Until the early 1970s the economic literature on saving and investment mainly considered industrialized countries, with the common presumption being that the same analysis applied to developing countries. According to this analysis, the relationship between savings and interest rates could be ambiguous, in light of the opposing influences of the income and substitution effects, but the relationship between investment and interest rates was unambiguous. Low interest rates would promote investment spending and economic growth in developed and developing countries alike, in accordance with the Keynesian and neoclassical theories. Although the empirical evidence on the interest sensitivity of investment was mixed, economic policymakers in developing countries frequently adopted policies of low interest rates as a way of promoting economic growth.1 McKinnon and Shaw were the first to challenge seriously this conventional wisdom.
McKinnon (1973) analyzed an economy with very limited possibilities of external finance for the vast majority of investors. He argued that, because of the lumpiness of physical capital, savers may find it convenient to accumulate funds in monetary assets until they have enough resources to invest in higher-yielding physical assets. In McKinnon’s words, deposits may serve as a “conduit” for capital formation, making deposits and capital complementary assets.2 The availability of deposits with positive real rates of return may thus encourage both saving and capital accumulation.
Shaw (1973) also stressed the importance of positive real deposit rates as an inducement to save in financially repressed economies; but, unlike McKinnon, he emphasized external rather than internal financing possibilities as the effective constraint on capital formation. Focusing on the role of deposits as a source of funds for financial intermediaries, Shaw argued that high deposit rates may stimulate investment spending by allowing the supply of credit to expand in line with the financing needs of the productive sectors of the economy.
Although much theoretical and empirical work has criticized, tested, or elaborated on the McKinnon-Shaw propositions in the past ten years, there remains today some misunderstanding or confusion over the interpretation of these hypotheses.3 This confusion is due mainly to a neglect of the intertemporal nature of McKinnon’s complementarity hypothesis and to the common presumption that this hypothesis is inconsistent with Shaw’s debt-intermediation view.4
The purpose of this paper is to sharpen understanding of these two hypotheses through use of an analytical model that allows explicit treatment of the intertemporal aspects of the propositions. The model underscores the lag in the effect of interest rates on investment, savings, and asset-holding decisions and shows that the McKinnon-Shaw theses are mutually compatible. The model implies that the existing empirical work has failed to test adequately the validity of these two theses, in part because of misspecification of the appropriate lag structure. On the basis of this model the paper provides suggestions on how to specify aggregate investment, savings, and money demand functions in developing countries, so as to capture more accurately the complexity of saving-investment decisions.
The paper is organized as follows. Section I develops a three-period life-cycle model of consumption and saving behavior. This model allows study of the effects of deposit rate changes under various circumstances, including the McKinnon case of pure self-finance and the case of partial debt finance. The model is extended in Section II, which gives an informal derivation of some testable implications for aggregate economic behavior. Section III contains some concluding remarks and suggestions for further research.
I. A Microeconomic Model of Saving-Investment Behavior
A three-period life-cycle model of consumption with no bequests is developed in this section. The model differs from the conventional Fisherian model in that it allows for more than one asset, with different rates of return. The multiple asset feature is important because it makes possible the study of the complementarity hypothesis, which involves accumulation of savings first in the form of deposits and later in the form of physical capital. In the case of certainty about all rates of return, the complementarity relationship is derived under pure self-finance, and it is shown that the introduction of borrowing possibilities need not alter the nature of this relationship. For the case of pure self-finance, it is also shown that the complementarity relationship may break down when capital is a risky asset and deposits are a safe asset.
Consider a consumer who expects to live three periods and has a utility function that depends on consumption in each one of these periods. The consumer earns income only in the first two periods and saves part of it for consumption in the third period.
Savings can be invested in deposits with financial institutions or in physical capital. The consumer can finance some of his asset holdings with loans in the first two periods, and he can repay all his loans, liquidate all his assets, and consume all his wealth in the third period, since there are no bequests.
When capital markets function smoothly, the consumer’s asset holdings will reflect his rate of time preference, the rates of return on the various assets and liabilities, and, if these returns are uncertain, his attitude toward risk. In a financially repressed economy,5 however, the consumer may be constrained to borrow less than he wishes to, and, at the limit, he may have to finance his acquisition of all goods and assets with his own resources. Moreover, in the absence of a well-developed equity market, physical capital may be much less divisible an asset than in more sophisticated economies. This lumpiness, together with borrowing constraints, may make it necessary for investors to accumulate savings for some time in the form of deposits before investing in physical capital.6 It is precisely in regard to this capacity that McKinnon (1973) argues that deposits serve as a conduit for capital accumulation.
