Visit Title Page
Students Corner
Back to first page of Module: 3
|
A Decade of Economic Reforms - Review by RBI
[Source: RBI Report on Currency and Finance 2001-2002 dated March 31, 2003]
Module: 3 - Monetary Reforms - Money, Credit And Prices
Transmission Channels of Monetary Policy
In India, as in many other developing countries, credit markets were typically segmented till the 1980s, with interest rate controls and directed lending stipulations. During the 1990s, financial sector liberalisation brought out greater linkages among the various segments of the financial markets paving the way for the emergence of an interest rate channel of monetary policy transmission. With the increasing opening of the economy, the exchange rate channel of the monetary policy transmission is also expected to reinforce the traditional interest rate channel (Box V.2). The transmission channels of monetary policy are, therefore, expected to have undergone changes, consequent upon the liberalisation of financial markets. Against the backdrop of these changes, this Section makes an attempt to examine the transmission process in the 1990s vis-a-vis the 1980s.
Box V.2
Monetary Transmission Mechanism: An Overview of Empirical Evidence
The transmission mechanism describes as to how monetary policy action affects output and inflation, which are the final objectives of monetary policy. Monetary policy action transmits to the ultimate objectives through two broad sets of channels, financial prices (e.g., interest rates, exchange rates, yields, asset prices, equity prices) and financial quantities (money supply, credit aggregates, supply of government bonds, foreign denominated assets).
An important financial market prices channel is the interest rate channel. A contractionary monetary policy leads to a hike in nominal short-term interest rate, which given nominal rigidities (sticky nominal wages and prices) and the expectations model of the term structure, translates into higher real interest rates. The resultant increase in the price of currently purchased goods as compared with goods purchased in the future reduces investment and consumption and contracts output. As wages/goods prices adjust over time, real GDP returns to the potential level and the real interest rate and the real exchange rate also return to their fundamental levels.
The efficacy of the interest rate channel would, however, be reduced in case banks ration credit. But even though interest rates charged by banks do not change, the amount of lending by banks would increase if the supply curve of funds shifts to the right; to that extent, even with unchanged interest rates, the economic activity will have a positive effect.
The transmission of monetary policy through interest rates is augmented by changes in the exchange rates and balance sheet effects. Higher interest rates, for example, induce an appreciation of domestic currency which leads to lower prices of imports (the direct effect) and a reduction in net exports and, hence, in aggregate demand and output leading to a decline in prices (the indirect effect). Changes in interest rates impact on the balance sheet in a variety of ways. Higher interest outgoes, for example, lead to lower cash flows of firms and personal disposable incomes of households and, hence, reduce investment and consumption demand. Similarly, higher interest rates reduce asset prices and erode the net worth of borrowers restricting their ability to borrow.
On the other hand, the monetarist view of transmission argues that interest rate is just one of the many relative prices in the transmission mechanism. It is not just a single short-term interest rate but actual and anticipated prices on a variety of domestic and foreign assets that also undergo a change. Monetary impulses are, therefore, transmitted through relative price changes and changes in real money balances. The particular pattern of relative price changes varies from cycle to cycle and from country to country.
