Page 637 -

636 Chapter 15 Estimation of Dynamic Causal Effects

by generalized least squares (GLS). Both the ADL and GLS methods, however, require
a stronger version of exogeneity than we have used so far: strict exogeneity, under
which the regression errors have a conditional mean of zero given past, present, and
future values of X.

Section 15.6 provides a more complete analysis of the relationship between
orange juice prices and the weather. In this application, the weather is beyond human
control and thus is exogenous (although, as discussed in Section 15.6, economic
theory suggests that it is not necessarily strictly exogenous). Because exogeneity is
necessary for estimating dynamic causal effects, Section 15.7 examines this assumption
in several applications taken from macroeconomics and finance.

This chapter builds on the material in Sections 14.1 through 14.4 but, with
the exception of a subsection (that can be skipped) of the empirical analysis in
Section 15.6, does not require the material in Sections 14.5 through 14.7.

15.1 An Initial Taste of the Orange Juice Data

Orlando, the historical center of Florida’s orange-growing region, is normally
sunny and warm. But now and then there is a cold snap, and if temperatures drop
below freezing for too long, the trees drop many of their oranges. If the cold snap
is severe, the trees freeze. Following a freeze, the supply of orange juice concen-
trate falls and its price rises. The timing of the price increases is rather complicated,
however. Orange juice concentrate is a “durable,” or storable, commodity; that is,
it can be stored in its frozen state, albeit at some cost (to run the freezer). Thus the
price of orange juice concentrate depends not only on current supply but also on
expectations of future supply. A freeze today means that future supplies of con-
centrate will be low, but because concentrate currently in storage can be used to
meet either current or future demand, the price of existing concentrate rises today.
But precisely how much does the price of concentrate rise when there is a freeze?
The answer to this question is of interest not just to orange juice traders but more
generally to economists interested in studying the operations of modern commod-
ity markets. To learn how the price of orange juice changes in response to weather
conditions, we must analyze data on orange juice prices and the weather.

Monthly data on the price of frozen orange juice concentrate, its monthly
percentage change, and temperatures in the orange-growing region of Florida
from January 1950 to December 2000 are plotted in Figure 15.1. The price, plot-
ted in Figure 15.1a, is a measure of the average real price of frozen orange juice
concentrate paid by wholesalers. This price was deflated by the overall producer
price index for finished goods to eliminate the effects of overall price inflation.