To determine the nature of the complementarity between deposits and physical capital, the consumer’s optimization problem is set up formally. The consumer maximizes
subject to the constraints7
|Ci||Spending in the beginning of the ith period on goods that are consumed during the period, i = 1,2,3|
|Di||Funds deposited with banks in the beginning of the ith period, i = 1,2|
|Ki||Physical capital purchased in the beginning of the ith period, i =1,2|
|Kmin||Minimum possible investment in physical capital|
|Li||Borrowing in the beginning of the ith period that is repayable in the beginning of period i + 1, i = 1,2|
|Maximum amount of borrowing possible in the ith period, i = 1,2|
|rd||Rate of return on bank deposits, rd < rL|
|rk||Rate of return on physical capital|
|rL||Rate of return on loans, rL > rd|
|Si||Savings in the /th period, i = 1,2|
|Yi||Noninterest income, received in the beginning of the ith period, i =1,2|
It is assumed that the marginal utility of Ci is positive, decreasing in Ci and independent of Ci for i = 1,2,3 and j ≠ i It also is assumed that the rate of return on deposits is lower than the loan rate (rd < rL), which implies that the consumer will never choose to borrow and hold deposits at the same time. Constraints (2) and (3) simply state that income in each period must be allocated between consumption and savings. Equations (4) and (5) are the financing constraints that require that first-period assets be financed out of borrowings and current savings, and second-period assets out of previously accumulated wealth plus new savings and borrowings. In the third period there is no current income, and all consumption is financed by the liquidation of second-period investments as is indicated by equation (6). Finally, inequalities (7) and (8) reflect the lumpiness of physical capital, and inequalities (9) and (10) are the borrowing constraints that are likely to be effective in financially repressed economies. If the consumer’s desired investment in physical capital falls short of the required minimum (Kmin), then he can hold either more capital than he wishes to (Kt = Kmin) or none at all (Kt =0). Of course, if the minimum capital requirement is not binding (Ki > Kmin), then inequalities (7) and (8) are inconsequential. Because the purpose is to highlight the conduit role of deposits, it is useful to assume that desired savings and borrowing possibilities are insufficient to make investment in capital positive in the first period, while in the second period enough resources have been accumulated to make inequality (8) ineffective. More specifically, one has
In other words, all investment in physical capital takes place in the second period while all first-period savings are placed in deposits. Thus, deposits can be viewed as a temporary abode of funds that facilitates the accumulation of resources necessary to satisfy the minimum capital requirement.
It is already clear at this point that the conduit role of deposits and their possible complementarity with physical capital result from the intertemporal nature of the process of capital accumulation. Deposits and capital in the second period (D2 and K2) may thus be perfect substitutes, whereas first-period deposits and second-period capital (D1 and K1) are complementary. To derive these relationships more explicitly, separate cases are considered that depend on the riskiness of physical capital and on the existence of borrowing possibilities.8
The Case of Certainty
In this case investors are, or act as if they were, certain about the rates of return that they will realize on all assets and liabilities. Two cases are distinguished that depend on the availability of borrowing possibilities.
The Model with Pure Self-Finance
A saver-investor may be confined to self-finance,9 either because he has no access to external funds (
Solving the consumer’s optimization problem yields
Conditions (16), (17), and (18) simply require that the marginal rate of substitution between consumption in any two periods equals the relevant discount rate. The deposit rate gives the rate at which first-period consumption can be exchanged for second-period consumption, since all savings are originally accumulated in deposits. In the second period, the minimum capital requirement is no longer binding, and the exchange rate between second-and third-period consumption is determined by the higher rate of return on capital.
Looking at the effects of interest rates on consumption, saving, and asset demands reveals that the model differs from the simple Fisherian framework in two important ways. First, two different assets are allowed, with distinct rates of return. This feature makes it possible, at least in principle, for the two rates to affect each variable in different directions. Second, in the model saving takes place in both the first and second periods. In addition to the ambiguities that typically are associated with the income and substitution effects of changes in interest rates, this feature raises the possibility of offsetting changes in saving in the two periods.
More specifically, the only unambiguous effects of changes in the deposit rate are those on second-period demand for capital and second-period saving:
In other words, a rise in the deposit rate discourages second-period saving and encourages capital formation.
The positive relationship between demand for capital (K2) and the rate of return on deposits (rd) is a direct result of the conduit role of deposits. This relationship is unambiguous because, in this version of the model, deposits can never be viewed as a competing asset for capital and are used exclusively as a temporary abode for savings. Thus, a rise in rd raises the quantity of funds available for investment in the second period without leading to a substitution away from capital and into deposits.11 Similarly, second-period saving (S2) is inversely related to rd as a result of this mode of accumulating wealth, first in the form of deposits and then in the form of capital. A change in rd can have no substitution effect on S2 because all wealth is held in physical capital in the second period. The income effect, then, obviously dominates, and a higher rd increases second-period wealth and, thereby, second-period consumption.
The other effects of rd and rk on consumption, saving, and demand for capital are subject to the usual qualifications with respect to the relative magnitude of the income and substitution effects. Three cases can be distinguished.
First, the substitution effect dominates:
Second, the substitution effect dominates for rd but not for rk:
Third, the income effect dominates:
In the first case, both interest rates have positive effects on first-period saving, and rk has a positive effect on saving and on demand for capital in the second period. The second case is slightly more interesting, since first-period saving is positively related to rd but negatively related to rk. This difference in the effects of rd and rk is because the substitution effect is stronger in the first period for rd than it is for rk. A rise in rd, on the one hand, stimulates a relatively larger amount of saving in the first period because it changes the rates of substitution between both first- and second-period consumption (1 + rd) and first- and third-period consumption [(1 + rd) (1 + rk)]. A rise in rk, on the other hand, leaves the terms of trade between first- and second-period consumption unchanged and has a weaker overall substitution effect on first-period consumption and saving. Finally, in the third case the income effect dominates for both interest rates in both periods.
The apparent paradox in the last case is that, even though a rise in rd discourages saving in both the first and second periods, it still stimulates demand for capital (K2) because inequalities (19) always hold. Since in the model all investment is self-financed, a question arises concerning the availability of resources for K2 to rise. K2 is financed with second-period savings and with first-period savings that have been placed in deposits:
This equation implies that a change in rd affects K2 not only through changes in savings out of first- and second-period non-interest income but also through changes in interest income from first-period saving:
Thus, even if S1 and S2 are inversely related to rd, the effect on interest earned on S1 allows a positive relationship between K2 and rd. With regard to consumption, inequalities (19) and (22) imply that a rise in rd raises C1 and C2. The assumption of diminishing marginal utility implies that it is optimal for C3 to share some of the gains of the increase in rd. But C3 can rise only if K2 rises because, with D2 = 0, constraint (6) becomes
Hence, a rise in rd leads to a higher K2 irrespective of the effects of such a rise on saving.
