Friday, April 11, 2008

Empirical evidence on the impact of INR/USD changes on the WPI

Surjit Bhalla has written an article in Business Standard questioning the empirical evidence on the link between INR/USD appreciation and the WPI. Here's the empirical evidence, and the steps required to replicate the numerical results:

  1. The mechanism of exchange-rate pass-through (ERPT) critically depends on low trade barriers. There was less ERPT in the older data, given the bigger quantitative restrictions and tariff barriers which were in place. Hence, let's focus on the latest 10 years, i.e. 120 observations of monthly data. Define the reading for each month as the average of the month (i.e. the default notion of `monthly' data in CMIE's Business Beacon database). Owing to lags of data release, this works out to the time period from Dec-1997 to Dec-2007.
  2. Start with the differences of logs of series, and do seasonal adjustment. Think of this as the time-series of point-on-point percentage changes in the series, but seasonally adjusted. The seasonal adjustment is not essential to the result, but it's logically sound, and helps to improve the statistical significance.
  3. We're going to do a two-variable vector autoregression. Four different criteria for selection of the lag length of the VAR (AIC, HQ, SC, FPE) all agree that the right lag order is 1. Estimate the VAR.
  4. Testing for granger causality shows one-way causality from INR/USD changes to WPI changes. H0: INR/USD does not cause WPI has a prob value of 0.005954, i.e. you can be pretty sure that's a null you would reject. This is one-way causality - on the other direction the prob value is 0.5514.
  5. Compute the impulse response function, orthogonalised, cumulative, using bootstrap inference. The dashed lines are the 95% confidence interval.

Click on the picture to see it more clearly. What does it say? (a) Over a horizon of roughly 4 months, a 1% shock to the INR/USD yields a 0.2% change in the WPI, and (b) The 95% confidence interval is all above 0; you can reject the null of no-effect at a 95% level of significance.

This same result comes out nicer when using weekly data.

This is a quick take on the problem. For a more thorough attack on it, and a review of the literature and international evidence on ERPT, look at this paper on exchange rate passthrough by Rudrani Bhattacharya, Ila Patnaik and myself [paper] [slideshow], a draft of which was presented at the 2nd Research Meeting of the NIPFP-DEA Research Program on Capital Flows and their Consequences.

6 comments:

  1. The impulse responses seem to be based on a 'short run' VAR in first differences. But I think WPI and exchange rate would be cointegrated (at levels) in which case the VAR is not very informative. Both short run and long run effects need to be analyzed from the VECM model alone.

    ReplyDelete
  2. Yes, the paper that's mentioned in the last para of the blog entry does do VECMs on this.

    ReplyDelete
  3. Precisely! Which is why I think the unrestricted VAR estimates don't mean much since the ECT is significant in the VECM.

    ReplyDelete
  4. It is really amazing that surjit bhalla is not even understanding pre.k.g elements of economics about imported inflation. He is in such a big position. GOD SAVE OXUS.

    ReplyDelete
  5. However I have noticed a drastic difference between the result presented here and the PDF file, which is named here for ref. Look at page 15 of that file. Here the IRF shows that there is short term response at abt 0.1% of WPI due to 1% impact on ex. rate and this effect diminishes (almost) after 10-15 month. Whereas here IRF shows that the same effect is persistent for long period. So which result is correct?

    ReplyDelete
  6. Inr conclusion has implications for the usd. Good explanation

    ReplyDelete

Please note: Comments are moderated; I will delete comments that misbehave. The rules are as follows. Only civilised conversation is permitted on this blog. Criticising me is perfectly okay; uncivilised language is not. I delete any comment which is spam, has personal attacks against anyone, or uses foul language.