RDP 9506: The Liberalisation and Integration of Domestic Financial Markets in Western Pacific Economies 6. A Discussion on Domestic Integration

Even a glance at the graphs of domestic money, deposit and lending interest rates set out in Appendix 2 shows that rates in Australia, Canada, Japan, Malaysia, the Philippines, Singapore, Taiwan and the US are closely linked to each other. Rates in Indonesia and Thailand, on the other hand, were not well-linked for most of the period but have become more so in the 1990s, especially in Thailand. Posted institutional rates in Korea are barely linked with money market interest rates^[13], but curb loan rates do tend to follow money market developments. This pattern is reflected in the correlations between money and institutional rates, the nature of the equilibrium relationships between rates, and the speed of adjustment of institutional rates to changes in money market rates.

The average correlation between both deposit and loan rates and money market rates across both time and the Western Pacific region is substantially below one, indicating that shocks, foreign or otherwise, which affect domestic money rates are not immediately transmitted in full to the rates on non-traded domestic financial assets. This can be explained by the static nature of correlation analysis, the existence of interest rate ceilings or controls, imperfect competition in the banking sector, by shocks which only affect part rather than the whole of the term structure, or by implicit contracts between financial intermediaries and their customers to smooth retail deposit and loan rates either over the cycle or during periods of volatility in money markets.

On the other hand, correlation co-efficients, error-correction and the time taken to complete adjustment differ substantially by country and over time. By country, correlations are higher and the adjustment to equilibrium is relatively fast in Australia, Canada, Japan, Malaysia (deposit rates only), the Philippines and the United States. Adjustment is slower in Hong Kong, where deposit and loan rates are set by an officially sanctioned bank cartel (HKBA), and in Singapore, where branching and foreign competition are tightly controlled. Interestingly, the result that the speed of adjustment of Japanese institutional rates is similar to that of other non-cartelised banking systems suggests that its ‘main bank’ system does not generate abnormal behaviour in institutional rates. While cross-country comparisons are difficult to make, since the maturity profile of the instruments differs across countries, correlations tend to be negative or zero and adjustment negligible when countries set retail rates in a way which does not conform with market rates. In Korea, for example, the correlation between changes in money and deposit rates was significantly negative in the early 1980s, and not different from zero otherwise. Correlations for Thai and Indonesian rates are also not different from zero.

By period, correlations and the speed of adjustment to equilibrium have increased for Australia^[14], Canada, Indonesia, Japan, Malaysia and the Philippines. In Malaysia in the first half of the 1980s, for example, a third of the expected adjustment of the deposit rate to a rise in the money market rate had taken place by the end of the first month after the rise, but, by a decade later, this had risen to three-quarters of the expected adjustment. Moreover, while there were no equilibrium relationship between rates in Indonesia and Thailand during the 1970s or 1980s, such a relationship emerged in the 1990s. All this points to substantial and increasing integration of domestic financial markets. In Singapore and Taiwan, on the other hand, equilibrium adjustment appears to have slowed, and the adjustment of the loan rate has slowed in the US. In the case of Singapore, this is partly due to the insulating domestic institutional rates through 1988, 1989 and 1990 from money market rates. Excluding this period, the co-movement of rates rises slightly. The deterioration in the case of Taiwan may be due to an increase in the volatility of money market rates in the 1990s,^[15] while the slowdown in the adjustment of the US prime rate may reflect the decreasing importance of that rate for pricing bank loans. Nonetheless, there has been a notable increase in the co-movement of institutional rates with money market rates in most Western Pacific economies over the past decade, largely attributable to deregulation and a greater focus on competition in the banking sector.

It is worth noting that regulation and control per se are only impediments to domestic rate integration if they are not market-conforming. Japan and Taiwan are cases in point. While deposit rate liberalisation started in Japan in 1985 and was only completed in 1994, the rate was based on the CD rate, and so moved fairly closely with interbank rates. On the other hand, the margin between the deposit rate and CD rate only narrowed after deregulation, suggesting that the aim of regulation was to subsidise the cost of bank funds. Taiwan's deposit rates also seem to have been set with money market developments in mind. The story is less clear with loan rates. In Japan's case, until 1989 the short-term prime rate was set with respect to the ODR, below the money market rate, which is in violation of a market model of loan rate determination. This rate was formally liberalised in January 1989, but informal practices ensured that it initially remained relatively inflexible: banks met considerable borrower resistance in trying to implement a market-based lending rate when the rate was first liberalised and were forced to forego the requirement that borrowers place compensatory balances with them (so that the effective cost of borrowing rate was less affected). At the same time, risk of default increased markedly in 1992 as the economy deteriorated and the number of bankruptcies jumped, and the loan-call rate spread widened. The speed of adjustment also increased in the 1990s, especially when compared to the second half of the 1980s. One further reason why the deposit rate may have conformed more to the market than the loan rate is that Japan, like some other East-Asian countries, consistently sought to maintain positive real rates of interest on financial assets in order to promote saving, while at the same time trying to subsidise industry with cheap credit.