With respect to the properties of the asset demand functions, recall that all first-period savings are placed in deposits. Because demand for deposits is identical to savings in the first period, the properties of the deposit demand function can be obtained from inequalities (20)-(22). It follows that dD1/drk may be positive or negative, whereas dK2/drd is always positive. These results confirm McKinnon’s complementarity hypothesis. It should be emphasized, however, that this is a peculiar sort of complementarity because it refers to assets that are held at different points in time.
The results imply that demand for capital may be positively associated with deposit rates even if saving out of noninterest income responds negatively to these rates. Higher deposit rates increase the wealth of prospective investors and, thereby, the amount of resources available for future investment. This wealth effect may be especially important in countries with underdeveloped capital markets, where the nonexistence of alternative financial assets may force savers to hold deposits for some time, even if the real deposit rate is negative. The compound wealth effect of deposit rates is likely to be substantial, and a change in the authorities’ policy stance from negative to positive real rates may dramatically increase resources available for self-financed investment.
To illustrate the importance of this type of wealth effect, it is useful to consider a real-world example. Two countries have been chosen, one (country A) that followed a positive real rate policy and one (country B) that had negative real rates during the period 1978-81 (Table 2).
|Year||Real Deposit Rate|
|Country A||Country B|
We consider two savers that deposited 100 local currency units each, in bank accounts in countries A and B, respectively, in 1978. By 1981 these deposits would be worth 122.5 units in 1978 prices in country A but only 46.4 units in country B. The depositor in the high-interest-rate country would have almost three times the wealth he would have had if he had lived in the low-interest-rate country. In light of the magnitude of this wealth effect, the theoretical possibility that higher deposit rates will stimulate investment in physical capital even if they do not generate more saving seems quite plausible.12
The Model with External Finance
In this case the investor is assumed to be able to obtain one-period loans not exceeding
In the absence of risk, this interest rate structure induces the investor to borrow as much as he can in the second period in order to increase his holdings of physical capital:
Finally, it is assumed that the loan rate exceeds the deposit rate by a fixed margin (α) that covers the cost of financial intermediation:
This equation implies that changes in deposit rates are always reflected in equal changes in loan rates.
A rise in the deposit rate now increases interest income from first-period savings but also raises the cost of borrowing in the second period. In addition to the substitution effect, which encourages first-period saving, a negative income effect may now apply that would encourage saving in both periods. The sign of the income effect depends, however, on the relative importance of loans as a source of financing, and, as in the case of pure self-finance, the effects of changes in interest rates on savings are ambiguous when the income effect is positive.
The availability of external financing does not affect the results with respect to demand for physical capital—which continues to be related positively to the deposit rate even when loans account for a large share of the financing of physical capital—because of the assumption that actual borrowing is not affected by interest rate changes when there is excess demand for loans. As long as the rate of return on capital exceeds the loan rate, the investor wishes to borrow as much as he can and is restricted to the maximum amount of loans available to him (
A relaxation of the borrowing constraint, through an increase in
The existence of external financing possibilities allows one to highlight some of the distinctions between the various channels through which interest rates affect saving and investment. McKinnon’s complementarity hypothesis, on the one hand, emphasizes the role of deposits in encouraging self-financed investment. A rise in the deposit rate stimulates demand for capital by making savings accumulation more rewarding and by increasing the amount of internally financed investment. Shaw’s debt-intermediation view, on the other hand, focuses on the role of deposit accumulation in expanding the lending potential of financial intermediaries. Higher deposit rates encourage the inflow of deposits to banks, which in turn can increase lending, thereby stimulating externally financed investment. Although the Shaw and McKinnon theses emphasize different aspects of the process of accumulation of financial assets and liabilities, it is clear from the discussion so far that these theses should be viewed as complementary rather than competing theories. A more systematic attempt to integrate the two theories is taken up in Section II below.
The Case of Uncertainty
In general, the returns to capital and other assets may be associated with a great deal of uncertainty. If deposits are considered safer than capital, the two assets will not be perfect substitutes. A risk-averse investor may then choose to hold both capital and deposits if the expected rate of return on capital is higher than the deposit rate.
We can extend the model of the previous subsection to account for risky investments by allowing the rate of return on capital to be a random variable.14 We assume that the deposit and loan rates are known with certainty, which again implies that the investor will never wish to borrow and hold deposits at the same time. If the minimum capital requirement can be met only in the second period, the investor will place all his savings in deposits in the first period, and, in the second period, he may choose to hold a diversified portfolio of deposits and capital or to hold all his wealth in capital. The possibility of portfolio diversification is especially interesting because it allows deposits and physical capital to be complements in the first period but substitutes in the second period. The effects of changes in interest rates, for this case, can be determined by solving the consumer’s optimization problem.
The marginal rate of substitution between C1 and C2 is the same as in the certainty model because all first-period savings are still held in deposits. The rate of substitution between C2 and C3, however, can vary depending on the relative size of deposit and capital holdings in the second period. This substitution rate attains its minimum (1 + rd) when no capital is held (K2 = 0) and its maximum (1 + rk) when no deposits are held (D2 = 0). Finally, the rate at which expected third-period consumption can be traded for a reduction in riskiness decreases as the margin between rk and rd widens. A reduction in riskiness can only be attained by shifting from capital to the safe asset. Given any fall in C3, the larger is the gap rk-rd, the smaller is the corresponding fall in K2, and the smaller is the accompanying risk reduction.
