Monday, October 22, 2012

September 2012 Consumer Prices

Consumer price inflation continued to slow marginally in September (log annual and monthly changes; 2000=100):

01_indexes

It’s now been just over a year since we last saw an increase in consumer inflation, and the current rate of increase is nearly two thirds less. What inflation there is, is being driven by the usual suspects – food and transport costs, although food prices have held steady between August and September (log annual and monthly changes; 2000=100):

02_food

Instead, September’s pain was delivered almost single-handedly through higher transport costs (log annual and monthly changes; 2000=100):

03_trans

It’s not hard to figure out why: RON97 petrol was bumped up 10% to RM3.00 per litre on September 6. The other big mover was medical care and health, which rose 0.5% in log terms in September alone.

Just out of curiosity, I checked the elasticity of the transport cost index against global crude oil prices (in this case, West Texas Intermediate), and the relationship is statistically strong – a 1% increase in the WTI price results in a 0.048% increase in transport prices.

That may not sound like much, but a USD12 movement per barrel like we’ve seen over the past three months, should have caused a 0.6% increase in petrol prices, instead of the 0.36% that’s actually occurred. That’s evidence of the distortions in the domestic market caused by the fixed price of RON95 petrol.

Having looked at this, I’m wondering whether the government’s petrol subsidy estimate for this year might still be too conservative at RM25 billion.

Technical Notes:

September 2012 Consumer Price Index Report from the Department of Statistics

2 comments:

  1. Hey Hisham,

    Colleague spotted your WTI v Transport CPI estimate on your blog and got a bit curious. I ran the (admittedly, simple excel) regression and the R-square and the estimator (0.048% on your sample) is not consistent - i.e. it changes depending with the sample time frame.

    Sorry, didn't mean to discount your results, I just got curious and had to point it out.

    ReplyDelete
    Replies
    1. Jason, the sample was 2000m1 to 2012m9 - I didn't try any robustness checks, it was just an on-the-fly estimate. Now that I've checked, it does move around a lot doesn't it?

      One thing I might add is that I used an AR(1) term to correct for serial correlation, which substantially lowered the estimate (not something you can check on in Excel I'm afraid).

      Delete