Cause and Effect

Identifying the cause (or causes) of political events is one of the central tasks of political science.  Talking about cause and effect seems easy – one thing leads to another -- but defining what a cause is gets harder the more you think about it, especially in politics.  In this video lecture, I will outline the technical definition of causality, focusing on three key elements:

-        Separating cause and effect

-        Eliminating alternate explanations

-        Thinking about likely, rather than absolute, causes

Identifying causes

One reason we focus on causal questions in this class is that they are often really hard to answer.  Take the question of whether economic development causes democracy.  After we define our terms, there are many ways that development might lead to democracy.  Development might lead to a larger middle class, who starts to demand political rights to match their new economic power.  Or it might empower the working class, who protest to improve their conditions. Or, we might see a relationship between development and democracy, but democracy might be the cause, not the effect, of economic development, if it creates political institutions that promote innovation and growth.

So, when thinking about causality, we have to identify different possible causes and whether we are really talking about a cause or an effect.

Defining causality

I don’t have the definitive answer to this question, but we can start tackling it by being clearer about what we mean by a cause. 

This is debated in the philosophy of science, but in social science, we tend to use a formal definition of causality developed by the psychologists William Shadish, Thomas Cook, and Donald Campbell. 

There are three things you need to call something a cause:

1)     Correlation.  Your cause needs to be related to your effect.  Using more formal language, the “independent variable” (potential cause) and “dependent variable” (potential effect) should change together.  When your potential cause changes, your effect should also change.  Correlation alone doesn’t mean that one thing causes another, but without some kind of relationship, there’s definitely not a cause.

2)     Temporal precedence.  Your cause needs to happen before the effect.  As with correlation, this is fairly obvious, but not always easy to show when we are talking about things like democracy and development, which can happen slowly.

3)     Finally, we need to eliminate alternate explanations.  This is the most important of these three criteria, and the hardest one to prove.  Let me give you a few examples to demonstrate this.

The most famous example of how correlation is not causation is the fact that ice cream sales are correlated with shark attacks.  Just because people are eating more ice cream doesn’t cause them to be eaten by sharks (even though you could create a story in which that makes sense).  There is an alternate explanation that makes more sense:  a third factor (hot weather) can explain both why people are eating more ice cream and why they are more likely to be attacked by a shark (since they are more likely to be at the beach).  Here we have two variables that are related, but we don’t see one clearly happening before the other and we can’t rule out an alternate explanation.  This is what is called a spurious correlation.  If you remember one thing about causality, it is that correlation is necessary for causation, but it is not the same thing.  Spurious correlation is when two things are related, but only because they are both caused by a third factor.  This is a confounding variable or a control variable, and this is why we must eliminate alternate explanations if we want to show one thing causes another.

To come back to democracy and development, then, if we wanted to show that a rising middle class leads to democratization, we would need to show that the cause isn’t a more educated population or labor protests.  This is hard because it can be difficult to separate variables like education from economic power – they are all correlated with democracy – or to come up with a complete list of all possible explanations for why one country became democratic.  In political science, we usually don’t prove causality, but instead rule out alternatives.  In doing so, each time we examine a question we can get closer to the truth.

I want to touch on the final two parts of this formal definition.  These aren’t requirements, but are nice to have.  Manipulation means running an experiment.  The easiest way to think of this is a clinical trial for a vaccine.  We can conclude a treatment is or is not effective by randomizing who receives it during a test period.  We will talk later about how we can do this in political science and public policy, but it’s not possible to do this for many if not most political questions, which is why it’s nice to have, but not necessary.

Last, it is also helpful to be able to describe a causal mechanism – the process by which a cause leads to an effect.  This is something we often are able to do in politics – we can describe the history of a country, for example – but it isn’t a formal requirement because it is often missing in medicine.  We knew that aspirin reduced headaches long before we understood the biology of how it worked, for example.

So those are nice to have, but the key parts of demonstrating causality are:

·   showing that the cause and effect change together,

·   that the cause comes before the effect, and

·   that nothing else can explain that relationship.

Types of causes

I want to end with a brief discussion of some of the adjectives you may see associated with the word cause in our readings.  We also talk a lot about proximate and distal causes.  A proximate cause is a short-term cause – protests lead to the resignation of a leader, for example.  A distal cause is a long-term one – for example, that changes in the structure of an economy created the conditions for democratization.  Both proximate and distal causes can be true, but you should be clear as to which one you are interested in.

Finally, in politics we almost never study deterministic relationships.  That means that when one thing (a cause) happens, an effect always happens.  There are always exceptions in politics, so we almost never see the kind of definitive relationships we study in physics or chemistry.  Instead, we study what conditions make an effect more likely (or increase the probability). 

I wanted to both familiarize you with these terms, but the key things to remember from this lecture are:

1)     Be clear about what you think is a cause vs an effect – separate them out.

2)     The key factor in determining causality is eliminating alternate explanations.

3)     In politics, we need to think probabilistically – what conditions makes an outcome more likely.  We can never guarantee something will happen.