More generally, there is an apparent difference between the adjustment process of deposit rates and loan rates. Correlations tend to be higher, the adjustment process simpler, and the adjustment to equilibrium faster in deposit markets than in loan markets. This is most apparent in Indonesia, Malaysia and Thailand, but it also occurs in other countries. The fact that it occurs generally is consistent with a number of hypotheses^[16]: firstly, the maturity matching with money market rates is more precise with deposit rates than with lending rates; secondly, the smoothing of rates under implicit contracts is more important in the loan market than the deposit market, since borrowers may be more concerned with fixing costs than depositors are with fixing income; thirdly, it is easier and less expensive for a depositor to change accounts or financial intermediary than it is for a borrower, which means that arbitrage between deposit rates will be faster than for lending rates for any given change in market rates (see Lowe and Rohling (1992)).

The fact that the difference is most pronounced in Indonesia, Malaysia and Thailand suggests that it is difficult to enforce controls when substitutes to the controlled instruments can be readily created. In South-East Asian countries, there have usually been close substitutes for domestic currency denominated deposits, either in the form of foreign currency (US$) deposits at local banks or access to off-shore foreign currency deposits, for example in Singapore. If the authorities hold deposit rates below the ‘market’ rate, they risk hollowing out the banking sector and increasing the volatility and size of capital flows and exchange rate fluctuations. Substitutability is typically greater for deposits than loans, given the additional contracting costs and information asymmetries in the loan market, and this implies a lower adjustment coefficient on loan rates. However, this is not tenable over time if the banking system is to be stable. Moreover, one would expect that the ability of the authorities to insulate the domestic market has declined over the past decade as capital inflows to South-East Asia have increased (so the range of foreign substitutes for loans has increased) and as domestic capital markets have grown apace (so the range of domestic substitutes for both deposits and loans has expanded).

As discussed in Section 2, banks in most countries in the region have experienced periodic deterioration in the quality of their assets. The loan pricing rule for a free market predicts that this forces the loan rate up, and the positive constant in the loan rate equation should rise, which is what actually occurs. The constant term in the Philippine loan equation is positive, relatively large and statistically significant in all sub-periods, as expected. The constant term increases substantially in the Japanese, Canadian and Australian loan rate equations in the 90M1–94M12 sub-period, coincident with a substantial rise in business risk and non-performing loans (Okina and Sakuraba 1993; Lowe 1995). Indonesia has also experienced severe problems with non-performing loans, but the constant term for 90M1–94M12, while positive, is not significant, and this may indicate that the pricing of risk and the recovery of the banking system is being effected through non-market mechanisms (such as central bank bailouts). The constant term in the Thai loan equation is very large, positive and significant in the 80M1–84M12 sub-period, reflecting the series of banking crises from 1983 to 1986. The constant subsequently falls over successive periods, following substantial and effective reform of bank supervision, though it is still statistically significant. The constant is positive, significant and mostly stable in Singapore and the US over the sub-periods.^[17] The constant is positive but insignificant in Malaysia in 90M1–94M12, apparently reflecting sound banking practices and effective supervision in that country.

There are two additional issues to be considered. The first is whether foreign rates should be included in the institutional rate estimating equation. Consider the loan rate. The simple bank balance sheet from which the rules for institutional rates are derived ignores foreign liabilities, which can be important sources of funds when the capital account is open. This is unlikely, however, to be important in practice, since banks typically cover foreign currency borrowing in the forward exchange market, and so the cost of foreign funds will be the same as the money market rate when covered interest parity holds.^[18] In regard to the deposit rate, on the other hand, it is reasonable to think of the foreign interest rate as a determinant of the domestic deposit rate when foreign currency deposits are close substitutes for home currency denominated deposits, as in Indonesia and Malaysia. The foreign interest rate, however, is not significant in deposit rate equations.

