One reason why people pay their taxes is that they’re afraid of being caught, and fined, if they cheat.
In the simplest possible model of tax evasion, people only comply with the tax code if the benefits of evasion – the savings in taxes paid - are less than the costs – the probability of getting caught times the penalty if caught.
The major problem with this simple model (and more sophisticated variants of it) is that it doesn’t fit the facts. Far fewer people evade taxes than one might expect, given that the probability of audit is fairly low (not that IRS, CRA or any other revenue agency will reveal the actual number), and the penalties for small-scale avoidance are not all that high.
Why doesn’t the simple model work? One hypothesis is that people pay taxes because they believe others are paying taxes: they contribute because it’s fair. Another hypothesis is that people are really lousy at estimating the probability getting audited.
Gareth Myles at University of Exeter is doing some fascinating research, along with co-author Nigar Hashimzade, on the effects of social networks on tax evasion. They hypothesize that people form estimates of the probability of being audited based on their own experience, and the experience of their co-workers, friends and neighbours. If I don’t know anyone who has been audited, then I will believe that the probability of being audited is low. On the other hand, if I get audited, or if my friend, neighbour or co-worker is audited, then suddenly my own perception of the probability of being audited will shoot up.
It seems to me that this research has a fairly obvious policy implication: revenue agencies should audit people who have lots of friends, and who will tell all of them that they are being audited. The more people know about an audit, the more effective it will be as a deterrent to others.
But how can a revenue agency know how many friends a person has?
The Canadian Community Health Survey asks respondents the following question:
About how many close friends and close relatives do you have, that is, people you feel at ease with and can talk to about what is on your mind?
It turns out that the size of a person’s social network is significantly related to a number of readily observable economic and demographic factors. The complete results of my analysis using the 2009-10 CCHS are available here; the associated stata .do file is here.
If tax collection agencies want to audit people with large social networks, they should begin by auditing males. I was surprised at the gender effect, but it's quite strong, and persists after controlling for a variety of factors. The average male between 20 and 64 in the CCHS sample has 8.5 close friends and family; the average female has 7.6. Now it could be that, even though women have fewer friends, those friends are closer, but simply looking at the reported number of friends, men have more.
What else? Predictably, higher income people tend to have larger social networks, as do people whose main source of income is employment or self-employment. (Unfortunately since evaders by definition are those who are underreporting their income, targeting high income individuals has some limitations as an audit strategy.) Married people have more friends and family than people living common-law, or singles.
The number of close friends and family a person reports declines with age until people are in their 40s, then picks up again a little, but never approaches the size of the social networks people have in their 20s. Auditing people young, when they will share the experience with many friends - hopefully via Facebook or Twitter - would be a good strategy for a revenue agency aiming to increase people's perceptions of the probability of being audited.
Immigrants, and especially recent immigrants, tend to have significantly smaller social networks. I would suspect, though I can't test it with the data, that people who move within a country also tend to have smaller social networks than those who stay put. Hence targeting stayers rather than movers might be a good strategy for a revenue agency looking to audit people with lots of social contacts.
Of course all of this presumes that the experience of being audited leads to greater tax compliance. If people's reaction to being audited is "it was no problem, the revenue agency didn't work out that I earned $20,000 in cash last year and didn't report it", then the experience of being audited might tend to decrease compliance - and the more people who know the auditing is unlikely to lead to seriously averse consequences, the more serious the tax compliance problem is likely to become.