In an often cited paper, Hall (1978) observed that, under rational expectations and the life cycle-permanent income hypothesis, the first-order condition for utility maximization implies that the marginal utility of consumption follows a random walk. If single-period utility is quadratic, that is, if U(Ct) = -.5(B Ct)2, where Ct is time t consumption and U is utility, then consumption itself follows the random walk C, = a + bC,1 + et. If true utility is not quadratic, it may still be approximated by a quadratic. In this model b is the ratio of one plus the rate of time preference for utility to one plus the interest rate, and a = (1 b)B. The parameter B is interpreted as the bliss level of consumption because utility takes on its maximum value when C = B. This utility function is undefined for C > B because marginal utility is negative at these consumption levels. Given estimates of a and b, the corresponding bliss level is simply obtained using B = a/( 1 b). Yet authors who employ Hall's methodology and quadratic utility have not calculated B. Their oversight is corrected here. Using real per capita U.S. consumption of services and nondurables, Hall estimates a = 14 and b = 1.011 (0.003) (standard errors are in parentheses). These numbers imply a bliss level of B = $1,270, while C is in the range of $2,000-$4,000 (Hall actually gave the estimate of a as 0.0 14, but his data were in thousands). For consumption of durable goods, Mankiw (1982) uses Hall's methodology to derive the model Ct = ad + bCt1 + et + (d 1)et1, where Ct is real per capita U.S. consumption of durables and d is the rate of depreciation of durable capital. Setting d = 1, Mankiw estimates a = 2.65 (5.04) and b = 1.002 (0.011), which implies a bliss level of B = -$1,325. When d is estimated along with the other parameters, Mankiw finds ad = -4.37 (3.49), b = 1.015 (0.011), and
Read full abstract