RDP 9604: Issues in Modelling Monetary Policy 3. The Monetary Policy Transmission Process

Our second broad task in this paper is to review empirical evidence on the monetary policy transmission process. As noted at the outset, this review draws mainly on research originating at the Reserve Bank. The topic can usefully be divided into three parts: the influence of cash rates on other financial prices; the role of money and credit; and the effect of financial variables on activity and inflation.

3.1 The Influence of Cash Rates on other Financial Prices

Changes in the cash rate can be expected to influence the entire structure of interest rates and yields on financial assets. A useful starting point is to consider the effect of cash rates on other interest rates relevant to private-sector decisions. Lowe (1995) examined the extent and speed of ‘pass-through’ from cash rates to sixteen interest rates of securities markets and financial institutions. In general, the pass-through to short-term money-market interest rates is rapid and complete, while most deposit and lending rates are much less responsive, particularly for low-balance deposit accounts, housing and consumer loans. The business indicator rate is the most responsive lending rate, but pass-through is nonetheless found to be incomplete. Since intermediaries' interest rates are likely to be the most important ones for private-sector decisions, this general pattern of slow pass-through would help to account for part of the overall transmission lag from cash rates to the economy.

The relationship with longer-term interest rates can be thought of as reflecting two effects: a ‘liquidity effect’, essentially representing the impact of cash-rate changes on near-term expectations of future cash rates; and an effect on longer-term inflation expectations. The expectations hypothesis of the term structure suggests an overall impact representing the expected net effect of these factors on future short rates averaged over the life of the security. Broadly consistent with this, results of simple level-regressions by Lowe indicate positive pass-through from cash rates to various longer-term securities rates, but that the size of the coefficients declines the longer is the maturity of the bond. Results from regressions using changes in interest rates are less clear, particularly for longer term bonds, with significant cash-rate effects only found on security yields out to a maturity of two years.

Formal tests of the expectations hypothesis of the term structure, by Macfarlane (1988) and by Whitelaw, de Roos and Groeger (1995), suggest that the hypothesis can be rejected with Australian data, though the rejections are not dramatic. These studies find some evidence of predictable excess returns in bond markets related both to prior movements in foreign bond yields and to domestic short-term rates. The first of these results might be consistent with models of financial risk premia, while the second can be interpreted as evidence of an ‘excess’ sensitivity of long-term interest rates to current short-term rates. Regarding international linkages, there is clear evidence of significant cross-country co-movements in levels of bond yields (Fahrer and Shori 1990; Orr, Edey and Kennedy, 1995) arguably due to a combination of correlated shifts in inflation expectations and common factors influencing real interest rates internationally. The highly internationally-correlated rise in bond yields during 1994 provides a powerful illustration of these linkages, but research by Gruen (1995) and by Orr et al. suggests that it is hard to find a fully convincing explanation for these movements in terms of observable fundamentals.

The study by Orr et al. uses a panel regression to estimate a model of long-term interest rate determination in 17 OECD countries. The authors distinguish between slow-moving or ‘fundamental’ explanatory factors and dynamics largely driven by interactions among interest rates in different countries. It is the latter that largely account for the 1994 episode while the fundamental factors (principally fiscal and current account positions and a measure of policy credibility) explain longer-term trends and cross-country differentials. An important result is the significance of a policy-credibility variable based on countries' past inflation record, which includes lags of inflation of up to 10 years. The implication is that credibility takes a long time to establish, and the detrimental effects of a poor track record of inflation-control take a long time to drop out of market expectations of inflation. This is a result that is at odds with strong presumptions of forward-looking behaviour made by theorists, but one that rings true with policymakers.

Taken together the sources cited point to strong and fairly consistent effects of cash rate changes on security-market rates out to around two years maturity, which might be thought of as the period over which markets can form reasonably well-based expectations of policy responses to foreseeable developments. Results on the determinants of long-term bond yields do not really provide a coherent picture but suggest that these are best characterised as reflecting a combination of forward and backward looking factors along with significant international influences. The backward-looking elements should not be underplayed, and in particular it seems that the lags in the effect of past inflation on nominal bond yields can be quite long. Effects of cash rates on key borrowing and lending rates are also subject to variable lags attributable to interest-rate smoothing by financial institutions, but there is some evidence of these lags shortening as markets become more competitive and contestable.

