In the coming days critical data will be announced: first October industrial production index, than third national income for the third quarter and November capacity utilization rate, and finally unemployment data for September. Two types of comparison will be made using the first three: comparison with the same period last year and with the previous quarter.

Making comparison with the year before can sometimes prove problematic particularly in periods where we expect upward or downward turn. One way to overcome this problem is assessing figures in comparison with the previous month (with previous quarter for quarterly data).however, it the series we examine show fluctuations for reasons other than economic reasons, such a comparison does not have any meaning as the growth/contraction rates resulting from such a comparison reflect the mixture of two elements: economic factors and seasonal factors. But we are interested only in economic factors. For instance, take tourism revenues: it rises in summers and drops in winters. So, is the fall in revenues in winter a normal drop owing to the season or is it a result of a rise in hotel fees or a reduction in the interest of foreigners in Turkey; or maybe both?

National income, industrial production, and capacity utilization data evidently involve seasonal movements. For instance, in the first quarter national income of Turkey always falls compared to the fourth quarter. Moreover, in the fourth quarter a prominent drop in comparison with the third quarter is seen. This applies for the other two series declared monthly. However, we see that after the announcement of the capacity utilization rate and industrial production index particularly, press compares figures with the month before. This of course leads to misinformation.

In that case, it is necessary to utilize series net of seasonal effects. But this also incurs some problems: in real life, we also observe the series itself; we do not know the seasonal and non-seasonal impacts. Therefore, we have to separate the series in 'one piece' in reality into 'two pieces' using statistical methods. And as these two pieces do not exist in reality, we have to estimate their values. We can reach different results depending on the method we use to let aside seasonal effects. And it is clear that this way we will reach different growth rates in comparison with the previous period.

Let me exemplify: Take the real value of Turkey's GNP. There are two well known adjustment methods. If you use the older one known as X12, you see that GDP decreased by 0.2 percent from the first quarter to the second quarter of 2007. But according to the Tramo-Seats method which is more recent, GNP 'appears to' have increased by 1.7 percent in the same period. So, what is it? Did Turkey's economy grow or contract in that period? Such a situation can also be observed for the first quarter of 1980. For the first quarter of 2008 both methods indicate that the economy contracted. However, the former suggests 4.4 percent contraction while the latter suggests 1.7 percent: this is a considerable difference.

Actually this is observable for all indicators which come in 'one piece' in real life but which must be divided into more than one piece. For instance, to determine the growth and contraction cycles of the economy, GDP data must be net of seasonal movements as well as the trends. The remaining gives the cyclical part we are interested in. An academic study conducted in early 1990s calculated different cycles for GDP using different trend calculation methods. The interesting part is it showed that there is a negative relationship between different cycles of the same series. That is one method said that the economy was expanding while the other signaled a tightening. Put yourselves in the shoes of economic policy makers for instance who think that the economy is in expansion and tighten monetary and fiscal policy!

Thank God that no one makes such mechanical decisions. Nonetheless, we should still be careful in assessing the results derived via such estimation methods. For example, raw data on industrial production index, GNP, and capacity utilization rates should not be immediately compared with the quarter before. Second, if the growth (contraction) rate calculated using series net of seasonality is low; 'deep' conclusions like 'growth revived or contraction again  shall not be made.

This commentary was published in Radikal daily on 03.12.2009