This Selected Issues paper and Statistical Appendix examines the channels of monetary policy transmission in Thailand. The main findings are that changes in monetary policy are associated with changes in real output, and that the main channel for transmission is not bank lending but asset prices. The paper takes stock of the performance of the Thai corporate sector emerging from the crisis and discusses remaining challenges and vulnerabilities. An assessment of Thailand’s fiscal vulnerability is also presented.

Abstract

This Selected Issues paper and Statistical Appendix examines the channels of monetary policy transmission in Thailand. The main findings are that changes in monetary policy are associated with changes in real output, and that the main channel for transmission is not bank lending but asset prices. The paper takes stock of the performance of the Thai corporate sector emerging from the crisis and discusses remaining challenges and vulnerabilities. An assessment of Thailand’s fiscal vulnerability is also presented.

II. The Channels of Monetary Policy Transmission in Thailand1

A. Introduction

1. Recent changes in monetary policy highlight the importance of understanding the monetary policy transmission mechanism. Following the crisis, the Bank of Thailand (BOT) progressively eased the stance of monetary policy. Short-term interest rates declined from over 20 percent in early 1998 to about 1.5 percent by mid-2001. However, bank-lending rates adjusted slowly and private credit grew sluggishly. Given continuing problems in the corporate and financial sectors, an important question is whether the transmission mechanism is working, and, if so, through which channels. This paper addresses these questions by using vector auto-regressions to study the transmission of monetary policy.

2. The main findings of this paper are:

  • Monetary policy is associated with changes in real output. A typical shock to the key policy rate of the Bank of Thailand (equal to 1.7 percent, as identified in the model) has a peak impact of about 1 percent on real GDP after 4 to 6 quarters.

  • Bank-lending is not transmitting monetary policy—asset prices are playing an important role. Changes in the policy rate weakly affect the supply of private credit, which in turn only modestly affects output. However, interest rates affect asset prices and corporate balance sheets, which in turn affect firms’ investment decisions.

3. The following policy implications follow from the analysis:

  • Accelerating reforms of the banking system will strengthen the transmission mechanism. As the current economic turnaround unfolds and the economy picks up steam, credit demand will likely rise. An acceleration of ongoing reforms in the financial sector will allow the bank-lending channel to become more fully operative and prevent supply side constraints.

  • While asset prices appear to propagate monetary shocks, monetary policy should not target asset prices. Asset prices cannot be fully controlled by central banks in general. Ineffective attempts at targeting asset prices can therefore undermine the credibility of the monetary framework.

4. The paper is organized as follows: the next section briefly discusses the different channels through which monetary policy affects economic activity; section C presents the evidence; and the last section summarizes the results and discusses the potential for further work.

B. The Channels of Transmission

5. The channels through which monetary policy works include the following:2

  • Interest rate channel. This is the standard Keynesian channel of monetary transmission. Contractionary monetary policy raises short-term nominal interest rates. Real interest rates rise as well due to price rigidities in the short-term. The increased cost of financing depresses expenditure on business investment, housing, and consumer durables, leading to a fall in aggregate demand and output.

  • Exchange rate channel. A rise in domestic interest rates attracts foreign capital which appreciates the real exchange rate and leads to a fall in net exports. However, the exchange rate can also have opposite effects on domestic demand when substantial corporate debt is foreign currency denominated. A depreciated exchange rate weakens balance sheets, leading to a fall in investment spending and output.

  • The “credit” view. As one of the more recent contributions in the literature on monetary transmission, the channels grouped under this heading emphasize problems of information asymmetries in credit markets. Two channels under this heading include the bank-lending channel and the balance sheet channel. The former arises when problems of information asymmetries prevent some firms, especially small firms, from tapping securities markets directly for their financing needs. Banks play a special role in financial markets by maintaining close relationships with borrowers and overcoming these information asymmetries. A monetary contraction which reduces bank reserves causes banks to reduce the supply of loans. Since such firms do not have access to other sources of borrowing, this leads to a fall in investment spending. The balance sheet channel focuses on the weakening of corporate balance sheets when interest rates rise and free cash flows fall. Lower net worth borrowers create problems of adverse selection (borrowers in effect are offering lower collateral) and moral hazard (borrowers’ have less stake in the firm) for creditors, causing them to reduce the supply of lending. Balance sheet effects can also directly affect firms’ planned investment spending by reducing retained earnings available for financing investment.

