I have been working, like most economists, on rational models where people are fully aware of the economic environment and all the policies they face. That 'full information' assumption is deeply embedded in many areas of economics, especially public finance, my area of focus.

I began getting interested in this assumption partly because of the work of my wife, Sundari, who is doing her PhD at Berkeley in neuroscience, in particular the effects of stress on the brain. I understand from conversations with her that the brain is sufficiently complicated that we cannot expect people to be perfectly optimising economic agents in all environments. So, naturally, I began to question myself and the theoretical models I was writing: 'Does it make sense that people are fully aware of all the taxes and other government policies that they face?' Talking with non-economist friends, my sense was that many of them do not think about taxes at the level of detail the way economists that study them do. I knew that this could substantially change theories of optimal taxation, and began to think about a research project on this problem.

The first step in challenging a longstanding assumption is to come up with clear evidence that the assumption is not correct. This led to an experiment at a grocery store in California. If you go to a store in most states in the US, say, to buy razor blades, the price says $8.99 on the shelf. But you pay something like $9.72 at the register, because of 8 percent sales tax. We did a simple intervention where we added a tag below the existing tag showing the tax-inclusive price, that is, total price: $8.99 + CA sales tax = $9.72. We did this for a thousand products that in categories that are very price elastic, eg cosmetics and hair care accessories. The traditional economic model would say that this intervention should have no effect because people should already be taking the tax into account. After all, we are just giving them redundant information. We found that quantity sold did in fact fall quite substantially, by about 8 percent relative to sales in previous weeks. The implication is that most people apparently don't normally take the sales tax into account when making purchasing decisions.

The concern with the experiment is that it may be a short-run effect, one that occurs only because we violate people's norms by posting these 1,000 new tags. So we tried a different strategy. Suppose I increase the price of a product by a dollar, and I increase tax on that product by a dollar. The traditional economic theory would say that both should have the same effect because you should only care about the price plus the tax.

We test that by prediction by looking at consumption of alcohol. Alcohol is subject to two taxes in the US: One that is included in the price called the excise tax and one that is added at the cash register, the sales tax. Using legislated changes in these tax rates going back to 1970, we find that excise tax increases reduce demand about 10 times as much as the sales tax. The effect of sales taxes is much smaller than excise taxes even in the long run (two or three years).

Together, the two lines of evidence show that individual's responses to taxation depend quite significantly on the salience or visibility of the tax. If even these simple taxes pose such problems, one must be concerned that similar effects will emerge for the much more complicated income and capital gains taxes. So now we had some evidence to challenge the longstanding assumption. The next step in the project was: What does all this mean? What are the consequences of economic policies in view of this finding?

And what did you conclude?

There are a few interesting results. The first is the answer you get if you ask an economist what kind of tax produces inefficiency. Normally, economists think that when you tax products it leads to inefficiency because it reduces total economic output (people buying less) and growth. One of the most basic ideas if that if people change their behavior in response to a tax, that's a distortion and leads to an efficiency loss. Economists will tell you that if we find no link between the tax and people's purchasing behavior, that means there is no inefficiency.

I've found that this logic is wrong when people don't pay attention to taxes. Think about it like this: Suppose you don't respond to tax because you are not paying attention to the tax. And suppose, say, you're spending your money on cars and food and there are these complicated taxes on cars, which there are actually in practice in the US. So you end up overspending on cars as opposed to what you actually thought you spent. Later on, you are consuming food and you have less money left over than you thought you would. Now you could actually end up with lower overall utility or happiness than you would have, had you known about the tax from the beginning and changed your behaviour accordingly.

So this very basic prediction - that we should tax things where we don't see any response - 'inelastic goods' is not necessarily correct, if people are not paying attention to things. Essentially, you can end up reducing people's welfare through taxes even though they don't change their demand for the car, because they do not recognise the true cost of what they are doing.

Another example is that if you ask economists whether the government should tax the company or the consumer, they will tell you that it makes no difference. The price is going to adjust and the company is going to pass the tax on to the consumer to some extent anyway. But in the case where not everyone pays attention to all taxes, it makes a big difference if you tax the consumer or you tax the company. Say you tax the individual for a cell phone plan. The price for the cell phone plan will be quoted by the company as say $39.99, but you end up paying $48.00 because of the taxes that are added later. But suppose I levied the same taxes and fees directly on the company. It would be harder for them to raise the quoted price by $8 and pass the whole tax on to the consumer because the consumer is paying attention to the quoted price and will buy less. Whereas, if the government levies the tax on the consumer side, it's easy for the company to say to the consumer that the price you are paying is only $39.99; the tax ends up being paid completely by the inattentive consumer. So the so-called "tax neutrality result" - that it doesn't matter who you tax: the supply side or the demand side, the company or the consumer, does not hold. This suggests that we should think much more carefully about which side of the market we tax.

Image: Talking to his non-economist friends helped Chetty realise that many of us do not think about taxes in detail. He then deduced that this could substantially change theories of optimal taxation. Photograph: Yoshikazu Tsundo/AFP/Getty Images