Working Paper: NBER ID: w14368
Authors: Timothy Erickson; Ariel Pakes
Abstract: Until recently the Consumer Price Index consisted solely of "matched model" component indexes. The latter are constructed by BLS personnel who visit stores and compare prices of goods with the same set of characteristics over successive periods. This procedure is subject to a selection bias. Goods that were not on the shelves in the second period were discarded and hence never contributed price comparisons. The discarded goods were disproportionately goods which were being obsoleted and had falling prices. Pakes (2003) provided an analytic framework for analyzing this selection effect and showed both that it could be partially corrected using a particular hedonic technique and that the correction for his personal computer example was substantial. The BLS staff has recently increased the rate at which they incorporate techniques to correct for selection effects in their component indexes. However recent work shows very little difference between hedonic and matched model indices for non computer components of the CPI. This paper explores why. \n \nWe look carefully at the data on the component index for TVs and show that differences between the TV and computer markets imply that to obtain an effective selection correction we need to use a more general hedonic procedure than has been used to date. The computer market is special in having well defined cardinal measures of the major product characteristics. In markets where such measures are absent we may need to allow for selection on unmeasured, as well as measured, characteristics. We develop a hedonic selection correction that accounts for unmeasured characteristics, apply it to TVs, and show that it yields a much larger selection correction than the standard hedonic. In particular we find that matched model techniques underestimate the rate of price decline by over 20%.
Keywords: Consumer Price Index; Hedonic Regression; Selection Bias
JEL Codes: E31; L11; L16
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| selection bias (C24) | overestimation of inflation rates (E31) |
| exit of goods (F19) | overestimation of inflation rates (E31) |
| product characteristics (L15) | price predictions (Q47) |
| unobserved characteristics (D80) | systematic biases in price predictions (G41) |
| hedonic techniques (C90) | larger selection correction (C52) |
| hedonic techniques (C90) | price indexes with faster decline rate (E31) |
| introduction of new goods (D40) | amelioration of biases (D91) |
| newly entering goods (Y20) | faster rate of price decline (E31) |
| selection bias (C24) | inaccuracies in inflation measurement (E31) |