  • Asset price channel. This channel emphasizes the behavior of equity and other asset prices. A monetary contraction is generally associated with a fall in equity prices, as investors move from stocks to bonds. To the extent that firms compare the market value of capital to its replacement cost (Tobin’s q theory of investment) in making investment decisions, this reduces investment expenditures, and hence output.

C. Evidence

6. The channels of monetary policy transmission can be studied using vector autoregressions (VARs). A large empirical literature, primarily on the U.S., Europe, and selected other countries, uses VARs to study transmission issues.3 This methodology has both strengths and weaknesses. The key strengths are that it allows for flexible dynamic modeling of the key macro series of interest and affords a useful graphical way to present the results. Impulse response functions (IRFs), through which VAR results are generally presented, show the dynamic response of the variables in the system due to a shock in one of the variables of interest. The shape of the IRFs shows the magnitude and persistence of the effects of shocks. The key weakness is that VARs, by their nature, are somewhat a-theoretic and it is difficult to identify exogenous monetary policy shocks. Also, IRFs show the response to unanticipated changes in monetary policy, as identified in the model. In using the results to consider actual policy changes, due attention has to be paid to the extent to which the policy change is not already anticipated.

7. This paper first estimates a baseline model and then considers extensions to study the different channels discussed above.

Baseline Model

8. A baseline VAR consisting of real GDP, the consumer price index, and the policy rate (14-day repurchase rate) was estimated on quarterly data from 1993:1 to 2002:1. Output and prices are seasonally adjusted and expressed in log levels.4 The IRFs can therefore be interpreted as percent deviations from baseline. A measure of world interest rates (the U.S. federal funds rate) was additionally included as an exogenous variable, since it is likely to be one of the important determinants of domestic rates.5

9. The VAR was identified by assuming a standard ordering of the variables: output, prices, and interest rates. The identification scheme is important for computing the impulse responses. If shocks to the different variables were uncorrected the impulse responses could simply be computed from the estimated coefficients in the underlying OLS equations. However, in such macro models, shocks to one variable are likely to occur with predictable changes in other variables. For instance, a negative aggregate supply shock would cause both a positive price shock and a negative output shock. An identification scheme imposes a partial structure on the model which allows for the extraction of the exogenous component of the shock. The scheme used here, known as a Cholesky decomposition, is common in the literature and recursively identifies exogenous movements in the variables with the first variable in the ordering treated as the most exogenous and the last as the least. Thus a monetary shock is identified as that part of the change in the interest rate which cannot be explained by contemporaneous changes in output and prices.6

10. The response of real GDP to a one standard deviation change in the policy rate (1.7 percent) peaks at about a 1 percent deviation from baseline after about 1½ years. The estimated 95 percent confidence interval shows that unanticipated changes in the policy rate produce statistically significant changes in the path of output.7 Additional measures of the monetary stance yield similar results. When the VAR is re-estimated using both the repurchase rate and money base (MO) as measures of monetary stance, a consistent picture for the output response emerges. Real GDP responds negatively to an interest rate shock and positively to a money base shock. However, a variance decomposition analysis—which seeks to determine which variable accounts for most of the forecasted variation in output—shows the interest rate to be a more relevant measure of the monetary stance.