In this formulation of the consumer’s maximization problem, deposits and capital need not be complements. In particular, it can be shown that under some conditions
A rise in the deposit rate now affects K2 not only through the conduit effect but also through the competing asset effect. Thus, a rise in rd may stimulate saving and demand for deposits in both periods, but it may also induce a shift from capital to deposits in the second period that offsets the savings effect. Notice that again it is possible for saving to be related positively to rd but related negatively to rk. A rise in rd induces the investor to shift some of his funds from K2 to D2; the shift implies a lower but more certain level of consumption (C3) in the third period, with the fall in C3 moderated through increases in saving in the first two periods. A rise in rk, however, has the reverse effect. Shifts from D2 to K2 raise expected consumption C3 and its variance in the third period and allow the individual to spread his increased consumption evenly by decreasing saving in the first two periods.
The model in this section can be generalized to take account of inflationary uncertainty, which is endemic in many developing countries. So far, deposits have been assumed to be a safe asset, and only capital has been associated with risk, with no distinction made between nominal and real rates of interest. In practice, even if the nominal deposit rate is known with certainty, the variability in the rate of inflation may account for large fluctuations in real interest rates. In such situations both deposits and physical capital may be viewed as risky assets, and inflation hedges—such as precious metals or real estate—may play the role of safe assets.15 Interest rate policy may then have a much more powerful effect on the public’s asset composition because the authorities may raise real deposit rates to consistently positive levels, thereby decreasing the riskiness of deposits. In time, deposits may replace inflation hedges in investors’ portfolios as the highest-yielding safe asset, and this substitution may further stimulate current savings and the process of financial deepening.
II. Extension and Macroeconomic Implications of the Model
The foregoing analysis has highlighted the differing roles of deposits in the various stages of the process of wealth accumulation. Deposits with financial institutions may be an investor’s only outlet for savings while he accumulates funds for the purchase of higher-yielding assets. The investor may later shift his position in the financial system from that of net creditor to that of net debtor, or he may choose to keep some deposits as the safe asset, along with higher-yielding, riskier assets in a diversified portfolio.
At any point in time some investors are at an early stage of the process of wealth accumulation while others are at a later stage. To explain aggregate economic behavior, one must take into account both types of investors as well as the possible interactions between them. In the preceding discussion it was assumed that each individual borrower had access to a fixed amount of loans. From the point of view of the economy as a whole, however, aggregate credit flows are bound to be related to aggregate deposit flows. Accordingly, interest rate policy will influence capital formation by affecting not only the returns to internally financed investment but also the availability of external financing. The former effect is the focus of the analysis in McKinnon (1973), whereas the latter is emphasized by Shaw (1973).
In this section the analysis attempts to integrate these two approaches and to determine their implications for aggregate economic behavior. Past, current, and expected future changes in interest rates are distinguished to account for the intertemporal nature of McKinnon’s complementarity hypothesis, and a positive relationship between the aggregate levels of deposits and loans is stipulated in accordance with Shaw’s debt-intermediation view. The effects of interest rates on saving, investment, and demand for money are considered in turn.
The aggregate savings function will reflect the behavior of the different types of depositors and borrowers described in Section I. Therefore, the savings function implied by our model can be written as
where St is the flow of saving in period t; Lt is the stock supply of loans in period t;
In addition to current interest rates and the current supply of loans, both past and expected future rates enter the savings function. Past deposit rates affect the level of currently available internal funds and thereby current saving, whereas expected future rates of return on capital affect the expected return of current deposits that are to be placed in capital in the future. Income and substitution effects cause the familiar ambiguities with respect to the effects on saving of changes in current and expected future interest rates. Changes in past rates, however, can have only income effects in the current period and, as a result, only negative effects on current saving. Finally, increases in the supply of loans will encourage current consumption and will discourage current saving according to the model:
A positive relationship between the stock of deposits and the supply of loans is assumed:
where Dt is the stock of deposits in period t. A deposit rate change that raises demand for deposits will thus lead to an increase in loan supply. In long-run equilibrium, interest rates can be expected to stabilize, so that
The ultimate effect on saving of a permanent deposit rate change can then be obtained from
It follows that, in general, the long-run impact of deposit rate changes on saving is ambiguous. Even if a higher rate stimulates current saving, equation (39) implies that this effect will be dampened over time, under the influence of the income effects of past interest rate increases and loan supply increases.
Aggregating the demands for capital by the various types of asset holders described in the previous section yields
According to the model, aggregate demand for capital is positively related to past deposit rates, but it may be negatively related to the current deposit rate. From the point of view of current holders of capital, deposits may have served as a conduit for capital accumulation in the past, but current deposits may be a substitute for the riskier physical assets. Of course, for those savers who are still accumulating all their funds in deposits, with the ultimate purpose of investing in capital in the future, demand for capital will also be positively associated with the deposit rate. It is future demand for capital, however, that will be affected by changes in current deposit rates. Thus, one might have
The model also implies a positive relationship between demand for capital and loan supply, and it is assumed that the aggregate demand for capital is positively related to its own rate of return. Thus,
This framework, which gives a theory of the determination of the stock demand for capital, can be easily extended to account for the flow of investment spending. Assuming that the marginal productivity (R) and the supply (K0) of capital are fixed in the short run, one can interpret
To determine the long-run effect of changes in deposit rates on capital formation, both direct and indirect interest rate effects on the demand for capital are considered. The former reflect the substitutability or complementarity between deposits and capital, whereas the latter capture the impact of interest rates through their effect on financial deepening:
Assuming that inequalities (37), (41), and (42) hold and that one finds that is a necessary and sufficient condition for dKt/drd to be positive in the short run, and a sufficient condition for dKt/drd to be positive in the long run. Inequality (47) ensures that the financial deepening effect of deposit rate increases on demand for capital more than offsets the effect of substitutions of deposits for capital. One can draw the following conclusions from expressions (45) and (47).