The influence of interest rates on the exchange rate represents an important component of the transmission process, but one where empirical analysis has not been very successful in linking theory to actual behaviour. A standard assumption in macroeconomic theory is that of uncovered interest parity, which relates expected exchange rate movements to interest differentials. This assumption allows a neat conceptual framework for exchange-rate determination whereby the steady-state exchange rate is thought of as being tied down by variables determining the long-run equilibrium, and the path of adjustment to that equilibrium is determined by the expected path of future interest differentials. Unfortunately, however, empirical evidence consistently rejects the joint hypothesis of rational expectations and uncovered interest parity, often finding that interest differentials have insignificant or wrong-signed coefficients as predictors of exchange rate movements.^[10] A recent example for Australia is a study by Gruen and Kortian (1996) which finds that predictable terms-of-trade movements generate predictable excess returns with respect to the $A exchange rate.

Notwithstanding these results it would seem undeniable that exchange-rate determination is strongly forward-looking and influenced by monetary policy settings, raising the question of whether or not the uncovered interest parity assumption might nonetheless be a useful approximation for the purposes of model-building. It is hard to give a categorical answer to this, since there is a lack of obvious alternatives despite the poor empirical performance of the hypothesis. Gruen and Wilkinson (1994) estimate a model of Australia's real exchange rate over the post-float period which they find to have good fit both within and out of sample, using the terms of trade and real interest differentials as explanatory variables.^[11] They report that long-term interest rates dominate short-term rates in the equation. Despite the good statistical properties reported, the implications of their results for a structural model are not clear, since the estimated equation is essentially a reduced form relationship among endogenous variables. In particular, interest rates and exchange rates are likely to be closely inter-related contemporaneously, making cause and effect relationships between them difficult to disentangle. For the purposes of policy analysis it makes sense to think of the exchange rate as being influenced by short-term interest rates, even though such effects may well be difficult to estimate empirically given the severe problem of simultaneous causation. As noted by Friedman (1995), this is part of a general problem of identifying the effects of monetary policy actions where those actions are largely systematic responses to movements in other macroeconomic variables that they also affect.

3.2 The Role of Money and Credit

A formal policy role for monetary aggregates (in the form of a ‘conditional projection’ for M3) was abandoned in Australia in 1985. This reflected an assessment that the aggregates were too unstable in their relation to final objectives to serve as a policy anchor. As noted above in Section 2, this does not necessarily preclude a role for money and credit as indicator variables, but their information content with respect to final policy objectives needs to be continually re-assessed.

The most comprehensive recent study on this topic was by de Brouwer et al. (1993) who tested the stability of a variety of money-demand functions and assessed the sensitivity of these relationships to alternative definitions of the key variables and to different testing procedures. Evidence of cointegration between money, income and interest rates was not particularly strong, and the results were sensitive to changes in the definition of activity and interest rates and to the testing procedure used. No evidence was found of cointegration between M1 and income, and only scant evidence for the money base. Of 192 feasible combinations of the real money base, real activity, nominal interest rates, and estimation techniques that were considered, evidence supported cointegration in only 11 cases. Results were similarly negative in the cases of M3 (4 out of 60 combinations) and Broad Money (5 out of 30). The strongest evidence was found for currency (8 out of 24 combinations) but this relationship has since deteriorated, with the ratio of currency to income shifting markedly beyond the sample period used for the formal tests. These results confirm the impression of instability apparent from eyeballing the data in Figure 2. Since cointegration is a minimal requirement for a stable money demand function, the results are not supportive of a stable role for money aggregates in the transmission process.

Alternative methods applied in earlier studies gave qualitatively similar results. A study by Bullock, Morris and Stevens (1989) using graphical evidence and correlation analysis emphasised the leading and coincident negative relationship between short-term interest rates and private demand, and found relationships involving the monetary aggregates to be generally unstable. VAR analysis by Stevens and Thorp (1989) looking at relationships between financial and activity variables found the role of money aggregates to be weak and highly sensitive to model specification and to the sample period.