The second issue is the relevance of the other institutional rate in the determination of institutional rates, discussed above in Section 4. If institutional rates are determined by fiat, then the money market rate and both institutional rates may contain information about the process for each of the institutional rates, whereas if they are market-determined, only the money market rate and the particular institutional rate under consideration should contain information about its process. Using data for Thailand and Australia, regressions which included the money market rate and both institutional rates were conducted for each country for the last two sub-sample periods, 85M1–89M12 and 90M1–94M12. The countries and periods were selected on the grounds that institutional rates were liberalised in both periods in Australia, but only the second period in Thailand, and so implicitly provide a test of the model.^[19] The estimation proceeded as follows. The lagged level of the other institutional interest rate and its lagged first-difference were added to the preferred equation estimated using equation (16), which effectively nests the hypothesis that the regulatory regime makes a difference to the determination of institutional interest rates. Table 5 reports the chi-square (2) statistic for excluding the other institutional rate.

The results are relatively straightforward. For Australia, it is unambiguous that, in both periods, including the deposit rate in the loan rate equation provides no additional information and vice versa. For Thailand, the loan rate (in this case, the lagged level) did provide information about the deposit rate before deregulation but not afterwards. On the other hand, the deposit rate provides information about the loan rate in the 1990s, though in this case the first lag of the differenced deposit rate provides information, and not the lagged level, which indicates the information concerns dynamics and not fundamentals. Moreover, the loan rate was only liberalised in mid 1992 and so the results are the average of two regimes.^[20] Overall, the evidence suggests that the rules are reasonable first approximations of institutional rate determination. Australian institutional rates are market-determined, as expected, and Thai deposit rates have been market determined this decade, although the evidence is less clear for Thai loan rates. The impact of money market rates on institutional interest rates has increased.

Footnotes

The Korean institutional interest rates analysed in this paper have been liberalised, but other rates are yet to be deregulated. The published figures do not report actual market rates. [13]

Note that the adjustment of the deposit rate to the bill rate slows down in the 85M1–89M12 sub-period. As official interest rates rose during 1985 and 1986, savings banks, whose assets chiefly comprised housing loans, became constrained by the 13.5 per cent ceiling on loans for owner-occupied housing. They responded to tighter margins by rationing housing credit and, taking advantage of price-making power in the deposit market, by limiting the rise in deposit rates, thereby driving a wedge between deposit and bill rates. The impasse was broken by providing special subsidies to savings banks and the liberalisation of interest rates on new loans for owner-occupied housing in April 1986, after which the wedge between deposit and bill rates narrowed. When the regression excludes this period, and is run from 87M1 to 89M12, β₁ is 0.48 and β₂ is 0.47, implying that 48 per cent of adjustment is completed after 1 month, 93 per cent after 4 months, and 100 per cent after 12 months. This is considerably faster than the 80M1–84M12 period but not as fast as the 90M1–94M12 period. [14]

The money market rate used in this case is the average of call rates across all maturities, which is dominated by very short call transactions. Since the term structure of money and institutional rates is not well matched, an increase in short-lived shocks to the money market rate will depress the adjustment coefficients. Other money market interest rate data, such as the 1 to 90 day NCD secondary market rate exhibits a similar pattern. [15]

In any sub-sample, the adjustment of deposit rates will be faster than that of loan rates if the deposit market is liberalised ahead of the loan market, as occurred in Malaysia (October 1978 compared to February 1991) and Thailand (March 1990 compared to June 1992). The 90M1–94M12 sub-sample for the loan rate is re-estimated for Malaysia and Thailand with the starting date being the date the loan rate was liberalised. The coefficients are almost identical for Thailand but change substantially for Malaysia. In this case, β₁ is 0.28 (up from 0.17) and β₂ is 0.19 (up from 0.13). These coefficients are still significantly lower than the corresponding values for the deposit rate (0.48 and 0.50), and so the different speed of adjustment is not simply due to the timing of liberalisation. [16]

In the case of the US, however, the adjustment coefficient declines over sub-periods, which may be due to risk or the declining significance of the prime rate as an indicator lending rate. [17]

This is generally a reasonable assumption for most of these markets – see Chinn and Frankel (1994a) and de Brouwer (1995). [18]

This test was also applied to the other Western Pacific economies from 90M1–94M12. As for Australia and Thailand, the other institutional rate is not significant for Japan, Malaysia and the Philippines, but is for Hong Kong, Indonesia, Singapore and Taiwan. Rates are market-determined in the latter set of countries, but these markets are the relatively more closed or cartelised. For Hong Kong and Singapore, where cartels are dominant, the interbank rate becomes statistically insignificant in the loan rate equation once the deposit rate is included, but is significant in the deposit rate equation along with the loan rate. [19]

When the first lags of the differenced money and loan rates are included in this equation, the lag of the differenced deposit rate remains significant but the coefficient (0.28) is offset by the other dynamics on the loan rate (−0.20) and on the interbank rate (−0.05). [20]