If you are like many people, you are perhaps puzzled how the government's most significant measure of inflation, the consumer price index (CPI), showed inflation slowing down in April to 0.2% (a pace of 2.4% per year) from 0.3% (a pace of 3.7% per year) in March. What about the price of food and gas? In fact, the most amusing statistic in the report is that according to the Bureau of Labor Statistics (BLS), gas prices fell 2.0% from March to April. In the same time period some saw unleaded gasoline go from $2.19 per gallon to $2.35 per gallon. So what is happening here?
The CPI numbers quoted above, as well as by most of the news media, represent seasonally adjusted (SA) numbers. Economists tend to look at most things through the lens of SA numbers. At certain times of the year, prices are expected to go up for a number of reasons. For example, the price of heating oil typically goes up in the winter as people need the commodity to heat their homes. It is for these reasons that economists typically adjust the month to month changes to try to gauge the main trend.
This is great in theory, but there is a big problem with this: seasonal trends are very hard to define. The seasonality exists alongside major trends from year to year, and the two effects are inherently difficult to separate. There are techniques that can be used to attempt to separate them (like the autoregressive integrated moving average used by the BLS), but each of these techniques do involve statistical error. This is the main reason that, in the real world away from economics textbooks, individual stock analysts and business managers avoid seasonally adjusted numbers if at all possible.
A more stable way of analyzing economic data, and data in general, is looking at year over year percent changes. When companies report quarterly earnings reports they nearly always speak in terms of year over year increases in revenue and earnings, not seasonally adjusted increases from the prior quarter. If you are comparing time period to time period, the seasonality is automatically adjusted for and one gets the true underlying trend of the business. By comparing current the year over year change to that of the prior period, one then can get a better sense of the underlying trend.
Applying this principal to the CPI, one can pull up the full press release from the BLS webpage (www.bls.gov) and look for the raw non-seasonally adjusted (NSA) numbers. For example, the unadjusted numbers for gasoline were up 20.7% from levels a year ago. In contrast, the year over year change in March was 26.0%. The report also indicates that the unadjusted percent change was 5.6% from March to April, which would probably be closer to recent experiences at the pump. So the overall message is that prices in gasoline in April were still going up, but not at as steep a rate as in March.
The most interesting analysis was on the headline CPI number. The stock market rally on the morning of the release was generally attributed by the media to the belief that inflation had been meaningfully reduced, with the annualized rate derived from the SA number down to 2.4%. However, according to the raw NSA data, not only were prices up 0.6% in absolute terms in April versus March, but the year over year growth in inflation was 3.9%. This compares to a 4.0% year over year increase in March. Considering the measurement errors in the statistic, the rate of inflation was essentially unchanged and still high in April.
This is a prime example of how economists' calculations can give a totally distorted view of the world. This is why next time you are looking at important economic numbers, examine the raw NSA numbers and take the SA numbers with a grain of salt.