An issue that has received increasing attention in international literature recently is the possible role of a ‘credit channel’ in monetary transmission.^[12] Theories of the credit transmission channel focus on market imperfections such as information and monitoring costs which limit borrowers' access to credit. These imply that access to credit for financing investment is likely to be affected not only by the risk-adjusted profitability of a given project, but also by factors such as the borrower's available collateral, cash flow, reputation, and other indicators of credit-worthiness, with credit rationed in equilibrium on the basis of a range of such criteria. This has two important implications. First, changes in interest rates and in business cycle conditions could be expected to have distributional and efficiency consequences related to systematic variation in borrowers' access to credit. A rise in interest rates will tend to reduce investment in projects that are least profitable, other things being equal, but the effect will be strongest in sectors where credit-enhancing factors such as collateral are scarce. Second, because the severity of credit rationing is likely to be an increasing function of the general level of interest rates, the credit channel represents a potential amplifying factor in the transmission of monetary policy. It is also likely to interact with macroeconomic developments, such as asset-price swings and trends in balance-sheet strength, which influence assessments of borrowers' credit-worthiness.

Tests of these propositions reported in the international literature do not focus strongly on the role of credit aggregates in macroeconomic outcomes; proponents of the ‘credit view’ argue that the role of these aggregates is not central to the story, since the stock of intermediaries' credit outstanding does not really measure variations in the impact of credit rationing. Rather, the emphasis in empirical work has been on examination of factors affecting the availability of credit at the firm or industry level. Results summarised by Bernanke and Gertler (1995) for the US are supportive of a role for credit-availability factors in explaining borrowing and investment at this level, and are thus consistent with the operation of a credit channel in monetary transmission.

Recent Australian studies are also supportive of such a channel. At an aggregate level, Lowe and Rohling (1993) find significant effects from various measures of balance-sheet strength on the availability of finance to business and, in particular, that rises in asset prices and in corporate equity tend to make finance easier to obtain. Panel data studies by Shuetrim, Lowe and Morling (1993) and Mills, Morling and Tease (1994) give a more detailed picture of the role of credit-factors in business behaviour at the firm level. They find that, in addition to macroeconomic influences, firms' borrowings and investment are influenced by a number of plausible indicators of credit availability. Most important among those are firm size, growth, collateral and cash flow. The influence of these factors on investment behaviour is found to be particularly strong for small firms. All these channels are probably difficult to capture in macroeconomic models, but a key lesson seems to be that interest rate changes interact with other macroeconomic factors in a way that potentially reinforces their impact on investment. The recent asset-price cycle is a leading example.

3.3 Effects of Financial Variables on Activity and Inflation

A recent review by Grenville (1996) identifies five main channels by which monetary policy affects activity and inflation:

• ‘intertemporal substitution’, broadly defined as the effect of interest rates on the incentive to postpone expenditure;
• effects operating through the exchange rate on the tradeable sector;
• cash-flow effects on liquidity-constrained borrowers;
• a wealth channel operating via the effect of interest rates on asset prices; and
• credit supply effects.

Further to these could be added an expectations channel, referring to the direct influence of policy decisions on expectations of inflation and growth. The discussion that follows cites evidence concerning the overall magnitude and timing of monetary transmission.^[13]

Two recent studies give estimates of the overall impact of interest rates on activity. Gruen and Shuetrim (1994) estimated a model of the Australian business cycle which specified real growth as a function of a weather indicator, the terms of trade, foreign output and real cash rates, finding a significant impact of real cash rates on GDP with lags of 6 to 18 months. The estimates imply the peak effect of a temporary 1 percentage point decrease in real cash rates (sustained for one year) raises four-quarter-ended GDP growth by about 0.4 per cent, and the level of GDP by about 0.65 per cent. Unpublished analysis by the authors suggests that similar results are obtained using alternative interest-rate maturities but that the two-year bond rate gives perhaps the best explanatory power. No independent role is found for the exchange rate in this aggregate equation, although effects of the exchange rate on the tradeable sector are well documented elsewhere.

