On Twitter, Russ Roberts said something I tend to agree with:

Just a curious coincidence that economists who like stimulus want bigger government and those who oppose it prefer smaller. 

In fact, he said something very similar back in 2011:

The evidence for the Keynesian worldview is very mixed. Most economists come down in favor or against it because of their prior ideological beliefs. Krugman is a Keynesian because he wants bigger government. I’m an anti-Keynesian because I want smaller government. Both of us can find evidence for our worldviews[.]

Now, Roberts doesn't actually know Krugman's motives - he's necessarily making a guess. But he definitely does know his own! Basically, he's saying that he adopts an anti-Keynesian stance not because be thinks stimulus actually fails to fight recessions, but because he wants to shrink the size of the government in the long term. 
Furthermore, he strongly implies that he will selectively display evidence against the effectiveness of stimulus as a stabilizer, in order to ward off the long-term expansion of government. That's motivated reasoning.
I have long believed that stuff like this happens in economics arguments all the time. It's probably not (usually) an intentional subterfuge, but more of an unconscious bias. An economist is presented with the proposition that countercyclical fiscal policy (stimulus in recessions, austerity in booms) increases overall efficiency. He is generally against increasing the size of the government. So when he sees that countercyclical fiscal policy will (temporarily) increase the size of the government in a boom, it triggers warning sirens in his subconscious. "What? Increase the size of government? Never!", says his subconscious. And so that motivates him to argue against the proposition that countercyclical policy is effective at stabilizing booms and recessions. 
So Russ Roberts is being honest and introspective, which is good. Introspection is difficult and often unflattering, so not enough people engage in it.
In fact, this is similar to the first part of Paul Romer's "mathiness" argument. Romer argues that some economists make their modeling choices for reasons related to academic politics - for example, he says that researchers who want to believe in a world without market power will construct growth theories that make silly assumptions just to avoid putting market power in the equation. 
This effect is a lot stronger if there's no skin in the game. As Matt Yglesias pointed out back in 2011, you see Republican politicians - who probably want to shrink the government - doing Keynesian policies fairly often. Bush enacted stimulus in 2008, and probably Reagan in the early 80s. 
Another example would be the fact that nearly everyone on Wall Street was eager to tell you back in 2012 that QE would cause a big rise in inflation. But when you looked at TIPS spreads, it was clear that the marginal investor wasn't putting his money where his mouth was.
This is consistent with the finding that partisan belief gaps go away when you pay people to get things right. A bet really is a tax on bullshit (although not the optimal tax). Or, as Nassim Taleb so memorably put it, macrobullshitting is reduced by having skin in the game. The cynical side of me says that Romer's "mathiness" manifests mainly in fields where the data is not good enough to exert discipline on theory.
So when you read econ arguments, always be a little wary of the motivations of the arguers.

P.S. - This is unrelated to the main point of my blog post, but an alternative, non-Keynesian theory of stimulus is that it boosts output because it expands the government, in the ways that need expanding (public investment).

Updates

Paul Krugman responds to Russ Roberts.

Adam Ozimek responds to Roberts, defending Paul Krugman, and the econ profession, from Roberts' cynical allegations. The two have an interesting (but brief) twitter debate about how evidence should interact with ideology.

Roberts writes an additional tweet that I like:

Conceding the reality of self-deception isn't cynical. It's realistic. Leads to humility and caution.

I agree, but just because some nonzero degree of self-deception is inevitable doesn't mean it's benign. Instead of just accepting it, I say people should try to fight against it. And if you realize that you yourself engage in a considerable degree of self-deception, I say you should focus on reducing it, rather than focusing on demonstrating that your rhetorical opponents are equally self-deceptive.