Welcome#
to the knowledge economy#
In an introductory economics class, we learn about markets, different market structures, and why firms enter and exit those markets. Important features of markets include any “barriers to entry” that new firms face.
Think of the arts and sciences as an economy, where each discipline, or field, is a market. And each researcher within a field is a firm that produces a product, e.g., art or knowledge, that competes with the output of other firms. New researchers face barriers to entering a field just as a new firm might when they enter a market. Most of the barriers are conventional things like startup costs (e.g., tuition and equipment), regulatory hurdles (e.g., admissions and degree certification), and strategic barriers (e.g., the peer-review process for publishing books and papers).
But the main barrier to entry is just the language used in each field. Even within economics, the language used in various fields appears to change with their specialized concepts, data, tools, and methods. Researchers must be familiar with, read, and speak the language of a field before entering that field to produce useful and novel contributions and ultimately becoming a competitive force.
and the language of time series econometrics#
What is time? You probably take the definition of time for granted, but time “is the continued sequence of existence and events that occurs in an apparently irreversible succession from the past, through the present, and into the future.”[1] It is the particular sequence of events that will set the data and methods in this class apart from others.
A time series is data that is a function of time. In jargon, a time series is “a series of data points indexed (or listed or graphed) in time order. Most commonly, a time series is a sequence taken at successive equally spaced points in time. Thus [a time series] is a sequence of discrete-time data.”[2]
However, defining and collecting data is only the beginning; we need the tools to study it. Time Series is a system of concepts, tools, methods, and processes to turn time-series data into something understandable and meaningful. The great thing about time is that we can partition the past from the future. We can use time series analysis to study and try to understand the past and time series forecasting to think and make predictions about the future. Although, that is not limited to economics; all disciplines might use time-series data to analyze the past and forecast the future.[2]
“Econometrics is an application of statistical methods to economic data in order to give empirical content to economic relationships.”[3] Combining the above definitions, Time Series Econometrics is an application of statistical methods to economic time-series data in order to give empirical content to economic relationships over time.
And the big question is: How do economic policy changes affect economic outcomes (e.g., quantities, prices, employment, income, consumption, well-being)? Do they rise or fall and by how much? Well the keyword there is “affect,” or causes, causation, a causal relationship. It would be good to know whether an economic policy change actually resulted in, or caused, a statistically significant change in outcomes. You can ask that question in both the cross-section and time dimensions. Thus, time series econometrics may be used to analyze the effects of past events on present economic outcomes, or forecast the effects of past and present events on future economic outcomes.
Cross section vs. time series#
Cross-sectional data “is a type of data collected by observing many subjects (such as individuals, firms, countries, or regions) at a single point or period of time.”[4] Data with both cross-section and time dimensions is known as longitudinal data, or panel data (if the subjects do not change across time).[5]
In practice, data may have a cross-section dimension, \(N\), and a time dimension, \(T\). Until now, your econometrics classes mostly used and studied data without a time dimension, i.e., \(T = 1\). We will study models for data with a time dimension, \(T > 1\).
Time series data and models: \(N\) is “small” relative to \(T\). A key feature of time series models is that they capture serial dependence, i.e., dependence across time, in the data.
Panel data and models: \(T\) is “small” relative to \(N\). These are traditionally a subfield of microeconometrics. A key feature of panel data models is that they try to capture cross-sectional heterogeneity, and, in some instances, also spatial correlation.