The Case for Bathroom Mandates

After walking through SF

Walking around San Francisco has me thinking about public bathrooms. It seems to me that the provision of them will suffer from a collective action problem, and that it is very easy to slide into an equilibrium where you can’t find a bathroom anywhere, even paid bathrooms.

Suppose that there are two types of customers who want to use the bathroom. One has a very elastic demand for bathrooms, and will go at home if the shop doesn’t have one. The other has very inelastic demand, and will continue down the street if the shop they stop in doesn’t have a bathroom. Businesses provide a bathroom because it increases sales, and pay a cost to clean the bathroom. Now, we will stipulate that the cost of cleaning is higher for the inelastic group than the elastic group. This maps onto the homeless (in San Francisco) having a higher likelihood of messing up the place.

Now, suppose everyone has a bathroom provided. This could plausibly be an equilibrium – every business receives some flow of payments in excess of the costs. I am going to abstract away from entry, and just say that the number of businesses is fixed. Now, suppose that the costs to businesses are independent, and are a first order Markov variable. Some will have their costs exceed their benefits, and they will close the bathroom. This leads the proportion of expensive, inelastic customers at the other restaurants to increase, and some other businesses close their restrooms, and so on. The market totally unravels.

Strikingly, this holds even if businesses are able to charge for the right to use the bathroom, if we assume that costlier consumers are less responsive to a charge than the less costly consumers. This is a fairly standard sort of adverse selection problem. This assumption is one that I am less confident, however, as it seems likely to go the opposite way.

In the toy model I have given above, though, a bathroom mandate would raise welfare.

I spent an afternoon dithering around trying to put numbers on this. However, usage data would have to be collected in the field (which makes BLP type estimation impossible), and the homeless are only recorded at the district level, and in any case regressions would be hopelessly endogenous. It would be cool to pick this up again in the future, though. Something which stands out as a possibility are Starbucks’ policy changes for their bathrooms.