First, a rise in rd will discourage capital formation in the short run if loans finance a relatively large increase in current consumption or fail to increase in line with deposits. The latter may happen if, for whatever reason, financial intermediaries use some of their new deposits to acquire assets other than loans (for example, cash reserves or foreign assets).17
Second, in the longer run, capital formation will be positively related to the deposit rate if the conduit role of deposits dominates the substitute asset role. A deposit rate increase thus may discourage investment immediately while encouraging it in the long term. The time it takes for an interest rate change to have its full effect will be longer the longer the period over which self-financed investors accumulate their savings in deposit form.
Third, McKinnon’s complementarity hypothesis addresses inherently intertemporal aspects of the process of capital accumulation and has very little to say about the short run.18 Shaw’s debt-intermediation view, however, has some short-run implications because asset substitutions in favor of deposits may occur in a relatively short period of time. The approach taken here shows that the McKinnon-Shaw theses are mutually compatible and can be integrated into one framework.
Demand for Money
Assuming that money consists mainly of deposits with financial intermediaries, one can draw the model’s implications for the money demand function.19 Deposits are held either as a safe asset in conjunction with riskier assets or as a temporary abode of funds that are to be used for the acquisition of higher-yielding assets in the future. In their former capacity, deposits are a substitute for capital, but in their latter use they are complementary to capital. If all investors follow the same pattern of asset accumulation over their planning horizons, then the relationship between deposits and capital is a function not only of relative rates of return but also of temporal considerations. Current deposits may thus be a substitute for current capital but a complement to future capital.
Accordingly, demand for money may be negatively related to the current rate of return on capital but positively related to the expected future rate of return on capital. The current supply of loans does not enter the money demand function because depositors and borrowers are distinct groups at any point in time. Money demand, however, is negatively related to the expected future loan supply, since a rise in this supply lowers the need for current accumulation of internal funds. Thus, one might have
The foregoing discussion highlights three important aspects of the saving-investment process in financially repressed economies. First, interest rates affect individual asset holders’ decisions with a complex lag structure. For example, an investor’s demand for capital may be positively related to past deposit rates but negatively related to current rates; similarly, his demand for money may be negatively related to the current rate of return on capital but positively related to the expected future rate of return on capital. Second, at the aggregate level, external financing possibilities are closely related to savings flows and, as a result, are also affected by interest rates. Investment spending may thus be affected by deposit rates both directly, through the competing or complementary asset effect, and indirectly, through the flow of loans. Finally, the analysis suggests that, in the presence of financial constraints, saving and investment decisions are intimately related and can be viewed as different stages of the same process. Accordingly, any theory of saving behavior in financially repressed economies should also seek to take into account the investment motive.
These findings point to some serious shortcomings of most empirical studies of the complementarity hypothesis.20 These studies tend to disregard the investment motive by excluding the rate of return on capital from the estimated equations, thereby blurring the interrelationships among the saving, investing, and money-holding decisions.21 As a result, these equations may yield biased estimates that distort the view of the relative importance of financial and real sector developments. This problem may be most acute in cross-sectional studies because there may be large crosscountry variations in the rate of return on capital. Another general shortcoming of the empirical literature is that it disregards the intertemporal aspects of the saving-investment process, which are central to the complementarity relationship.22 The resultant estimates thus cannot possibly capture the intricacies of this relationship23 and may be subject to misinterpretation.24
III. Conclusions and Suggestions for Further Research
The McKinnon-Shaw models emphasize different aspects of the effects of interest rate liberalization in a financially repressed economy. McKinnon (1973) focused on the linkage between internally financed investment and the deposit rate, whereas Shaw (1973) highlighted the importance of financial deepening and external financing. The two approaches complement each other because most projects are financed in part with own funds and in part with borrowings. This paper has illustrated how the two views can be integrated without altering their basic conclusions.
In the context of a life-cycle model of consumption, the financing constraint applies over each individual’s planning horizon, which comprises several periods. The possibilities of substitution between current and future consumption and between different assets allow for a great variety of saving-investment patterns. In particular, when physical capital is lumpy, deposits may serve as a “conduit” for its accumulation, and high deposit rates may stimulate investment even if they do not encourage saving. Negative real deposit rates may thus constitute a tax on saver-investors that decreases the amount of internally generated resources available for investment.25 The other side of this tax, of course, is the corresponding subsidy to borrowers that have access to artificially cheap credit. The net effect of interest rate changes on aggregate investment then depends on the fraction of total investment that is financed by credit.26
The complementarity of deposits and physical capital is intertemporal, with current deposits intended to finance future investment. This relationship implies that even if higher deposit rates discourage investment in the short run they may encourage it in the long run. Instantaneous portfolio shifts into deposits may have an immediate negative effect on investment, whereas the increase in the rate of accumulation of internal funds will result in increased investment only gradually. The latter response may be extremely sluggish in the presence of uncertainty about inflation. If the government is not unequivocally committed to a policy of positive real deposit rates, the risk associated with accumulating savings in deposit form may outweigh any short-run interest rate advantage. Investors then know that unanticipated inflation may reduce the real value of nominally denominated assets quite rapidly, before they get a chance to switch to inflation hedges.27 Thus, a policy of high interest rates will have its full effect only after enough time has elapsed to eliminate any problems of credibility. The longer the public’s experience with negative real rates and inflationary uncertainty, the longer this policy lag may be.
