Seasonal adjustments: why bother? Without some accounting (or at least an acknowledgement) of the seasonal influence, it makes no sense to talk about month-to-month movements in wholesale prices. For example, if someone says wholesale prices fell from September to October, that’s not news – prices invariably do. The question we want to know is whether they fell more, or less, than normal.
Some analysts try to avoid the seasonal adjustment issue by focusing on year-over-year changes. But, that gives no insight into short-term trends. And, if the year-ago period was abnormal, the results are distorted.
Seasonal adjustments: how are they done? The seasonal adjustment process generally involves sophisticated statistical routines. And, when adjusting for retail unit volumes or revenues, the process is further complicated by the need to account for trading days, holidays, calendar shifts, and significant industry or economic shocks.
Since the Manheim Index is adjusting for average wholesale prices, our statistical analysis has shown it unnecessary to include these other factors. Thus, we are left with a purely statistical routine (Census X-12, multiplicative, and default trend factor) with no judgmental input.
Seasonal adjustment routines “learn”, so to speak, and give higher weighting to more recent time periods. So, without outside inputs, our methodology can produce factors that are slightly distorted, on a temporary basis, by major shocks. That’s because the calculation will misinterpret part of the shock as a seasonal force. To handle this situation, one can include dummy variables to represent one time events like 9/11, industry strikes, or 2008’s financial market collapse. To date, we have not found that exercise to materially influence the statistical stability, significance, or size of the seasonal adjustment factors. Thus, we have left them out.
The seasonal factors. The graphics below show the seasonal adjustment factor for each month averaged over the past 16 years and the specific factor for selective years. To read, note that a seasonal adjustment factor of .98 means that, all other things equal, wholesale prices in that month are 2% below the average of all months. A factor of 1.02 would mean 2% above, etcetera.
Keep in mind that these seasonal adjustment factors were developed after prices had already been adjusted for changes in mix and mileage. Clearly, the seasonal pattern of sports cars differs from that of SUVs.
Over time we have seen a slight moderation in the seasonal differences between months and a relative improvement in third quarter pricing. Possible explanations for that include:
- model year introductions being more evenly spread over the year,
- better management of carryover inventories,
- a more geographically diverse buyer base due to online buying, and
- a larger share of retail sales (and, thus, wholesale buying) coming from dealers located in less seasonally-influenced areas of the country.
As noted earlier, we have not included a dummy variable to account for the collapse in the financial markets in the fourth quarter of 2008. That has artificially reduced the seasonal factor for November and December of 2010.