Thursday, July 21, 2011

The (Non)Transmission of Monetary Policy (UPDATED)

While twiddling my thumbs last Saturday waiting for a function to start, I started reading this recent paper from the IMF on the monetary transmission mechanism. Just for the heck of it, I started applying the methodology used in the paper for Malaysia, and got some interesting results.

I’m not going to go the whole hog just yet replicating , as the data requirements are pretty extensive and I’ll need to do a lot more reading of the available literature before doing a full assessment, but the initial step (as taken in the paper) is a simple calculation of elasticities. Specifically, how much impact does a change in the policy rate affect the range of interest rates offered by the banking sector and the interbank/money market? In the terminology, what is the extent of the pass through of policy rates into other interest rates such as deposits and loans?

The basic elasticity calculation is a simple OLS regression:

y = α+βx+ε

…where x is the log of the first difference of the policy rate, and y is the log of the first difference of the particular interest rate we’re looking at.

I’ve calculated the elasticities across two different sample periods – 1998:9-2005:7 and 2005:7 to 2011:5 – corresponding to changes in BNM’s monetary policy regime. Since the pass-through between policy rates and the interbank rate is more or less complete, I’m using the overnight interbank rate as a proxy for the policy rate across both samples.

Starting with deposit rates, we find some curious results:

Interest Rate 1998:9-2005:7 2005:7-2011:5
Savings Deposits 0.293 0.586
Fixed Deposit 1 month 0.825 0.639
Interbank 1 month 0.937 0.941
Fixed Deposit 3 month 0.832 0.630
Interbank 3 month 0.843 0.738

Interpreting this is straightforward – taking the interbank 3 month rate as an example, a 1% rise in policy rates resulted in a 0.738% increase in the 3 month interbank rate after July 2005.

Looking at the range of results obtained, you can see the differences. While pass through of monetary policy changes is high for interbank interest rates, and little changed between the two periods, there are some pretty dramatic differences for the rates offered to customers. The pass through for savings rates for instance improved considerably between one period and the next, but in both periods is fairly incomplete. The FD rates are harder to fathom, as the pass-through, although high, actually fell after July 2005.

Yet, if deposit rates are hard to figure out, the effect on lending rates is pretty shocking, to me at least:

Interest Rate 1998:9-2005:7 2005:7-2011:5
BLR 0.360 0.364
Average Lending Rates 0.256 0.246
MGS 1 year 0.505 0.583
MGS 3 year 0.125* 0.007*
MGS 5 year 0.038* -0.260*
*Not statistically significant

The problem here is not the changes between periods – not much to speak of – but the fact that it is so incomplete. I’ve included here MGS, because 3-5 year MGS is the benchmark for corporate borrowing in the bond markets. Yet the pass through of policy rates into market driven 1 year MGS yields is actually better than the impact on the lending rates offered by the banking system.

These numbers call into question the appropriateness of using an intermediate interest rate target to affect output and inflation, as changes in the policy rate don’t appear to affect the cost of credit a whole lot. Perhaps other channels of transmission are in play – that would involve modelling everything out, and seeing the impact in each channel (e.g. the exchange rate, asset prices). But as it stands, relying on interest rates alone to affect growth and inflation would require much greater than the 25 basis point moves that BNM has been making.

This is worth investigating in greater detail.

[UPDATE]:

Forgot to add the link for the paper!

Technical Notes:

Acosta Ormaechea, Santiago ; Coble Fernandez, David O, "Monetary Transmission in Dollarized and Non-Dollarized Economies: The Cases of Chile, New Zealand, Peru and Uruguay", IMF Working Paper No 11/87, April 2011

6 comments:

  1. interesting post bro hishamh

    Is the picture of interbank money market complete?

    Not sure how the FX rules are now but zaman dulu folks used to borrow USD and Swap it back to MYR using the USD/MYR Overnight Swap market to cover MYR short positions. Sometimes its much cheaper to enter this trades than borrowing direct in O/N money or even term.

    I kinda expected the effect on Lending side,cause these Bankers semua driven by Greed anyway..believe one must go deeper into the type of Credit Exposure and the Rating Exposure group instead of stopping at BLR alone....

    Not many folks as well have implemented proper Fund Transfer Pricing on a consolidated basis to be able to factor all these movements and react by either adjusting rates or shifting portfolio exposure..

    ReplyDelete
  2. Bro satD,

    There's not a whole lot of action for interbank deposits over 3 months, so it's hard to say. But judging by the 3-month rate, changes in the OPR are largely being properly absorbed.

    I think one of the things going on for the lending side is dis-intermediation. Because corporates are borrowing more from the bond market and less from banks (current ratio is something like 1 to 2), loan portfolios are falling as a ratio to bank balance sheet size. That puts less pressure on lending rates as the main source of profits, and more pressure on treasury departments to trade and find yield.

    But this means the ability of central banks to influence the cost of credit is being diminished, because 1) banks aren't the only players on the bond market, and 2) yields are more market driven, and securities cover a far longer maturity.

    I need to collect some data and put together some kind of model to figure this out. There's a few potential channels involved (not just interest rates) in monetary policy transmission, and it would be interesting to see which one's the strongest for Malaysia.

    ReplyDelete
  3. Hisham,

    This is really cool stuff. I may be asking something silly here, but I was wondering if the BLR is considered more of a longer-term rate. Typically, borrowings based on the BLR are for the long term (i.e. housing loan, though I am not sure what other kinds of borrowings follow BLR). Based on this assumption (I could be wrong), perhaps it is not so surprising that the pass through rate is relatively small. Also, I don't think that a change in the O/N rate typically shifts the yield curve upwards by a parallel amount across all maturities. Usually, long-term yields tend to relatively more "stable" compared with short term rates. I am using terminology very loosely here, but I hope you understand what I mean.

    I also don't know how the BLR would be perceived on the yield curve as it is not exactly a bond.

    Based on the first table, it appears to be a significant difference between the elasticities for the FD rates and the interbank rates. Just wondering if you tested the significance?

    Thanks in advance for your input.

    ReplyDelete
  4. Shihong,

    Yes, BLR (and the average lending rate) are longer dated than interbank rates. But based on the paper I was reading (sorry, just put the link up!), 25% is on the low side - New Zealand gets around 62%, and Chile near 67%, which seems to be about the level for advanced economies. The low pass-through is more in keeping with a poorly developed financial system, not typical of one as advanced as ours.

    My preliminary workup using just GDP, inflation and the interbank rate suggests that monetary policy moves are having an effect, and in the desired direction, but whether this is through the policy rate itself, or through some other channel, is something I have to work out.

    On the last point, are you talking about differences between the two variables, or just the significance of the coefficients? The latter's automatically generated by the software I use.

    ReplyDelete
  5. Thanks Hisham for the link.

    On the last point, I was referring to the difference between two variables.

    ReplyDelete
  6. Only the difference between the 1 month rates test as significant - the standard deviations are between 0.05-0.075.

    ReplyDelete