Broadly similar results concerning the impact and timing of interest-rate effects on activity were obtained by Lowe (1992) in a study of the predictive power of the yield curve. Consistent with international evidence it was found that the slope of the yield curve (proxied in this study by the difference between yields on 10 year Treasury bonds and 180 day bank bills) has predictive power for a wide range of real variables including GDP, consumption, investment, a production index, dwelling approvals and car registrations. Predictive power for most variables is significant in the range of 9 to 18 months ahead, with somewhat shorter lags evident in the case of dwelling approvals. At its peak, a 1 percentage point increase in the yield spread predicts increased GDP growth by 0.6 percentage points. If it is presumed that policy influences the slope of the yield curve mainly by moving the short end, these results suggest effects of similar magnitude to the Gruen-Shuetrim equation. Investment and dwellings are found to be the most interest-sensitive sectors, although significant predictive power for consumption is also obtained.

Lowe also tested for information in the yield curve about future inflation. Significant predictive power was found although the estimated lags were perhaps implausibly long, the strongest effects occurring at around 24 to 36 months out.

Econometric studies of the inflation process by Stevens (1992), Cockerell and Russell (1995) and de Brouwer and Ericsson (1995) have adopted the general approach of modelling inflation as a function of domestic labour costs, import prices and measures of demand pressure, implicitly incorporating the view that policy affects inflation indirectly through its impact on these variables. Although the specifications adopted in these studies differ in important ways, the results emphasise a number of common points. First, the inflation process is subject to considerable inertia, as represented by the adjustment lags built into the estimated wage-price dynamics. Second, the estimated equations show a strong role for import prices in overall inflation, substantially bigger than the import share of expenditure (de Brouwer and Ericsson report a long-run coefficient on import prices of 0.4). This represents a powerful channel by which monetary policy can influence inflation via the exchange rate. Third, significant demand effects are captured by output gap measures, which generally enter with lags of one or two quarters. This is qualitatively consistent with Lowe's result, described above, that policy lags are longer on prices than on activity.

It is possible also to imagine monetary policy having a direct effect on inflation by influencing expectations. To the extent that the policy framework does influence inflation expectations directly, it seems sensible to think of this as working through the numerical objective itself, rather than through any intermediate variable such as money growth. Hard evidence on the importance of an expectations channel is difficult to come by, but the apparent inertia in inflation rates and sluggishness of expectations suggests the effect is not particularly strong. This seems to be confirmed by econometric evidence for Australia, New Zealand and Canada which could not find a role for the effect of policy targeting regimes on expectations independent of their effect on inflation itself.^[14]

3.4 Summary

The preceding discussion suggests a reasonably clear qualitative picture of the working of the policy transmission process. This can be broadly summarised as operating, in the first instance, via the effect of short-term interest rates on private expenditure, both directly and through effects on the exchange rate and on credit supply. Inflation is viewed as being influenced by monetary policy primarily through effects on the output gap and through the exchange rate. The inflation process appears to be subject to considerable inertia but, over time, the policy framework may also play a direct role in shaping inflation expectations. This description leaves no special role for monetary aggregates except as part of a menu of variables that might convey information to policy-makers about price and output trends. Plausible efforts can be made to quantify some, though not all, of the various parts of the transmission process and to estimate the length of the relevant lags. The areas where quantification seems to be relatively plausible (though the accuracy should not be overstated) include the transmission of cash rate changes along the yield curve, the effects of interest-rate changes on expenditure and the mechanics of the inflation process. Much greater uncertainty applies in other areas such as the effect of monetary policy on the exchange rate and the role of expectations in private sector decisions.

Footnotes

Studies by Gruen and Gizycki (1993) and Gruen and Menzies (1995) argue that these failures could be consistent with ‘near-rationality’ or with the presence of a subset of less-than-fully rational participants in the foreign exchange market. [10]

Other work by Blundell-Wignall, Fahrer and Heath (1993) and Tarditi (1995) built on these results. [11]

For recent surveys see Bernanke and Gertler (1995) and Cecchetti (1995). [12]

Grenville (1996) gives a detailed discussion of the separate channels identified above. [13]

See Debelle (1994). [14]