Collecting Relevant and Useful Data

• Pages: 45-64

45

CHAPTER 4

COLLECTING RELEVANT AND USEFUL DATA

A. Introduction

Data are critical inputs in econometric analysis. Even the most

sophisticated statistical analysis cannot make reliable inferences from

bad data. For certain econometric methods, specific types of data are

essential inputs without which the econometric analysis does not work.

Although data issues are frequently complex, it is extremely helpful

for lawyers working with econometric experts to understand them, along

with the advantages and disadvantages of particular types and sources of

data. Not only will that allow lawyers to provide guidance to their clients

as to the likely utility of particular data collection efforts, but it will

allow those who are frequently the primary interface between the

econometricians and the data source to communicate effectively and

prioritize the experts’ data requests. Informed lawyers should also be

able to better help experts understand what types of data are available

from the client, how they are kept, and where problems exist in the data

set that could raise questions as to the reliability of the analysis.

This chapter discusses the nature, types, and sources of data

commonly used in antitrust analysis. It aims to highlight some of the

issues most relevant to lawyers dealing with econometric evidence. The

chapter opens with a discussion of the three formats of economic data

used in econometric analysis. This is followed by a description of the

different types of data available for econometric analysis and the sources

of data available to the economist. It then provides a discussion of data

considerations that may affect econometric analysis.

B. Nature of Economic Data

Three kinds of data are often used in econometric analysis: cross-

sectional data, time series data, and data containing both cross-sectional

and time series data.

46 Econometrics

1.

Cross-Sectional Data

Cross-sectional data are observations collected for a number of

observational units at the same point in time, for example, the prices for

a particular product sold in twenty major metropolitan statistical areas

(“MSAs”) on January 1, 2010.

Cross-sectional data allow some empirical investigations into

questions about a particular point in time. For example, was a product

priced the same across MSAs? If not, what are the factors that

contributed to this difference? Although a properly constructed

econometric model may be able to explain much of the price variation

across regions, it is often difficult to explain price variations that are due

to immeasurable effects that are specific to a particular MSA. In such

circumstances, an econometric study based on cross-sectional data alone

would provide limited or even misleading results.

2.

Time Series Data

Time series data are observations collected for a single observational

unit at different points in time, for example, the daily price of a particular

product sold in the Washington DC metropolitan area in 2010. Time

series data observations are most often observed at the same intervals,

such as yearly, quarterly, monthly, or daily. The proper length of the

intervals depends on the situation being studied and data availability. For

example, household income data may be collected annually to smooth

out the uneven distribution of different sources of income throughout the

year, while price data in a stock market such as New York Stock

Exchange or in a commodity trading exchange such as Intercontinental

Exchange (ICE) can be collected virtually continuously in a real-time

fashion. Time series data allow an analysis of how a particular variable’s

value changes over time, and an analysis of what potential factors

contribute to such changes.

Time series data are particularly useful when the analysis of interest

is to determine the effect of a particular change in the market over time.

Such analysis is commonly used in price-fixing litigation or a

retrospective analysis of a consummated merger. However, time series

data have their limitations because they can mask the effect of

differences across the observational units that are aggregated across time.

For example, the observational unit in a time series data set can be

the average sales price of a particular product for a particular period of

time. Averaging price, even a weighted average price (i.e., an average