There's no doubt that 2009 was a rough economic year for the United States. Facing a slowing economy, the government created the "Cash for Clunkers" program, which offered a $3,500 or $4,500 credit to consumers who traded in a less fuel efficient (<18 MPG) vehicle and purchased a more fuel efficient (>22 MPG) vehicle. Proponents argued that by providing immediate economic incentives for people to purchase cars, the government would 1) stimulate demand for cars and 2) kick-start the economy. Now, I'm not disputing the first part of the argument: if (in real terms) you make something relatively less expensive, more people will buy it, and thus demand will necessarily increase. However, there's a problem with the second part of that argument. Cash for Clunkers only really makes the something relatively less expensive "today". Prices go back to normal "tomorrow"--thus, the only effect should be that consumption that would have occurred "tomorrow" will now happen "today" (referred to as an intertemporal shift). In theory, the net effect is zero, but the cost of the program is $3 billion. Do the data support my argument?
Click to enlarge
A little bit of background information: Cash for Clunkers officially started on July 24, 2009, and ended on August 24, 2009. Consistent with the U.S. Census Bureau's data, we see a surge in motor vehicle sales in August. In September, we see a significant decrease in auto sales. Note that both the sales increase in August (+7.75%) and sales decrease in September (-14.1%) are the highest percentage changes in the past ten years, significantly different from the averages of +0.02% and -1.40%, respectively.
Now, I can't prove with a simple graph like this that all Cash for Clunkers did was shift consumption forward by one month, having no net effect yet still incurring a cost. I acknowledge that there are alternative explanations for the graph: for example, perhaps consumers waited during September, hoping that the program would be extended; when they realized it wouldn't be, they made their purchases in October, accounting for the unusually high percentage increase in October sales. Though this "waiting effect" is likely present, it's unlikely to be of a larger magnitude than the intertemporal consumption-shifting effect. But furthermore, even if it was, it wouldn't mean that Cash for Clunkers was successful--rather, the opposite! If these people were going to buy cars regardless of the credit, it makes no sense to subsidize their purchase because the idea behind the credit is to create new demand in the hope that this new demand will kick-start the economy. At the point where the program is just subsidizing demand that already existed, it fails at stimulating the economy.
In sum, this graph isn't meant to be a substitute for rigorous analysis on the costs and benefits of the program. Rather, it's a supplement for this kind of analysis, which has found that overall, Cash for Clunkers cost taxpayers $2,000 for each of the 665,000 cars that qualified. Even though this wordplay is tired and overused, I'm going to say it: Cash for Clunkers was a clunker.
Note: this post was inspired by a similar analysis which I found to fall short because it looked at non-seasonally adjusted data (though it claimed to be seasonally adjusted) and didn't provide a historical context for comparison. The problem with using non-seasonally adjusted data is that one could easily counter that auto sales always fall in September because that's when new models are released, etc.