Warren Bickel and other researchers have argued on the basis of experimental economic studies that drug addicts have higher rates of time preference than others in the community and higher rates of time preference, in particular, for drugs like heroin than for money. Look here.
In a conversation (over a bottle of excellent Chalambar Seppelts Shiraz (2003)) last Friday, my friend, Professor Peter Bardsley from Melbourne made the good point that a rate of time preference is like a price, and that, for prices to differ, markets must be missing. If markets are complete a law of one price should hold.
For a drug addict to attach a much higher price for delaying the consumption of heroin now than delaying the receipt of money now must suggest that, for he/she, the possibility of taking the money and using it to buy heroin must somehow be prohibited. Otherwise you would think that they could borrow now and use the money to buy heroin in the future making a pure arbitrage gain.
Maybe the circumstances of addicts prevent them doing this. Maybe they cannot set aside money now to buy drugs in the future because they are income-constrained. Its a possibility that sounds half-plausible and is worth investigating.
We should also look carefully at the way Bickel and his co-authors (and many others) designed their experimental studies. Did they allow for the possibility of credit? More work needed on this and the chance to get involved in sopme interesting experimental economics.