Empirical Evidence
Recent empirical research has confirmed the early findings of Friedman and Schwartz (1963) that monetary policy actions are followed by movements in real output that may last for two years or more. Furthermore, real effects of the monetary shocks are not only substantial but also long-lived (though not permanent) with the effects remaining up to three years (Romer and Romer, 1989). The recent vector auto regression (VAR) literature confirms these results: monetary policy shocks have a persistent effect on output while inflation displays an inertial response. Output, consumption and investment display a hump-shaped response, with the peak effect occurring about 1.5 years after a monetary policy shock. Inflation also displays a hump-shaped response, with the peak response after about two years (Christiano et al, 2001
The results for the euro area broadly conform to this pattern. The peak effect of output occurs after one year while inflation hardly moves during the first year. The delayed response of prices relative to that of output suggests that studying the transmission of policy to spending and output is a logical step, even if the aim of monetary policy is defined primarily or exclusively in prices (Angeloni et al, 2002). Although the persistence of inflation has declined per se in the US and the UK, the lags in the impact of systematic monetary policy action on inflation still persist despite numerous changes in monetary policy arrangements and advances in information processing as well as financial market sophistication (Batini and Nelson, 2002). A distinguishing feature of the euro area and the Japanese transmission vis-à-vis that of the US is that real output changes in the former are brought about largely by the response of investment in contrast to the predominant role of consumption in the latter. A comparative analysis of the alternative channels for the euro area, as a whole, suggests that the interest rate appears to be a very prominent, although not the dominant, channel of transmission; for some individual countries though, it turns out to be the dominant channel. For the euro area, the exchange rate channel is the dominant channel of transmission in the first two years, both in terms of its impact on output and on prices; from the third year onwards, the user cost of capital channel is dominant in terms of impact on output.
The ‘credit channel’ is found to operate significantly in Germany and Italy but irrelevant in some other euro area countries. Thus, the role of the banks is found to be smaller than expected. On the other hand, evidence for Japan indicates a strong role for the ‘credit channel’ since borrowers have been unable to substitute bank borrowing with alternative sources and consequently, business investment is especially sensitive to monetary shocks (Morsink and Bayoumi, 2001). Moreover, a money shock is found to have a large impact on economic activity even when the interest rate is included in the VAR; this suggests that the interest rate channel does not fully account for the transmission mechanism in Japan. The overall evidence, therefore, strongly suggests that short-term policy nominal interest rates do have real effects in the short- and medium-term.
The monetary policy transmission channels in emerging market economies have also been affected by the process of financial liberalisation (Kamin et al,1998). The withdrawal of state controls facilitated the emergence of an interest rate channel, although the credit channel remains important, especially in the context of occasional financial fragility. At the same time, the opening up of the economy together with the withdrawal of balance sheet restrictions enhanced the role of asset prices, in particular the exchange rate, in the monetary transmission process.
|
Monetary Transmission in India
For India, the empirical evidence showed that changes in money supply lead to changes both in output and prices, although the price effects of an increase in money supply are stronger than the output effects (Rangarajan and Arif 1990; Jadhav 1994). The Working Group on Money Supply (RBI, 1998b) found a strong unidirectional causation running from real output to real money. Besides, the output response operating through the interest rate channel turned out to be stronger and more persistent than that of the credit channel. A comparison of monetary impulses transmitted through interest rate effects and through liquidity effects for the period 1961-2000 indicates that the interest rate channel has emerged as a significant factor for explaining the variation in real activity in the 1990s as compared with its negligible impact in the 1980s (Dhal, 2000). The liquidity effect, although significant, diminished in terms of magnitude. This subsection examines afresh the transmission mechanism in India in the recent years (April 1994 to December 2002) vis-à-vis the 1980s (April 1981 to June 1990). The analysis is undertaken in terms of a vector autoregression (VAR) framework (Box V.3).
Box V.3
Transmission Channels for India: Methodology and Data Issues
In the literature, the two competing approaches in assessing the transmission mechanism are structural modeling approach and vector autoregression (VAR) model. The structural model approach sometimes suffers from the imposition of identifying restrictions and arbitrariness regarding the variables that are assumed to be exogenous (or pre-determined) (Sims, 1980). Moreover, because most of the macroeconomic variables are non-stationary in nature, estimating a structural model may provide spurious estimates and misleading forecasts. As an alternative, formulating unrestricted VAR models, treating all variables as endogenous in order to avoid infecting the model with spurious or false identifying restrictions is preferred. Although the VAR methodology has also been subjected to criticism, the method remains popular since it offers a straightforward solution to the simultaneity problem and appears to yield a reasonable characterisation of the economy’s response to monetary policy (Kuttner and Mosser, 2002).11 Unlike a macroeconomic model, the VAR facilitates quantifying the impact of policy shocks. This method also makes the lag structure and dominance of the channels transparent. In view of lack of a consensus on the workings of the transmission, the preference for VAR methodology in the recent literature, therefore, comes from the minimum restrictions that it places on as to how the monetary shocks affect the economy.
