How Large is Your Reference Group

We discuss how specifications of interdependent preferences found in the literature yield biased estimates of parameters of the underlying consumption or choice models. We present new specifications which alleviate this problem and permit an estimation of the size of the reference group. This last point, a key element affecting the estimation biases, has been overlooked in most studies. Using French individual data on the reported subjective poverty level, we show that the reference group is likely to be very small.
[ - ]
[ + ]