Edmunds' Unrealistic Forecasts Hurting Auto Industry

Edmunds.com economists have come out with an evaluation of the Cash for Clunkers programs that states only an additional 125,000 vehicles were sold at a cost of $24,000 per incremental vehicle.

To arrive at this number Edmonds’ economists made forecasts of sales with and without Cash for Clunkers. While the reasons for doing this analysis are unclear (their conclusion hurts the chances of the auto industry getting additional stimulus) it is clear that the forecasts they made are either too high or too low.

In general, most people do a poor job of forecasting sales (95% of new products fail) but it is easy to track obvious bias in forecasts if you have ever had your bonus dependant on their accuracy. As a brand manager, I was carefully beaten about the head and shoulders by management to make sure I knew about forecasting errors and applied this learning to my subordinates when it was my turn. Let’s see how good Edmunds’ economists are at forecasting.

Here’s how they forecast sales would be with and without Cash for Clunkers.

Note that they forecast that Cash for Clunkers would permanently depress demand as shown by sales being lower than their “without” forecast in December, 4 months after the program end. This was a promotion, not an advertising campaign so changes to demand would have been restricted to people who were immediately considering buying a vehicle. Any event that is still affecting sales more than 6 or 8 week later is something that changes the basic demand.

The best way to check the accuracy of a promotion forecast is to compare it to unpromoted products whose sales move similarly (and for similar reasons to minimize the chance correlation is unrelated.) In this case, we have retail sales excluding auto from the Census Bureau which tend to go up and down similarly to auto sales, though auto sales are much more depressed. Because the past two years have not been typical, I compared sales versus 2006.

Here's the chart:

Total retail sales excluding autos are about the same level as 2006 but auto sales are running about 25% behind. The month-to-month trend show that auto sales tend to go up and down when other retail sales do. Edmonds forecast without Cash for Clunkers (in red) assumes that auto sales would have suddenly accelerated without Cash for Clunkers.

The trend line for retail sales ex-autos is overlaid on the forecast to show that Edmonds forecasted a significantly faster increase in auto sales than other retailers were experiencing.

(Note: October AUTO, OTHER MOTOR VEHICLES sales based on manufacturer sales reports. Other data from US Census.)

Now it’s pretty easy to show bad performance if you assume it is bad which is what Edmonds has done. Let’s hope that auto industry sales recover and they don’t need any more stimuli as this does not help their case.

Comments

[jfb:]

I think your forecast is less accurate than Edmunds.com's. You failed to lay sufficient background work supporting your theory of auto and retail sales following the same long term pattern. Edmunds.com's forecast is more believeable since there is recent historical evidence that a short term promotion doesn't do much more than pull-ahead sales from future months. Future month's sales are thereby negatively impacted by those that were pulled ahead. Just ask any marketing person from Ford, GM or Chrysler how this works. A suggestion to you is that you shouldn't spout off about subjects you know little about.

Nov 05 07:54 AM

[One Eyed Guide:]

I had a nice chat with an auto writer from a major newspaper and he pointed out that I failed to mention why the Edmunds forecasts were wrong. They based them on nonpromoted vehicles which are luxury and large suv sales. These move very differently than the total auto market so they ended up with a forecast that NADA said was wrong.

I did not actually make a forecast, just pointed out a more accurate basis in retail sales ex-auto. I'll fix this with another post that compares the different cost estimates, Edmunds, NADA, and CEA.

All forecasts are bad, some are just worse than others.


On Nov 05 07:54 AM jfb wrote:

> I think your forecast is less accurate than Edmunds.com's. You failed
> to lay sufficient background work supporting your theory of auto
> and retail sales following the same long term pattern. Edmunds.com's
> forecast is more believeable since there is recent historical evidence
> that a short term promotion doesn't do much more than pull-ahead
> sales from future months. Future month's sales are thereby negatively
> impacted by those that were pulled ahead. Just ask any marketing
> person from Ford, GM or Chrysler how this works. A suggestion to
> you is that you shouldn't spout off about subjects you know little
> about.

Nov 05 08:27 AM

[User 373847:]

Another consideration is the fact that the program was announced long before it was implemented so it likely took sales from months previous to the program. Oct. sales imply there wasn't many forward sales stolen. Edmunds evaluation seems pretty biased to me. I believe they had a conclusion and backed the data into it.

Nov 05 09:44 AM

[Dennis Virag:]

The situation is much more complex that presented. Since the program was announced early, sales in May and June were depressed. Cash for clunkers cleared out inventory, resulting in increased production plans for the 4th quarter. These production increases will not only helped the automakers but it will also help the supplier base. What most people don't understand is nearly 60% of suppliers do not meet their loan covenants, creating a potentially disastrous situation for both the supply base and all automakers due to the intertwined supply chain. While it created news and the resulting debate, the Edmund's analysis was very sophomoric in nature.

Nov 05 12:43 PM

[casey00001:]

You have provided nothing to disprove Edmunds. My money is on them. At least they are impartial.

Nov 05 04:55 PM

[karenc:]

Cash for clunkers didn't help the car industry, it just cost taxpayers, the poor and car donation charities a lot of money.

Nov 07 11:55 AM