As you may recall, the United States implemented a huge package of stimulus – about $787 billion worth – in 2008, President Obama’s first year in office. Included in that package was a fair amount of temporary tax cut measures, increased funding for programs like unemployment and food stamps, and a large list of spending on various projects identified as “Good Investments”. It all sounds great – but how could we measure the eventual effects if they don’t match the model – economics has no control group.
This is relevant for two reasons:
- President Obama at one point proposed $447 billion in new spending along similar lines to the original stimulus (do the math, that will be $1.2 trillion in ‘stimulus’, not including the plethora of other stimulus programs like Cash for Clunkers) .
- The Congressional Budget Office declared the original stimulus a success based upon the models which were used to propose the stimulus.
Why Does a Control Group Matter?
That last statement is extremely important, not just for intellectual curiosity but also in how it shapes future policy.
The CBO scored the stimulus in the same way the stimulus was scored when it was proposed. Administration projections predicted unemployment would peak at less than 8% with the stimulus. Of course, the actual maximum reached was 10.1% in October 2009.
What did that say about the stimulus? Pundits argued everything from ‘the economy was worse than anyone predicted‘ to ‘the stimulus was too small‘ (Ironically, the lower bound of stimuli ranges Krugman recognized and wanted in Economist Christina Romer’s work was $1.2 trillion). All of this from a near $800 billion stimulus which was proposed by a Democratic Congress as (‘only’) a $500 billion stimulus.
Which leads us to my point: I won’t argue the stimulus did nothing.
Money certainly was spent and there were definitely some stimulative effects. The problem I have is with saying that the stimulus was the best option, with saying that without the stimulus as designed we would be much worse off, or that the effect of the stimulus on the economy was greater than what was spent in the stimulus bill (declaring that there was a stimulus multiplier greater than 1+ the cost of interest).
First, declaring a stimulus multiplier greater than 1 carried to its intellectual extreme means that a tax rate of 100% directed to stimulus would create more than 100% of economic activity – something that doesn’t pass the smell test (while the opposite extreme, a 0% tax rate would increase economic activity, does seem to still make sense).
The problem with the stimulus was the same models that were used on the front end and declared unemployment would peak at 8% are being used to declare that the stimulus worked, and it prevented unemployment from reaching some even more extreme number. I’ll show you the problem with a simple 3 point graph.
PK Stimulus: A Proposed Stimulus Package from DQYDJ
I predict unemployment will go from 7% to 8% to 9% to 10%, and my model declares that unemployment is growing at a rate of 1 percentage point a year.
My proposed stimulus will reduce the growth rate by 1 percentage point a year – meaning unemployment will be 7%, then 7%, then 7%, then 7%. No brainer, huh?
Well, let’s say PK’s stimulus was enacted, and unfortunately the unemployment rate got worse (13%!) even with the stimulus.
However, the Budget Office, which was asked to score the stimulus based upon the original model, declares the stimulus a rousing success and pundits say, “if only it was bigger”/”if only the economy was as good as predicted”.
Here’s how the graph looks at the end (Year 0 is now the current year, and negative years are in the past.) Commentary concentrating on the fact that maybe the stimulus funds could have been spent elsewhere for more productive purposes (ahem, broken window theory) is derided or ignored.
So, to recap, the creators of Paul’s stimulus said it would reduce unemployment’s growth rate by 1 percentage point a year, and they predicted unemployment would grow at 1 percentage point a year. Therefore, when unemployment grew at 2 percentage points a year, instead of declaring that the stimulus wasn’t successful, the writers said that the unemployment growth rate would have been 3 percentage points. Therein lies the problem; a stimulus is always a success when judged using the original equations used to draft a stimulus.
Even in the case the unemployment rate goes to 100%, drafters will argue that it would have happened in a shorter time-frame without the stimulus – there is no way to escape the argument. According to Reason Magazine, the director of the CBO, Doug Elmendorf, responded to the question “If the stimulus bill did not do what it was originally forecast to do, then that would not have been detected by the subsequent analysis, right?” with, “That’s right. That’s right.”
Say it With Me: “Economics has no Control Group”
The problem, of course, is that economics has no control group. Economics has a tough time proving theories since the most convenient comparison is to something before and after some policy goes into effect. It is one of the reasons Economics has so much theory, and real world examples so disparate. Even when theories are widely accepted, it is hard to find perfect policy examples. One way to make synthetic control groups is to compare, say, types of stimulus enacted in various countries (say: do nothing, cut taxes, increase spending) versus the effect on the economies in aggregate. Although it doesn’t prove that one type of stimulus is best (“well, the U.S. is different from New Zealand because…”), it certainly makes a strong argument for specific policies. Also, it prevents the “best known” argument that “it would have been worse”.
(Sure, smart guys and gals, tell me about Econometrics in the comments – then explain how it applies to the United States’ economy, that is the global hegemon and issuer of the world’s reserve currency, as a whole.)