A02ufig01

Response of real GDP to one S.D. shock to Interest rate

(+/− 2 standard errors)

Citation: IMF Staff Country Reports 2002, 195; 10.5089/9781451836783.002.A002

11. Impulse responses with respect to the other variables in the system show reasonable patterns. The price level declines immediately following an output shock, reflecting a movement down the aggregate demand curve. Subsequently, as demand picks up, prices rise, reflecting a movement along the aggregate supply curve. Impulse responses of the policy rate are consistent with a Taylor-rule type framework: rates rise in response to a positive output and price shocks. The price level rises following an increase in the policy rate. This surprising response of prices to a monetary tightening is a frequent result in the VAR literature on monetary transmission. It can arise if the central bank uses additional information for judging inflationary pressures which is missing from the model. As the central bank expects an adverse shock it tightens rates while at the same time prices rise.

12. Investment responds more than private consumption to an interest rate shock. Decomposing real GDP into private consumption and total investment and estimating similar VARs shows that the peak response is one percent and 4 percent respectively.8 The greater impact on investment is consistent with the greater role of financing in investment expenditures, compared to consumption. As the economy matures and consumer credit becomes more important, the contribution from consumption is likely to grow.

A02ufig02

Response to one S.D. shock to interest rate

Citation: IMF Staff Country Reports 2002, 195; 10.5089/9781451836783.002.A002

13. The crisis period complicates interpretation of the results. However, a similar VAR estimated on post-crisis monthly data shows consistent results. The tumultuous movement of the key macro series during the crisis represents the operation of other factors not inside the model. At the same time, the limited number of time series data available prevents dealing with the issue in a sophisticated manner. One simple way is to examine if the output response holds in the post-crisis period. A similar VAR was estimated on monthly data for 1999–2001. As GDP is not available on a monthly basis, the manufacturing production index was used as an indicator of activity. Since the large variation in the series during the crisis may be complicating the estimated relationships, the start of the post-crisis period was chosen late enough to allow for interest rates to have adequately come down from the crisis highs. The accompanying figure shows a consistent output response. Moreover, the price puzzle, discussed above, also disappears, indicating that it was driven by the sharp rise of interest and inflation rates during the crisis period.9

A02ufig03

Response of manufacturing output to one S.D. interest rate shock, monthly data 1999-2001

Citation: IMF Staff Country Reports 2002, 195; 10.5089/9781451836783.002.A002

Bank-lending channel

14. There is little evidence that monetary shocks are transmitted through bank-lending. The baseline VAR was augmented to include private credit extended by commercial banks, adjusted for debt write-offs and transfers to asset management companies.10 This allows an examination of how the policy rate affects bank-lending, and in turn, how bank-lending affects output.11

  • There is weak transmission from the policy rate to bank-lending. In fact, bank-lending rises immediately following a rate increase, and then declines. Such a relationship could arise if the policy rate may in part be reacting to developments in credit markets. Indeed, when the bank-lending variable is ordered first in the VAR, to measure the interest rate shock as exogenous to contemporaneous movements in private credit, the positive effect disappears. Private credit reacts very weakly and after a considerable lag to changes in the policy rate. As the figure shows, a one standard deviation change in the interest rate has a peak effect at a 10-quarter lag of roughly 1½ percent. In cumulative terms this amounts to a 5 percent reduction in bank-lending after two years. When scaled by the standard deviation of bank-lending over the period, this is equivalent to a 0.17 standard deviation fall in bank-lending—a small effect.

  • Shocks to bank-lending have weak effects on output. On the other hand, bank-lending reacts to output shocks. Taken together, these impulse responses imply that demand side factors may be more important, than supply factors, in determining credit flows.

A02ufig04

Response of bank-lending to one S.D. shock to interest rate (+/−2 standard errors)

Citation: IMF Staff Country Reports 2002, 195; 10.5089/9781451836783.002.A002

15. A useful way to demonstrate the significance of bank-lending in the transmission process is to compare the impulse responses with and without the bank-lending channel operating. The bank-lending channel can be “closed” by preventing shocks from endogenously propagating through this variable. Operationally, this amounts to not estimating an equation for bank-lending, but treating it as an exogenous variable in the equations for the other variables. A big difference in the impulse response would indicate that bank-lending is an important channel.12 The figure shows that this channel contributes little to the output response. A different criterion to evaluate the significance of bank-lending is to decompose the forecasted variance of an output shock into the various channels—bank-lending accounts for only 9 percent of the total variation in output at a 2 year horizon.