For the purposes of the present empirical exercise, the VAR model includes five endogenous variables in the following order: Index of Industrial Production (LIIP), Wholesale Price Index (LWPI), Non-Food Credit (LNFC), Broad Money (LM3) and Call Money Rate (CALL). The ordering of output and prices before the monetary policy variables appears reasonable in view of the well-accepted lags of the monetary policy actions on to output and prices; given the use of the monthly data, the ordering is all the more appropriate. Since the operating framework of the monetary policy has undergone changes during the period, as discussed earlier in this Chapter, from a monetary targeting approach to a multiple indicator approach in 1998, both money and interest rate are included in the VAR to capture the monetary policy shock. To assess the changes in the transmission that could have occurred in the aftermath of the structural reforms initiated in the economy during the 1990s, the VAR is estimated for the pre-reform (1981:04 to 1990:06) and the post-reform (1994:04 to 2002:12) periods separately. The intervening period 1990:07 to 1994:03 was marked by external payments imbalances, subsequent macroeconomic stabilisation and structural reforms and was, therefore, excluded from the analysis. In addition to the endogenous variables, the primary articles price index is included as an exogenous variable in the VAR, given the dominance of supply shocks as well as the procurement pricing approach in respect of agricultural commodities. The Bombay Stock Exchange index (LBSES) and the exchange rate of the rupee vis-à-vis the US dollar (LEXCH) are also included as exogenous variables in the VAR as a control for the growing globalisation and financial integration across various segments of the market. Finally, appropriate dummies have been used to control for abrupt oil price changes, spikes in the call money market (induced by the foreign exchange market volatility) and the impact of mergers. The prefix ‘L’ denotes the logarithms of the variables.
All the five variables in the VAR are found to be non-stationary. The ADF test statistics (with lag selection criteria based on the AIC criterion) were: 1.6 for LIIP (12 lags), 2.4 for LWPI (8 lags), 1.8 for LNFC (12 lags), 1.6 for LM3 (12 lags) and 0.5 for CALL (13 lags) for the period 1981:04 to 1990:06. For the second period (1994:06-2002:12), the corresponding t-statistics were: 2.3 (10 lags), 1.5 (6 lags), 0.8 (3 lags), 0.4 (1 lag) and 2.5 (3 lags). The 1%, 5% and 10% critical values are 3.5, 2.9 and 2.6, respectively. Given the non-stationarity of the variables, the option of differencing of series produces no gain in asymptotic efficiency and throws away information. However, if the variables are cointegrated, the VAR can be estimated in levels. Since the cointegration tests using the Johansen-Jusselius framework reveal that there exists at least one co-integrating vector for each of the sample periods, the VAR is, therefore, run in levels rather than the differenced form.
|
The results of the empirical exercise in terms of impulse responses for both the periods are generally on the expected lines. A positive shock to broad money (i.e. , expansion) over time leads to higher output while a positive shock to the call money rate ( i.e., increase) produces the reverse effect. The effect is, however, more pronounced and sharp during the first period as compared with the post-1994 period. In recent years, shocks to the call money rate take almost one year to have the expected negative effect on output, reflecting the monetary policy lags. A positive shock to non-food credit ( i.e., expansion) has the expected positive effect on output and the response during the post-1994 period is quicker. This indicates the role of a narrow credit channel in the Indian context and hence supports the continuingpolicy stress on the provision of adequate liquidity to meet genuine credit requirements to support investment demand. As regards prices, a positive broad money shock results in higher prices in the post-1994 period. However, in the pre-reform period, the outcome is a counter-intuitive fall in prices and, at no horizon does the effect become positive, perhaps reflecting the greater degree of administered pricing in the 1980s. The monetary policy tightening through a positive shock to the interest rate has the expected stabilising influence on prices in both the periods. The shocks to non-food credit have a positive impact on the prices, with the effect being weaker in the post-1994 period, supporting the policy rationale of meeting credit requirements with continuous vigil on prices.
