [2], for definitive studies.) Today, not only term structure spreads, but various other types of differences or spreads in spot markets are used to provide data in calculations designed to uncover predictive information. It is no longer thought necessary to try to isolate influences and abstract from differences that might affect and distort empirical results. With powerful computers available, researchers mine the overwhelming supply of existing data to see if, empiricistically, there is information contained in yield structures. But the ostensible purpose of scientific investigation is not to observe what market participants expect, but rather it is to enable human beings to predict events under specified conditions. Predictions need to be in conditional form: If..., then.... It is not clear just how the new and popular research efforts in finance will evolve into models useful in scientific prediction. CONCLUSIONS We examined the deposit rates set by the Hong Kong Association of Banks to see if the term structure set by the cartel conforms to the expectations hypothesis. Since members of a cartel compete among themselves as well as operating as a monopolistic entity in a broader economy, cartel rate-setting should reflect some aspects of competitive markets. While the aggregate of information provided by competitive markets cannot be replicated in a cartel, its pricing is conditioned on the information it has about market forces. We found that the spread between one-month and two-month rates did not forecast the change in the one-month rate. However changes in the one month rate can be forecast with some degree of confidence by the spread between the 9 mo. and 1 mo. rates, and also by the spread between the 12 mo. and 1 mo. rates. It should be noted that R 2 coefficients are very low. We also found that changes in the 12 mo. rate were not predicted by the five spreads used as independent variables except for a small measure of predictability by the 12 mo. to 1 mo. spread

7 REFERENCES [1] Abken, Peter A., Innovations in Modeling the Term Structure of Interest Rates, Economic Review, Federal Reserve Bank of Atlanta, July/August 1990, pp Also found in Financial Derivatives: New Instruments and Their Uses, Federal Reserve Bank of Atlanta, [2] Using Eurodollar Futures Options: Gauging the Market s View of Interest Rate Movements, Economic Review, Federal Reserve Bank of Atlanta, March/April, 1995, pp [3] Fortune, Peter, Do Municipal Bond Yields Forecast Tax Policy? New England Economic Review, Federal Reserve Bank of Boston, Sept./Oct. 1996, pp [4] Greenwood, John G., Ho, Y. K., and, Law, C. K., Forum on the Negative Interest Rate Scheme, Hong Kong Economic Papers, Hong Kong Economic Association, No. 19, 1989, pp [5] Hong Kong Association of Banks, Rules on Interest Rates and Deposit Charges, 1st December [6] Hong Kong Monetary Authority, HKMA s Response to the Banking Sector Consultancy Study, Quarterly Bulletin, August 1999, pp [7] Kasa, Kenneth, Post-1997 Hong Kong: A View from the Financial Markets, Federal Reserve Bank of San Francisco, Economic Letter, November 29, [8] Page, Daniel E. and. Sirmans, D. F., Yield Differences on Fixed and Adjustable Rate Mortgages, Quarterly Review of Economics and Business, Autumn 1984, pp [9] PonArul, Richard, and Scott, Robert Haney, Does the Spread Between Fixed and Adjustable Mortgage Rates Predict Changes in the One Year T-Bill Rate? Euro-Asia Journal of Management, Vol. 13, Special Issue, Euro-Asia Business Studies Center, University of Macau, 1996, pp [10] Scott, Robert Haney, On the Forecasting Content of the Implied Forward Rate, Managerial Finance, Volume 16, Number 5, 1990, pp [11] Thorton, Daniel L., Does the Federal Funds Futures Market Predict Fed. Behavior? Monetary Trends, Federal Reserve Bank of St Louis, December 1996, p.1. [12] Timmermann, Jeffrey, and Van Zant, Eric, Hong Kong Economy Exhibits Surprising, Inexplicable Strength, The Wall Street Journal, August 30, 1999, p. A