A02ufig05

Response of real GDP to one S.D. shock to interest rate

Citation: IMF Staff Country Reports 2002, 195; 10.5089/9781451836783.002.A002

Asset price effects

16. Asset prices play an important role in the propagation of monetary shocks. The baseline VAR was augmented in a similar fashion to include the SET stock market index as a proxy for the movement in asset prices. The impulse responses show that stock prices react significantly to an unanticipated interest rate shock and that output in turn expands after an unanticipated rise in the stock market. A one standard deviation interest rate shock leads to a maximum deviation of 4 percent in the stock market index from baseline after about 6 quarters. The cumulative effect is a 18 percent reduction in the index (equal to a 0.4 standard deviation fall), a much bigger effect than the private credit response discussed above. In turn, a one standard deviation rise in the stock market index is followed by a peak one percent rise in output from baseline, an effect similar in magnitude to the direct effect of interest rate shocks on output. The figure shows the impulse responses with and without this channel operating. The output response is considerably muted and of a short duration. Asset prices account for about 30 percent of the variation in output at a 2-year horizon.

A02ufig06

Response of real GDP to one S.D. shock to interest rate

Citation: IMF Staff Country Reports 2002, 195; 10.5089/9781451836783.002.A002

17. Interest rate changes also directly affect firm balance sheets and hence the net worth of firms. A rise in interest rates increases the interest expense of firms, reducing cash flow and net worth of firms. This can directly impact output by reducing investment expenditure financed from retained earnings. The accompanying figure indeed shows the co-movement between interest rates and three indicators of corporate liquidity: the interest coverage ratio (defined as the earnings before interest and taxes divided by interest expense), quick ratio (ratio of liquid assets to short-term liabilities), and interest expenditure scaled by total assets. These indicators can be incorporated into the VAR framework, although the results have to be interpreted with caution since the small sample of firms may not be completely representative.13 Results show that while indicators of liquidity do decline in response to an interest rate shock, the response of output to liquidity indicators is weak. Closing the corporate balance sheet channel produces some impact on the impulse responses—however, a variance decomposition shows that very little variation can be attributed to movements in the liquidity indicators.

A02ufig07

Interest rates and corporate liquidity indicators

Citation: IMF Staff Country Reports 2002, 195; 10.5089/9781451836783.002.A002

Exchange rate channel

18. Evidence on the exchange rate channel is difficult to interpret and reveals mixed results. Interpretation is complicated by the change in the exchange rate regime during the sample period. Results therefore have to be interpreted with caution. The accompanying figure shows the impulse responses of output to the interest with and without the real effective exchange rate and indicates significant impact of the exchange rate channel. However, a variance decomposition analysis shows that the real exchange rate accounts for less than 10 percent of the variation in real output up to a 10 quarter horizon.

A02ufig08

Impulse response of real GDP to one S.D. shock to interest rate

Citation: IMF Staff Country Reports 2002, 195; 10.5089/9781451836783.002.A002

Summary Model

19. A direct comparison of the asset price and exchange rate channels shows the relative importance of asset price effects. A summary model was estimated to directly compare the contribution of asset price and the exchange rate channels. The appropriate metric to evaluate the relative importance of the channels is through a variance decomposition of real GDP. Such an exercise revealed that the asset price channel accounts for the greatest variation in output (40 percent) at roughly a 10-quarter horizon.

D. Conclusion

20. The collective picture of the transmission mechanism which emerges from this analysis is one where supply effects in private credit do not appear to be working. Instead interest rate changes appear to directly impact firms’ investment decisions, either through a Tobin’s q type channel as the market value of firms falls or directly by weakening corporate balance sheets and reducing available retained earnings for financing investment.

