Uncertainty Lowers the Discount Rate

PrintPrint

Uncertainty Lowers the Discount Rate

This module has covered the techniques for calculating the present value of future events. But, how uncertainty shows up in this is possibly surprising.

Suppose you estimate that damages of $1000 will occur in 100 years, and you want to estimate their present value as properly as possible. But, you aren’t sure whether the discount rate is D=1% or D=7%—you think that 1% and 7% are equally likely to be correct. You decide to split the difference, by assuming that the average of the present values of these cases is your best estimate of the present value.

You recall that earlier in this module we found that P=F/(1+d)t, allowing you to calculate the present value P from the future value F=1000 for time t=100 using the discount rate d=D/100.

You might be tempted to assume that the answer can be obtained by averaging 1% and 7% to get 4% and then calculating P for this discount rate of 4%. You would be wrong.

If you got out your calculator, you would find
For D=7%, P=1.15
For D=4%, P=19.80
For D=1%, P=369.71

You want the average of the present values for the 1% and 7% cases, which is (369.71+1.15)/2=185.43. That is almost 10 times larger than the value P=19.80 you get for a 4% discount rate.

What happened? For sufficiently high discount rate and sufficiently long time, the present value of any future event is very small, and you can’t lower a small value very much more by using a higher discount rate or longer time. Under those conditions, the average of the present values with an uncertain discount rate is not too far from being half of the present value, with the lower discount rate.

You could put the average present value, 185.43, into the equation above and solve for the discount rate that would give it. The result is D=1.7%, far below the 4% you would get by averaging 1% and 7%.

If you then decided to look at the present value of damages of 1000 occurring 200 years in the future rather than 100 years, you would find
D=7%, P=0.001 (that’s one-tenth of a cent)
D=4%, P=0.39 (that’s 39 cents)
D=1%, P=137.69

The average of 137.69 and 0.001 is 68.85, almost exactly half of the low-discount case. And if you calculate the one discount rate that would give this, it is 1.3%, even lower than the 1.7% for 100 years.

So, if you want to use a single number for the discount rate, but you know that there really is uncertainty about what that number should be, the single number is lower than the average of the possible discount rates, and the single number gets smaller as you look further into the future. In turn, our lack of knowledge about the future means that an economically efficient response involves more actions now to prevent global warming than you would calculate by simply taking a discount rate in the middle of the possible values.

For additional insights on why the discount rate should be made smaller when looking further into the future, motivating more action now to reduce global warming, see K. Arrow et al., 2013, Determining benefits and costs for future generations, Science 341, 349-350. (Note that in the Newell and Pizer paper referenced in the main text, and in this Arrow paper, time is made continuous and discounting is exponential; their equation looks different from ours, but the numerical difference from what we did here with annual values is very small—for example, Newell and Pizer get 34 cents rather than 39 cents for the present value of 4% discounting of 1000 in damages 200 years in the future.)