An assessment of the impulse responses brings out certain differences in the lag structure of monetary transmission. The peak effect of the money shock on output occurs after 35 months in the post-1994 period (compared to three months in the 1980s) while the peak effect of the interest rate shock takes 32 months (14 months in the 1980s). On the other hand, the peak effect of non-food credit at nine months in the post-1994 period was the same as during the 1980s. As regards the lags on to prices, the maximum effect in the case of the interest rate shock occurs after six months in the post-1994 period (11 months in the 1980s); for non-food credit shocks, the peak effect takes place after seven months (four months).
Apart from impulse responses, the VAR analysis is also undertaken through variance decomposition. This measures the percentage of forecast error variance in one variable that can be attributed to innovations in a particular variable under consideration. The proportion of output variance due to innovation to broad money, at 60 months horizon, shows a perceptible decline from 19.7 per cent during the 1980s to 2.8 per cent during the second period. Similarly, the proportion of output variance due to interest rate shocks declined from 2.0 per cent to 0.2 per cent over the same period (Table 5.10). These results indicate that the role of monetary policy shocks in output variability has been substantially lower during the post-reform period. In other words, the monetary policy has been stabilising over the period in the sense that the endogenous response of monetary policy to macroeconomic developments has been to minimise fluctuations in economic activity. This result is consistent with the evidence emerging from the studies in the context of the US (Boivin and Giannoni, 2002). Another viewpoint, however, interprets the decline in the role of the monetary policy shocks as suggestive of the view that the monetary policy does not matter and the improved outcome is attributed to better inventory management (Kahn et al, 2002). As regards the role of monetary policy shocks on prices, the evidence is inconclusive. While the contribution of the interest rate shocks to the variability of prices declined from 4.1 per cent during the 1980s to 0.7 per cent during the second period, that of the money supply shocks increased from 2.1 per cent to 23.6 per cent over the same period.
In brief, the preliminary evidence from the above VAR analysis throws some interesting results. The impulse responses are more consistent with a priori theory during the post-reform period vis-à-vis the 1980s. The variance decomposition analysis shows that the role of monetary policy shocks has declined, at least, in regard to output fluctuations. At the same time, it is important to note that given the evolutionary, rather than discrete and abrupt, nature of financial and other structural reforms in the economy, their impact on the monetary transmission will become evident only over relatively long periods of time. The scope of the formal tests of structural change is rather limited and the assessment of the change in the transmission mechanism, as in this exercise, is confined to economic, rather than statistical, significance of the changes (Kuttner and Moser, 2002). Finally, the above empirical exercise is constrained by the use of industrial production as a measure of output in the absence of a reasonably long quarterly time series on total GDP of the economy. In view of the significant structural shifts towards the services sector and the interlinkages between agriculture, industry and services, the results of this empirical exercise should be considered as tentative and would need to be ratified with a comprehensive measure of output, as also by considering alternative techniques.
The monetary policy transmission mechanism in the future would have to grapple with the ongoing revolution in the payments and settlement system. The gradual e-monetisation can shrink cash demand and thus the monetary authority’s balance sheet, reducing seigniorage revenues and restricting open market operations. On the other hand, it is argued that the central bank’s ability to influence the nominal rate of interest is eventually an issue of political economy and the government can always require settlement through central bank money (Goodhart, 2000; RBI, 2002c). Besides, as long as monetary transactions take place through the banking channel - for example, a credit card payment is essentially an advance from a bank - the spread of e-monetisation does not, per se, require redefinition of monetary and financial aggregates (RBI, 1998b).
|