21. Resolution of financial sector problems will contribute to the importance of the bank-lending channel. The current high levels of distressed assets on bank balance sheets have made banks highly risk averse, focusing only on the best credit risks. Quick restructuring of the remaining problem loans will allow the banking sector to support growth.

22. While the analysis has pointed to the importance of asset price effects, monetary policy can be more effective if it focuses on the nature of the shocks. Changes in asset prices can have important effects on aggregate demand and thus should be closely followed to evaluate the stance of monetary policy. An examination of the nature of asset price shocks and whether they are considered to be temporary or permanent would assist in devising the appropriate monetary policy response.14 In general, however, it is difficult for monetary authorities to identify asset price bubbles. It is also difficult for them to control asset prices since stock markets in particular are volatile and often move for reasons unrelated to monetary policy. If the central bank is perceived as trying to target the level of stock prices, in addition to other objectives, and is proved ineffective at doing so, it will likely undermine the credibility of the monetary framework. For these reasons it is problematic for monetary policy to target the level of asset prices.

23. Although the analysis has provided useful insights into the transmission mechanism, further work would enhance our understanding of different channels. One particular area would be to further examine the demand and supply side factors in private credit. Micro-data based evidence would nicely complement the findings in this paper.

Bibliography

  • Bemanke, Ben S., and Ilian Mihov, 1998, “Measuring Monetary Policy,Quarterly Journal of Economics, Vol. 113 (August), pp. 869902.

    • Search Google Scholar
    • Export Citation
  • Bernanke, Ben S., and Mark Gertler, 1995, “Inside the Black Box: The Credit Channel of Monetary Policy Transmission,Journal of Economic Perspectives, Vol. 9 (Fall), pp. 2748.

    • Search Google Scholar
    • Export Citation
  • Clements, Benedict, Zenon G. Kontolemis, and Joaquim Levy, 2001, “Monetary Policy Under EMU: Differences in the Transmission Mechanism,IMF Working Paper 01/102 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Cochrane, John, 1998, “What do the VARS mean? Measuring the output effects of monetary policy,Journal of Monetary Economics, Vol. 41 (April), pp. 277300.

    • Search Google Scholar
    • Export Citation
  • Kirakul, Suchada, 1996, “Interest Rate Administration and Liberalization and Money Market Development in Thailand,” in Mehran, Hassanali, Bernard Laurens, and Marc Quintyn (eds.), Interest Rate Liberalization and Money Market Development (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Mishkin, Frederic S., 1995, “Symposium on the Monetary Transmission Mechanism,Journal of Economic Perspectives, Vol. 9 (Fall), pp. 310.

    • Search Google Scholar
    • Export Citation
  • Mishkin, Frederic S., 2001, “The Transmission Mechanism and the Role of Asset Prices in Monetary Policy,NB ER Working Paper no. 8617 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation
  • Morsink, James and Tamim Bayoumi, 1999, “A Peak Inside the Black Box: The Monetary Transmission Mechanism in Japan,IMF Working Paper 99/137 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Patrawimolpon, Pichit, Theerapol Rattanalankar, Chutathong Charumilind and Pruchya Ngamchant, 2001, “A Structural Vector Autoregressive Model of Thailand: A Test for Structural Shifts,Bank of Thailand, Monetary Policy Group (Bangkok).

    • Search Google Scholar
    • Export Citation
  • Ramaswamy, Ramana, and Torsten Sloek, 1997, “The Real Effects of Monetary Policy in the European Union: What Are the Differences?IMF Working Paper 97/160 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Santiprabhob, Veerathai, 2001, “Bank of Thailand’s Inflation Targeting: Performance and Future Challenges,paper presented at the Bank of Thailand Symposium, Bangkok, Thailand, June.