The findings of this paper underscore the need for both further theoretical and empirical research into the McKinnon-Shaw propositions. The theoretical model in Section I is adequate for demonstrating some of the implications of the complementarity hypothesis, but it is clearly too simple to give a realistic account of saving behavior in repressed economies. Variables such as the demographic characteristics of the population, which are crucial determinants of aggregate saving behavior, can only be accounted for in a much more general, multiperiod framework. Such a framework would also allow a more rigorous treatment of the complementarity hypothesis. For example, the number of years required for a saver to accumulate enough funds to satisfy the minimum capital requirement may be an important determinant of his propensity to save. The larger is this time requirement, the longer is the horizon over which the saver will have to anticipate the rate of return on capital, and the lower are the rewards from saving. It would be a useful exercise to extend the model presented here and to derive its implications for life-cycle saving decisions under alternative assumptions about utility function parameters, the structure of interest rates, and the demographic composition of the population (for examples, see Tobin and Dolde (1971) and Summers (1981)).
The theoretical model presented here has suggested that interest rates affect expenditure-saving decisions through a complex and, possibly, very long lag. Moreover, in the presence of inflationary uncertainty, the ex ante current real deposit rate may be a function of ex post past rates, further complicating this lag structure. Statistical tests of the complementarity hypothesis are thus likely to require long interest rate series, which may be unavailable for many developing countries. In view of the serious data limitations, it is perhaps most desirable to try to estimate reduced-form savings and investment equations rather than to attempt a determination of the precise transmission mechanism for interest rate changes.
More generally, time-series savings equations are unlikely to yield reliable coefficient estimates. Even for the U.S. economy, with its advanced statistical collection methods, aggregate data have failed to provide conclusive evidence of the interest sensitivity of savings. (See, for example, Weber (1970), Howrey and Hymans (1978), Boskin (1978), and Gylfason (1981).) Auerbach and Kotlikoff (1983) have shown that, even with a “clean” data set generated from simulations of a theoretical life-cycle model, regression equations yield highly unstable estimates of the savings function. One promising approach for future studies of investment and saving behavior may be to conduct simulations of theoretical models, with parameters that are based on micro-economic data.28
AbeS.M.J.FryB.K.MinP.Vongvipanond and T.P.Yu “The Demand for Money in Pakistan: Some Alternative Estimates,” Pakistan Development Review (Islamabad) Vol. 4 (Summer1975) pp. 249–57.
AbeS.M.J.FryB.K.MinP.Vongvipanond and T.P.Yu “Financial Liberalization and Domestic Saving in Economic Development: An Empirical Test for Six Countries,” Pakistan Development Review (Islamabad) Vol. 16 (Autumn1977) pp. 298–308.
AkhtarM.A. “The Demand for Money in Pakistan,” Pakistan Development Review (Islamabad) Vol. 13 (Spring1974) pp. 40–54.
AuerbachA.J. and LaurenceKotlikoff “An Examination of Empirical Tests of Social Security and Savings,” in Social Policy Evaluation: An Economic Perspectiveed. byElhananHelpmanAssafRazin and EfraimSadka (New York: Academic1983) pp. 161–79.
BoskinMichael J. “Taxation, Saving, and the Rate of Interest,” Journal of Political Economy (Chicago) Vol. 86 (April1978) pp. S3–S27.
FryMaxwell J. “Money and Capital or Financial Deepening in Economic Development?” Journal of Money Credit and Banking (Columbus, Ohio) Vol. 10 (November1978) pp. 464–75.
FryMaxwell J. “Models of Financially Repressed Economies,” World Development (Oxford, England) Vol. 10 (September1982) pp. 731–50.
GalbisVicente “Money, Investment and Growth in Latin America, 1961-73,” Economic Development and Cultural Change (Chicago) Vol. 27 (March1979) pp. 423–43.
GuptaKanhaya L.Finance and Economic Growth in Developing Countries (London: Croom Helm1984).
GylfasonThorvaldur “Interest Rates, Inflation, and the Aggregate Consumption Function,” Review of Economics and Statistics (Cambridge, Massachusetts) Vol. 63 (May1981) pp. 232–45.
HowreyE. Philip and Saul H.Hymans “The Measurement and Determination of Loanable-Funds Saving,” Brookings Papers on Economic Activity: 3 (1978) The Brookings Institution (Washington) pp. 655–85.
International Monetary FundResearch DepartmentInterest Rate Policies in Developing Countries Occasional Paper 22 (WashingtonOctober1983).
JorgensonDale W. “Econometric Studies of Investment Behavior: A Survey,” Journal of Economic Literature (Nashville, Tennessee) Vol. 9 (December1971) pp. 1111–47.
KhatkhateDeena R. “False Issues in the Debate on Interest Rate Policies in Less Developed Countries,” Quarterly ReviewBanca Nazionale del Lavoro (Rome) No. 133 (June1980) pp. 205–24.
LeffNathaniel H. “Rates of Return to Capital, Domestic Savings, and Investment in the Developing Countries,” Kyklos (Basle, Switzerland) Vol. 28 (No. 41975) pp. 827–51.
