Month-on-month figures suggest that deflation in Hong Kong may have ended.
I started my public service career as a statistician. This lasted for five years. One of the first things that I came across, in internal debate about the role of the statistician in presenting and interpreting statistics, was a favourite saying (attributed to the British prime minister Benjamin Disraeli) about "lies, damned lies and statistics". There are so many ways of presenting and using figures that one should always exercise great care, particularly when important decisions are to be taken on the basis of the figures. During my years as a statistician I developed the habit, as much as time permits, of looking at the raw data as well as the interpretation that has been put on them, in order to satisfy myself that I am not missing something.
For those closely monitoring the economy, particularly developments on the price front, it may be worth doing just that, now and over the next few months. The regular press releases issued by the Census and Statistics Department typically feature the year-on-year changes as the headline message. For example, in the latest January press release, the CPI figures suggested that overall consumer prices fell by 1.9% in December 2003 from a year earlier, narrowed further from the 2.4% decrease in November 2003.
The impression that a layman reader will get from these figures is probably that "deflation in Hong Kong continues, although it may be less bad than before", or "Hong Kong is still experiencing a rate of deflation of 1.9%". To be sure, the press release does not say those things in those terms. The word "deflation" is not even used once. But this is how the figures are reported in the media and this is the impression many people get.
I do not know what readers are looking for when the monthly figures on consumer prices are released. For me, I wish to know whether, in the month that has just past, consumer prices have continued to fall and, if so, at what rate. And so when I am told that "overall prices fell by 1.9% from a year earlier" I wish to know when this fall amounting to 1.9% took place. Did it take place in the first half or the second half of the past year, or in recent months? May be prices have been falling in the first part of the year by, say, 3% and have since increased by 1.1%, so that the net decrease is 1.9%. If so, when did prices begin to turn back up? And how fast have they been increasing?
That is why I go one step further to look at the level of the Index and its movements from month to month, noting of course that there are seasonal and special factors that affect the figures for individual months. Seasonal adjustments to price indices are never complete and there is something called statistical noise that even statisticians cannot explain fully. And so one should never establish a trend or a turning point by looking at figures of one or two months. But when, for example, intuitive thinking suggests that a turning point may be near, one could benefit greatly by a closer look at the monthly changes.
What does all this statisticians' talk lead to? Consumer prices in Hong Kong may have stopped falling in around August 2003. They seem to have been increasing since. Longer observation is needed before a firm conclusion can be drawn, but deflation, taking the word to mean an ongoing (or may be even a spiralling) process of falling prices, may be over, and the turning point may have been in August 2003.
The Composite Consumer Price Index for December 2002 was 93.8. It fell almost continuously during the first part of 2003 and by August it reached the lowest level on record of 91.0. Since August, however, it has been increasing, and by December 2003 it reached 92.0. There has therefore been an increase over the period from August to December 2003 in the Composite Consumer Price Index of 1.1%. At this stage I would not try to annualise the increase and start to talk about a return of inflation. Turning points in economic data such as prices are difficult to identify and the behaviour of economic data at turning points is always erratic.
Joseph Yam
29 January 2004
Related Viewpoint Article:
Click here for previous articles in this column.
Document in Word format