    • Search Google Scholar
    • Export Citation
  • Sirivedhin, Tanya, 1998, “Financial reform and the monetary transmission mechanism; the case of Thailand,” in The Transmission of Monetary Policy in Emerging Market Economies (Basle, Switzerland: Bank for International Settlements).

    • Search Google Scholar
    • Export Citation
  • Taylor, John B., 2000, “Alternative Views of the Monetary Transmission Mechanism: What Difference Do They Make for Monetary Policy?Oxford Review of Economic Policy, Vol. 16 (Winter), pp. 6073.

    • Search Google Scholar
    • Export Citation
1

Prepared by Reza Baqir.

2

More elaborate discussion of the channels can be found in Mishkin (1995, 2001) and Taylor (2000), amongst others.

3

For a review of the empirical U.S. literature, see the Fall 1995 symposium in the Journal of Economic Perspectives. Morsink and Bayoumi (1999) present evidence for the importance of bank balance sheet effects in Japan. Clements et. al. (2001) find the interest rate channel to be the most important in the Euro area. Related work on Thailand includes Kirakul (1996), Sirivedhin (1998), Patrawimolpon et. al. (2001), and Santiprabhob (2001).

4

This specification follows the convention in the literature. Unit root test for most of the series do not reject. However, first-differencing often results in eliminating too much information. Moreover, since there is likely to be at least one cointegrating relationship in the series, and given problems in correctly estimating the cointegrating vector, estimating the model in first differences can lead to a mis-specified model. This is the principal reason why the literature generally uses a log-levels specification.

5

Given the small number of observations, the smallest possible lag order which adequately captures the dynamics is preferable to preserve efficiency in estimation. Two commonly used criteria to guide the choice of the lag order, the Akaike information criterion and the Schwaz criterion, indicate use of one lag. The impulse responses of the baseline VAR reported below remain qualitatively the same with two lags.

6

Such an identification scheme can also be interpreted as a reaction function for the central bank where the central bank observes information on output and prices in the current period before deciding the stance of monetary policy.

7

A different scaling of the data, with output expressed as ratio of potential output (as in Morsink and Bayoumi (1999), shows a similar response.

8

The series used here is total investment—private investment which was not available on a quarterly basis for the entire period. Annual data, however, show that private investment was 72 percent of total investment over the sample period.

9

The impulse response peaks earlier with the monthly data. Part of this may be explained by the use of the manufacturing production index which misses out on the lagged effects on the services and other sectors of the economy. All other impulse responses are by quarter.

10

Private credit was ordered last in the Cholesky decomposition as it is potentially one of the channels through which shocks to the interest rate affect the economy.

11

Bernanke and Gertler (1995) caution against interpreting movements in credit aggregates as evidence for or against the credit view of monetary transmission. They point out that credit flows are often counter-cyclical and that even though credit flows may not change, the terms of credit can move in ways predicted by the credit view.

12

The interest rate channel is treated as a residual—it is assumed to account for the output response which cannot be attributed to other channels.

13

To partially address concerns from use of an unbalanced sample, the median ratio is used for all three indicators as it is less sensitive to outliers than a simple average.

14

Mishkin (2001), amongst others, discuss the role of asset prices in the conduct of monetary policy.

Thailand: Selected Issues and Statistical Appendix
Author: International Monetary Fund
  • View in gallery

    Response of real GDP to one S.D. shock to Interest rate

    (+/− 2 standard errors)

  • View in gallery

    Response to one S.D. shock to interest rate

  • View in gallery

    Response of manufacturing output to one S.D. interest rate shock, monthly data 1999-2001

  • View in gallery

    Response of bank-lending to one S.D. shock to interest rate (+/−2 standard errors)

  • View in gallery

    Response of real GDP to one S.D. shock to interest rate

  • View in gallery

    Response of real GDP to one S.D. shock to interest rate

  • View in gallery

    Interest rates and corporate liquidity indicators

  • View in gallery

    Impulse response of real GDP to one S.D. shock to interest rate