Don't trust seasonal adjustment in July

The U.S. Bureau of Labor Statistics has reported the estimates of various consumer price indices for July 2011. Unexpectedly, the difference between the core and headline CPI has fallen to zero after a crucial turn to growth in June. There are many factors which allow us to consider the difference for July a blip.

The difference between the core and headline CPI, as shown in Figure 1, is for seasonally adjusted, SA, values. As a rule, seasonal adjustment allows to separate trend and fluctuations for a given month. For example, it is good to know that temperature in January and July is two degrees above their normal climatic values, but these values are different for January and July. Same is with the seasonally adjusted consumer prices – they depend on previous years. When seasonal factors are correlated over years the seasonally adjusted values are informative and help the broader audience to ignore constants.

However, when one value used in the adjustment for a given month is a spike then all further values are biased. This is the case for the headline CPI in July 2011. Figure 2 compares the seasonally adjusted headline CPI and not seasonally adjusted CPI, NSA, between 2001 and 2011. The NSA curve shows a small increase in July 2011. The large difference between the SA and NSA curves from May to July 2011 (also clearly seen in 2010) is driven by the turbulence in consumer prices in the same months in 2008 and 2009. Since August, the effect of 2008 and is opposite and the SA values will be below the NSA ones. Specifically, the SA-NSA effect is a major contributor to the fall in the difference between the core and headline CPI in July 2011.  Hence, do not trust the seasonally adjusted CPIs in July.

Figure 1. The evolution of the difference between the core and headline CPI since 2002.

Figure 2. The evolution of the difference between the core and headline CPI, both SA and NSA, since 2001.

Other factors implying a blip in the July’s difference between the core and headline CPI are related to actual changes in consumer prices. In August 2011, the price index of energy has to fall by approximately 10% as one can see from the price evolution during the past 20 days. Technically, this fall will reduce the headline CPI by at least 1% with the core CPI intact. Apparently, the fall in energy prices will indirectly affect other consumer prices which might result in a slightly larger change in the core CPI. A heavy crop in 2011 is also foreseen in the U.S. and Europe what will likely suppress the current growth in the price index of food in 2011-2012.

All in all, we foresee the difference between the core and headline CPI to grow through rest of the year. Since the core CPI will be growing at its current pace between 1% and 2% per year the headline CPI will sink below zero.

Figure 3 briefly repeats our concept of sustainable (quasi-linear) long-term trends in the difference between the headline and core CPI in the U.S. There were two clear periods of linear behaviour: between 1981 and 1999 and between 2002 and 2009. A natural assumption of the future evolution of the difference was that a new trend has to emerge around 2010 after a short period of very high volatility (see Figure 1). However, the difference is very volatile also in 2011.  There is no sign that the higher volatility will calm down any time soon.

Figure 3. Linear regression of the difference between the core CPI and CPI for the period from 1981 to 1999 (R²=0.96, the slope is 0.67) and a regression of the difference between the core CPI and CPI between 2002 and 2009 (R²=0.91 and the slope is -1.59).