McKinnonRonald I.Money and Capital in Economic Development (Washington: The Brookings Institution1973).
MikesellRaymond F. and James E.Zinser “The Nature of the Savings Function in Developing Countries: A Survey of the Theoretical and Empirical Literature,” Journal of Economic Literature (Nashville, Tennessee) Vol. 11 (March1973) pp. 1–26.
QureshiZia M. “Household Saving in Pakistan: Some Findings from Time-Series Data,” Pakistan Development Review (Islamabad) Vol. 20 (Winter1981) pp. 375–97.
ShawEdward S.Financial Deepening in Economic Development (New York: Oxford University Press1973).
SummersLawrence H. “Capital Taxation and Accumulation in a Life-Cycle Growth Model,” American Economic Review (Nashville, Tennessee) Vol. 71 (September1981) pp. 533-44.
SummersLawrence H. “The After-Tax Rate of Return Affects Private Savings,” American Economic Review Papers and Proceedings (Nashville, Tennessee) Vol. 74 (May1984) pp. 249-53.
SundararajanV. “Debt-Equity Ratios of Firms and Interest Rate Policy: Macroeconomic Effects of High Leverage in Developing Countries,” Staff PapersInternational Monetary Fund (Washington) Vol. 32 (September1985) pp. 430-74.
TobinJames “Liquidity Preference as Behavior Towards Risk,” Review of Economic Studies (Edinburgh, Scotland) Vol. 25 (February1958) pp. 65-86.
TobinJames “A General Equilibrium Approach to Monetary Theory,” Journal of Money Credit and Banking (Columbus, Ohio) Vol. 1 (February1969) pp. 15-29.
TobinJames and W.C.Brainard “Financial Intermediaries and the Effectiveness of Monetary Controls,” American Economic Review (Nashville, Tennessee) Vol. 53 (May1963) pp. 383-400.
TobinJames and W.C.Brainard “Pitfalls in Financial Model Building,” American Economic Review (Nashville, Tennessee) Vol. 58 (May1968) pp. 99-122.
TobinJames and W.C.Brainard “Asset Markets and the Cost of Capital,” in Economic Progress Private Values and Public Policy: Essays in Honor of William Fellnered. byBelaBalassa and RichardNelson (Amsterdam: North-Holland1977) pp. 235-61.
TobinJames and W.Dolde “Wealth, Liquidity and Consumption,” in Federal Reserve Bank of Boston Consumer Spending and Monetary Policy: The Linkages Conference Series No. 5 (Boston1971) pp. 99-160.
van WijnbergenSweder “Interest Rate Management in LDCs,” Journal of Monetary Economics (Amsterdam) Vol. 12 (September1983) pp. 433-52.
VogelR.C. and S.A.Buser “Inflation, Financial Repression and Capital Formation in Latin America,” in Money and Finance in Economic Growth and Development: Essays in Honor of Edward S. Shawed. byRonald I.McKinnon (New York: Marcel Dekker1976) pp. 35-70.
WeberWarren E. “The Effect of Interest Rates on Aggregate Consumption,” American Economic Review (Nashville, Tennessee) Vol. 60 (September1970) pp. 591-600.
Mr. Molho, an economist in the Central Banking Department, is a graduate of Yale University.
A detailed discussion of various rationales for a policy of low interest rates is given in Shaw (1973, pp. 92-112). For a survey of empirical studies on investment behavior in advanced countries, see Jorgenson (1971). Empirical studies on the interest sensitivity of savings and investment in developing countries were much more scarce until the early 1970s, in part as a result of data limitations. For discussions of the relevant literature, see Mikesell and Zinser (1973) and Leff (1975).
Even though McKinnon described an extreme case of financial repression, with no possibilities of external finance, his argument is intuitively appealing under much more general circumstances. For example, in the U.S. economy, with its highly sophisticated financial system, housing investment is subject to similar considerations. External financing is typically available for only a fraction of the price of a house, and some savers may need a few years to accumulate sufficient funds for a downpayment. For these savers, high money market rates may encourage accumulation of liquid assets for the purpose of buying a house. Even if higher rates do not generate more saving, they may increase investment by making more expensive houses affordable.
Gupta (1984, p. 3), for example, argues that “neither McKinnon nor anyone else has offered any rationale” for McKinnon’s assertion that “ ‘private saving (investment) is… sensitive to the real return on holding money,’” although this assertion is at the center of McKinnon’s argument. Van Wijnbergen (1983, p. 433), in a similar vein, refers to McKinnon’s analysis as eloquent but occasionally vague. For specific references to misunderstandings of McKinnon’s complementarity hypothesis, see footnote 18.
Fry (1982), for example, asserts that complementarity is incompatible with the debt-intermediation view. Gupta (1984) also refers to the two hypotheses as competing theories, although he makes an attempt to integrate them in one empirical model.
Financial repression is defined to entail artificially low deposit and loan rates that give rise to excess demand for loans and to nonprice credit rationing (McKinnon (1973, Ch. 7) and Shaw (1973, Ch. 4)).
This pattern of savings accumulation is by no means peculiar to financially repressed economies. Self-financing may be optimal even in economies with highly developed capital markets when borrowing rates are sufficiently high.
In principle, C3 may be a random variable if there is uncertainty with respect to interest rates. Constraints (5) and (6) can then be interpreted to hold for expected values. The cases with and without risk will be considered separately in the following analysis.
The detailed solution of the model and the derivation of the results presented in the following subsections (“The Case of Certainty” and “The Case of Uncertainty”) are available from the author on request.
