Types of Economic Data
Applied work uses four data structures. A cross section samples many units at one point in time. A time series follows one unit across many periods. A pooled cross section stacks separate cross sections from different years. A panel, or longitudinal, dataset follows the same units over time. The structure dictates the assumptions and the method, so naming it correctly is the first step of any analysis.
Why it matters
Picture a spreadsheet. One snapshot of 1,000 different workers this year is a cross section. One country watched every quarter for 40 years is a time series. Two separate surveys, one from 2005 and one from 2015 with different people, pooled together is a pooled cross section. The same 200 firms tracked every year is a panel. The shape of that spreadsheet decides what you can estimate and how.
Formulas
Worked examples
You want to study how a 2014 minimum-wage increase affected employment. What data structure helps most?
A panel of the same restaurants observed before and after the change lets you difference out fixed, unit-specific traits. A repeated survey of different restaurants each year would instead be a pooled cross section, which still allows before-versus-after comparisons but cannot follow the same establishment.
Classify each: (a) GDP for the US, quarterly 1960 to 2020; (b) test scores for 4,000 students in one year; (c) wages of the same 500 workers over five years.
(a) is a time series (one unit, many periods). (b) is a cross section (many units, one period). (c) is a panel (same units across periods). In Stata you would set the panel with `xtset id year` before using panel commands.
Common mistakes
- ✗Pooled cross sections and panels are the same thing. A pooled cross section draws new units each period, while a panel follows the same units, which is what allows fixed-unit effects to be removed.
- ✗Time-series data can be analyzed exactly like cross-section data. Time-series observations are usually correlated over time and may trend, so the classical independence assumptions need to be replaced.
- ✗A cross section always has more observations than a time series. Sample size is unrelated to the data type; a long daily time series can dwarf a small survey.
Revision bullets
- •Cross section: many units, one time period
- •Time series: one unit, many periods
- •Pooled cross section: separate cross sections stacked across years
- •Panel/longitudinal: the same units followed over time
- •Data structure dictates assumptions and method
Quick check
Tracking the same 300 firms every year from 2010 to 2020 gives which data structure?
Which best describes a pooled cross section?
Connected topics
Sources
- Wooldridge (2019), Ch. 1Wooldridge, J. M. Introductory Econometrics: A Modern Approach. 7th ed. Cengage, 2019. ISBN 978-1-337-55886-0.Introduces cross-sectional, time-series, pooled cross-sectional, and panel data structures.