Econometrics and Marketing Research

In the words of Investopedia, "Econometrics is the quantitative application of statistical and mathematical models using data to develop theories or test existing hypotheses in economics and to forecast future trends from historical data. It subjects real-world data to statistical trials and then compares and contrasts the results against the theory or theories being tested." 

Many econometricians work extensively with time-series data, which cover multiple points in time, not a single period as is the case with cross-sectional data. This specialization within econometrics is a crucial part of what has been dubbed macroeconometrics. Some time-series analysis methods used in macroeconometrics are employed in marketing research, for instance, in marketing mix modeling.  

The basic purposes of mix modeling, put simply, are to gauge how much marketing bang we’re getting for our marketing buck and to inform marketing planning. Mix models are also employed to forecast sales and market share under alternative marketing scenarios. These forecasts can be used to choose a marketing plan from several alternatives.  

Long and short term effects

Recently there has been much discussion regarding the long- and short-term effects of marketing, advertising in particular. Long- and short-term effects of government and corporate decisions are areas of interest in macroeconomics and some methodologies employed by economists have been adapted to mix modeling.  

Economic data are often utilized in mix modeling as well, inflation and GDP data for instance. One caveat is that many government statistics are published infrequently and often revised. Another is that it can be hard to decide which data to use, and whether the data should be local, national, regional, global or some geographic combination. This is a concern in any kind of research, not just mix modeling. 

Marketing researchers can use time-series analysis in other ways, for example with brand and customer satisfaction tracking. Some social media and customer correspondence data also lend themselves well to time-series analysis of various types. 

Many methods for analyzing time-series data and forecasting future periods have been developed over the years. Exponential smoothing, ARIMA, regression with ARMA errors, VAR/VECM, GARCH, and state-space models are a few of the ones used in macroeconometrics and other fields. Deep learning and other machine learning techniques are also employed in time-series analysis, which is a vast subject covering multiple disciplines, not just econometrics. 

We can also make use of methods designed for longitudinal data with a small number of periods, 6-8 for example. Consumers are occasionally interviewed more than once over a time frame, which could span just a few weeks or cover a longer period. These methods are distinct from time-series analysis, which typically requires at least 25-50 points in time. Sometimes there are thousands or even hundreds of thousands of time points in time-series analysis.

Spatial analytics and spatio-temporal models utilized in econometrics and other fields are also used in marketing research, in store placement research, for instance. 

Microeconometrics

There is also what is sometimes called microeconometrics, which is possibly closer to what most marketing researchers would associate with statistics in general. Many marketing researchers have been conducting microeconometrics throughout much of their careers without calling it such. 

Microeconometricians typically analyze data pertaining to corporations, households and individual consumers. Data may be cross-sectional or cover two or more periods in time. As with macroeconometrics, data are usually observational, though randomized experiments are also utilized. Microeconometrics makes heavy use of surveys.  

ANOVA, linear and logistic regression, principal components and factor analysis, discrete choice modeling, cluster analysis and random forests are some popular techniques. There are specialized methods, in addition, that may be unfamiliar to most marketing researchers. Arellano-Bond estimation is one example. 

I have listed many learning resources on cannongray.com/methods under Time Series Analysis and Marketing ROI and several other places. Mostly Harmless Econometrics (Angrist and Pischke), Introduction to Econometrics (Watson and Stock) or Introductory Econometrics (Wooldridge) are possibly the best places to start if you’re new to the field. Some coursework in statistics is assumed by these books, though it needn’t be extensive. 

As with any kind of research, clear objectives will go a long way! While data mining approaches can be useful in developing hypotheses they can also get us in trouble, as explained in Stuff Happens. Understanding our data and, to the extent possible, making sure they’re clean are other important fundamentals. 

How to communicate our results, like any research, depends very much on our audience and their expectations. My general preference is to concentrate on the key findings and implications to decision-makers and to avoid technical discussions as much as possible. These discursions can chew up precious time and distract our audience and readers from what is most important to them. 

Marketing research draws most of its methodologies from other disciplines. Econometrics is only one, but an important one.