3 Economics: Methods, approaches, fields and relevant questions

3.1 Economic theory and empirical work: What is it?

What is economic theory and what can it do?

Unlike “theory” in some other social science disciplines, economic theory is mostly based on mathematical modelling and rigorous proof that certain conclusions or results can be derived from certain assumptions. But theory alone can say little about the real world.

In Economics: Models = Theory = Mathematics… for the most part.

What is empirical work and what can it do?

In contrast, empirical work gathers evidence from the real world, usually organized into systematic data sets.

  • It tries to bring evidence to refute or substantiate economics theory,

  • It tries to estimate parameters such as price elasticity or the government spending multiplier in specific contexts

  • It rigorously presents broad “stylized facts”, providing a clear picture of a market, industry, or situation

Much empirical work itself relies on assumptions, either assumptions from economic theory, or assumptions about the data itself, or both. But empirical work does not “prove” anything. Instead, it presents evidence in favour of or against certain hypotheses, estimates parameters, and can, using the classical statistical framework, reject (or fail to reject) certain null hypotheses. What “rejecting” means is “if the assumptions underlying my estimation technique are correct, then it is highly unlikely that the null hypothesis holds.”

3.2 Normative vs. Positive

The word ‘Normative’, also called ‘prescriptive’, often refers to what ought to be, what an ideal policy would be, or how to think about judging whether this is a justifiable welfare function.

“Positive” work claims to be value-neutral and to address what is or what must be going on in the real world. Most modern economists would probably claim their work is “positive”, and in this sense, “prescriptive” is often used as a pejorative, In my experience. However, prescriptive papers can be very valuable if done well.

Note: There is also another context in which you will hear the expression ‘normative analysis.’ This may also be used to describe microeconomic analysis derived from the axioms of rational optimising behavior; this describes much of what you have covered in your textbook. This dual meaning of the word ‘normative’ is admittedly confusing!

3.3 Theoretical vs. Empirical (techniques)

Papers that use theory (modeling) as a technique typically start from a series of assumptions and try to derive results simply from these assumptions. They may motivate their focus or assumptions using previous empirical work and anecdotes, but these papers do not use themselves data nor do they do what we call “econometrics”. Remember that in Economics “theory papers” are usually highly mathematical and formal.

Empirical papers use evidence from the real world, usually to test hypotheses, but also to generate description and help formulate ideas and hypotheses.

3.4 Theoretical vs. Applied (focus)

“Theoretical” can also be used to describe a paper’s focus; a theoretical paper in this sense will address fundamentals of economic modeling. In theory, these may be widely applied across a range of fields, but they do not typically address any single issue of policy or focus on a specific industry. These papers are often very difficult to read and there is argument about whether many such papers will ultimately “trickle-down” to having practical use. These papers typically used theory and modeling techniques rather than empirics. However some empirical papers may be aimed at addressing fundamental theoretical issues and parameters.

Papers with an “applied” focus will directly target a policy issue or a puzzle or question about the functioning of certain market or nonmarket interactions. Nearly all of the papers you will read and work on as an undergraduate are “applied” in this sense.

3.5 Categories of empirical approaches

“Causal” vs. “Descriptive”

“Causal” papers try to get at whether one factor or event “A” can be seen to directly “cause” an outcome “B”. For example, “does an individual getting more years of schooling lead him or her to have higher income, on average?” A good way to think about this conception of causality is to consider the counterfactual: if a typical person who received a university degree had been randomly selected to not get this education, would her income have been lower than it now is? Similarly (but not identically) if the typical person without his education had been randomly input into a university program, would her income now be greater?

Since the real world does not usually present such clean experiments, “causal” empirical researchers rely on various techniques which usually themselves depend on" identification assumptions." See, for example, control strategies, difference in difference, and instrumental variables techniques.

“Descriptive” papers essentially aim to present a picture of “what the data looks like” in an informative way. Causal relationships may be suggested but the authors are not making a strong claim that they can identify these. They may present a data-driven portrait of an industry, of wealth and inequality in a country or globally over time, of particular patterns and trends in consumption, of a panel of governments’ monetary and fiscal policy, etc. They may focus on the ‘functional form’ of relationships in the data and the ‘residual’ or ‘error structure. They may hint at causal relationships or propose a governing model. They may identify a ’puzzle’ in the data (e.g., the ‘equity premium puzzle’) and propose potential explanations. They may use the data to ‘provide support’ for these explanations.5 They may devote much of the paper to providing a theoretical explanation (remember, in Economics these are usually mathematical models) for the pattern. They may also run statistical tests and report confidence intervals; one can establish a ‘statistically significant’ relationship between two variables even if the relationship is not (necessarily) causal. This is particular important when one sees the data as subject to measurement error and/or as a sample from a larger population. E.g., just because age and wealth (or height and head-size, or political affiliation and food-preference) are strongly related to one another in a random representative sample of 10 people does not mean they are strongly related to one another in the entire population.6

Structural vs. Reduced Form

This is a rather complicated issue, and there are long debates over the merits of each approach.

