I have been reading another academic economics blog, Overcoming Bias.  The author, Robin Hansen, is an Associate Professor of Economics at George Mason University. According to the tag line, “Overcoming Bias is economist Robin Hanson’s blog, on honesty, signaling, disagreement, forecasting, and the far future.”

Robin has written a number of posts arguing that many decisions in business and personal lives are ways of signaling otherwise hidden traits.  As an example relevant to accountants, in Just a Handshake, Robin talks about a paper showing that proposing a contract without penalties for every possible contingency allows the principal to signal that she trusts the agent, which can make the parties better off.

I haven’t seen many signaling models like this in the accounting literature lately.  Why not?

One obvious answer is that there simply aren’t that many people trying to publish any game-theoretic models in accounting these days.  But that may be a symptom, more than a cause, and moreover, it seems there are very few signaling models, even when measured as a fraction of the small body of analytical research.

In defending his own view that signaling models are useful, Robin Hansen may (unintentionally) give us an explanation for their flaws.  He writes (in the post linked above):

There are literally hundreds of papers out there showing how signaling can or does explain various details of human behavior.  In fact, fifteen years ago my Ph.D. thesis advisor tole me not to write such papers, because there were already so many of them that they weren’t very interesting. Yet people keep complaining everytime I mention a signaling explanation of something, that I’m too free with such explanations.  So I’m stuck between an academic discipline that considers such explanations too obvious to be worth publishing, and an audience that finds them too implausible to believe, even when backed by such publications.

My guess—and I would love to hear from the modelers who are more knowledgeable than I am—is that actually BOTH audiences feel that Robin is too free with signaling explanations, because signaling models don’t impose a whole lot of discipline on the modeler.  Any time someone wants to say ‘such-and-such action is a signal of something else,’ they need merely find some hidden attribute of the agent, and construct some way that people strong on that attribute will find the action less costly than those weak on that attribute.  In the classic example by Ross, academic success is less costly for smart people, so even if they don’t learn anything in school, smart people will go on to higher education simply to distinguish themselves from others who find academic success more costly.

The education model might seem sensible, but how hard is it to create signaling models for other attributes?  The answer is ‘not very’!  Basically any action that is differentially costly to different people (or firms) with different hidden attributes, and visible to others, can be tied into a signaling model.  Since wealth is a hidden attribute, and spending money is less costly to the rich than the poor, any action that costs money can be a signal of wealth–even if the signal has no direct or intuitive relationship to the message that the signaler is trying to convey.

We therefore run into the problem of way too many possible signals.  Suppose you wish to posit that hiring a more expensive investment bank for your IPO is a signal that your long-term prospects are strong.  Well, why would firms choose that signal in particular?  Why not put gold leaf on the prospectus, or simply film the CFO burning cash on the front steps of corporate headquarters for the six o’clock news?

Checking Robin’s blog again today, this new post seems to suffer from this type of problem.  Robin apparently views the entire pantheon of political activity as just a way to for individuals to strut their stuff. Key quote:

“In large polities, the main function of our politics in our lives is how it influences the way others see us; its influence on us via policy is far weaker.”

Thus, the ability of politically active people to change society, or to simply direct perquisites their own way, takes a back seat to their ability to show that their otherwise hidden characteristics of intelligence, hard work, persuasiveness, etc.

Allow me to pose a few questions to modelers.  First, do you agree that signaling models are fairly rare these days, or am I just not looking in the right places?  Have I indeed captured the problem with signaling models?  Or are there other issues?  Finally, has anyone succeeded in imposing more discipline on signaling models, in a way that would allow us to say “action X is a more reasonable way to signal hidden attribute Y than action Z?” even if Z satisfies the basic requirements of a possible signal (different costs for those with different hidden attributes, and visibility to the target audience).