This case is perhaps the most appropriate one for illustrating the nature of complementarity between deposits and capital described by McKinnon, who assumes that “all economic units are confined to self-finance, with no useful distinction to be made between savers (households) and investors (firms)—-These firm-households do not borrow from, or lend to, each other” (McKinnon (1973, p. 56)).
In addition to riskiness, divisibility, and the rate of return, there are other characteristics, such as taxability, liquidity, and transaction costs, that may make one asset dominate the other in consumers’ preferences. The model abstracts from taxes, and liquidity is not very important because the investor knows with certainty when he will be selling his assets. Thus, the rates of return, which can be defined to be net of transaction and other related costs, ultimately determine which asset will be held.
The implicit assumption here is that rd does not rise enough to overtake rk. If rd > rk, then the model implies that all wealth is shifted into deposits and that investment in physical capital falls to zero; this is one of the points made by Khatkhate (1980).
The partial equilibrium framework here is not well suited to a rigorous treatment of this issue. The other side of negative real deposit rates may be artificially cheap credit. The net effect of interest rate changes on capital formation, then, depends on how this subsidized credit is used. If borrowers allocate subsidized credit to investment spending exclusively, then low rates need not hamper capital formation. If, in contrast, credit finances consumer expenditures or capital flight, higher rates may be expected to promote domestic investment. Such general equilibrium considerations are outside the scope of this paper, but some of these issues are discussed informally in Section II.
An alternative way of modeling the borrowing constraint would be to assume a maximum debt-equity ratio. Although this might seem more plausible than a nominal ceiling on borrowing capacity, the latter has been chosen because it makes the algebra much more tractable.
The model of this section is based on Tobin (1958).
Vogel and Buser (1976) allow for three risky assets. Their one-period model, however, cannot adequately account for the conduit role of deposits.
The possibility that the expansion of the operations of financial intermediaries may be associated with a decrease in investment spending in the short run has been extensively discussed by Tobin and Brainard (1963) and, more recently, by van Wijnbergen (1983).
These aspects of McKinnon’s model have been subject to some misunderstanding. Vogel and Buser (1976, p. 36), for example, write that
Money and capital cannot be complements when they are the only two assets held in the portfolio and when the constraint on total assets is fixed. Thus, it is not surprising that the complementarity hypothesis has been overlooked in growth models where money is grafted onto the economy as the second of only two assets. However, the consideration of additional classes of assets introduces the possibility of limited complementarity. For McKinnon and Shaw, the additional assets considered are stores of goods or finished inventories labeled as inflation hedges.
As was shown above, temporal considerations may account for complementarity between money and capital in the absence of any other assets. Similarly, van Wijnbergen’s (1983) criticism of McKinnon (1973) is based on a one-period portfolio-balance model, which is unsuited for capturing the intertemporal complementarity that is at the core of McKinnon’s argument.
The balance-sheet constraint that can be obtained by aggregating equation (5) over all individual savers implies that the aggregate demand for money can be determined as a residual once the aggregate savings and demand for capital functions have been determined. The balance-sheet constraint can also be used to determine the interrelationships among the partial derivatives of the savings and asset demand functions.
See, for example, Abe and others (1975 and 1977), Akhtar (1974), Fry (1978), Galbis (1979), Gupta (1984), Qureshi (1981), Vogel and Buser (1976), and Yoo (1977). A survey of some of these and of other relevant studies is also included in Gupta (1984).
One possible reason for the exclusion of this variable may be the practical difficulties associated with its measurement. For a survey of the theoretical and empirical issues relating to the measurement of the rate of return on capital in developing countries, see Leff (1975).
This disregard may in part stem from the nature of the original presentations of the McKinnon-Shaw arguments, which were based on growth models describing steady-state equilibria, but data limitations may also account for researchers’ aversion to specifications with long lags.
Gupta (1984), for example, estimated investment functions for a number of developing countries, including only current financial savings and the current interest rate as explanatory variables; a more appropriate specification would also seek to relate current investment to past financial savings and past rates of return on financial assets, in accordance with the theoretical model presented here.
An example of the possibilities for misinterpretation of the empirical evidence on the complementarity hypothesis is Fry’s attempt to determine which of the McKinnon-Shaw models is more relevant by estimating different versions of the money demand function on data from ten Asian countries. Fry (1978, p. 470) started with McKinnon’s specification of real money demand as a function of real income, the real deposit rate, and the investment-income ratio. He estimated this function by regressing money demand on the contemporaneous explanatory variables, a procedure that yielded a negative and statistically significant coefficient for the investment-income ratio. The sign of this coefficient was interpreted as evidence refuting the complementarity hypothesis. According to the model presented here, however, complementarity would imply a positive association between current demand for money and intended future investment. To the extent that investment is self-financed, current investment might be associated with a decumulation of money balances, making the above result compatible with the complementarity hypothesis.
In industrialized countries, interest rate increases may be associated with a negative wealth effect for holders of bonds and stocks. Deposits, however, are not subject to such risk of capital loss. Unlike more illiquid assets, deposits have a fixed current nominal worth that can be fully realized upon withdrawal.
For a systematic analysis of how the effects of interest rates on investment depend on the debt-equity ratios of firms, see Sundararajan (1985).
Such experiences were all too common in the hyperinflations that preceded and followed World War II in several European countries. It is not surprising that, after seeing the value of paper assets reduced to practically nothing on various occasions, many Europeans are still reluctant to hold large sums of financial assets and have a special affinity for traditional inflation hedges such as gold.
For a description of some of the pitfalls in the statistical estimation of aggregate consumption functions, see Summers (1984).