In brief, structural empirical papers might be said to use theory to derive necessary relationships between variables and appropriate functional forms, often as part of a system of questions describing a broad model. They then “take this model to the data”, and estimate certain parameters; these estimates rely on the key structural assumptions and the chosen functional form (which is often selected for convenience) holding in the real world. They may also try to check how “robust” the estimates are to alternate assumptions and forms. Structural estimates can then be used to make precise predictions and welfare calculations.

Reduced form work may begin with some theoretical modeling but it will not usually try to estimate the model directly. Reduced form work often involves estimating single equations which may be “partial equilibrium”, and they may often use linear regression and interpret it as a “best linear approximation” to the true unknown functional form. Reduced form researchers often claim that results are “more robust” than structural work, while proponents of structural work may claim that reduced form econometrics is not theoretically grounded and thus meaningless.

Most of you are likely to focus on reduced form empirical work.

Quantitative vs. qualitative (the latter is rare in economics)

Quantitative research deals with data that can be quantified, i.e., expressed in terms of numbers and strict categories, often with hierarchical relationships.

Qualitative research is rarely done in modern economics. It relies on “softer” forms of data like interviews that cannot be reduced to a number or parameter, and cannot be dealt with using statistical techniques.

3.6 Methodological research

Methodological research is aimed at producing and evaluating techniques and approaches that can be used by other researchers. Most methodological research in economics is done by econometricians, who develop and evaluate techniques for estimating relationships using data.

3.7 Fields of economics, and some classic questions asked in each field

Economics is about choices under conditions of scarcity, the interaction of individuals governments and firms, and the consequences of these. [citation needed]

The standard official categorization of Economic fields and sub-fields is the ‘JEL code’ (from the Journal of Economic Literature). AEA guide, Wikipedia: JEL classification codes


Preferences and choices under constraints; e.g., “how do risk-averse individuals choose among a set of uncertain gambles?” … “How does consumption of leisure change in response to an increase in the VAT?”

Game theory, interactions; … “How do individuals coordinate in ‘stag hunt’ games, and are these equilibria robust to small errors?”

Mechanism design and contract theory; … “How can a performance scheme be designed to induce the optimal level of effort with asymmetric information about ability?”

Equilibrium; … “Is the general equilibrium of an economy with indivisible goods Pareto optimal?”


Stabilisation; … “how do changes in the level of government spending affect the changes in the rate of unemployment?”

Growth; …“Why did GDP per capita increase in Western Europe between 1950 and 1980?”

Aggregates, stocks, and flows; … “Does a trade deficit lead to the government budget deficit, or vice/versa, (or both, or neither)?”

Money and Banking … “Does deposit insurance decrease the likelihood of a bank run?”

Financial Economics (not as broad as the first two)

“Can an investor use publicly-available information be used to systematically earn supernormal profits?” (the Efficient Markets Hypothesis)

Econometrics (methods/technique)

“What is the lowest mean squared error unbiased estimator of a gravity equation?”

Experimental economics (a technique)

Do laboratory subjects (usually students) coordinate on the efficient equilibrium in `stag hunt’ games? Do stronger incentives increase their likelihood this type of play?

Behavioural economics (an alternate approach to micro)

“Can individual choices over time be rationalised by standard exponential discounting, or do they follow another model, such as time inconsistent preferences and hyperbolic discounting?”

Applied fields


“Has the legacy of British institutions increased or decrease the level of GDP in former colonies?”


“Do greater unemployment benefits increase the length of an unemployment spell, and if so, to what extent?”


“Does public support for education increase or decrease income inequality?”


“Why did the industrial revolution first occur in Britain rather than in another country?”


“Are protectionist infant industry policies’ usually successful in fostering growth?”


“Do floating (rather than fixed) exchange rates lead to macroeconomic instability?”


“What is the appropriate discount rate to use for considering costly measures to reduce carbon emissions?”

Industrial Organization

“Do firms innovate more or less when they have greater market power in an industry?”


“Do ‘single payer’ health care plans like the NHS provide basic health care services more or less efficiently then policies of mandated insurance and regulated exchanges like in the Netherlands?”

A more extensive definition and discussion of fields is in Appendix A of “Writing Economics”

Do you know?…

Which type of analysis typically uses the most ‘difficult, formal’ maths?7

  1. Microeconomic theory
  2. Applied econometric analysis
  3. Descriptive macroeconomics

  1. Another use of data: ‘calibrating’ models aka ‘calibration exercises’; I will not discuss this at the moment.

  2. Sometimes this can be confusing, particularly when the data seems to represent the entire ‘population’ of interest, such as an industry’s price and sales data in a relevant period. Without getting into an extensive discussion of the meaning of probability and statistics, I will suggest that we can see this as a ‘sample of the prices and sales that could have ocurred in any possible universe, or over a period of many years’. Ouch, this gets thorny, and there are strong debates in the Statistics world about this stuff.

  3. Answer: 1. Microeconomic theory