You are on page 1of 25

Political Science 110

Structural Causes vs. Triggering Events


The Black Lives Matter Movement
- Movement began as a response to the unpunished shooting of Michael
Brown in Ferguson, Missouri.
- Civil Unrest, protests, and riots from April to November 2014.
- Why did the Black Lives Matter movement emerge?
 Some possible causes…
o The shooting of Michael Brown?
o The acquittal of Darren Wilson?
 But what are we really talking about here?
o Growing inequality
o Institutionalized racism
o More…

Structural causes:
 Systemic Inequality
 Institutionalized racism
Triggering events:
 Shooting of Michael Brown
 Acquittal of Darren Wilson
Effects:
 Civil unrest
 Protests
 Rioting
Are triggers “causes” when we apply the counterfactual definition of
causality?
Not really, and here’s why:
1. Triggers are substitutable.
a. Any spark will start a blaze if there’s fuel.
b. When structural cause is present, tons of things could trigger the
outcome.
2. Some trigger may be necessary; but the specific trigger is usually
unimportant.

How to recognize a “trigger”:


A triggering event is an event that  generates an outcome ONLY in
combination with other (structural) causes.
- In other words, it is not sufficient.
And, it could have been substituted by other very likely events.
- Lots of sparks available, and any would have done.
- In other words, it is not necessary.
Some triggers are neither necessary nor sufficient.

The Arab Spring


Cause: Bouazizi self-immolation
Effect: Collapse of the Tunisian regime
Or… you can look at the broader structural causes and say:
Cause: Oppressive rule.

Structural causes:
- Oppressive Rule
- Inequality
- Social Media
Triggering event:
- Self-immolation of Bouazizi

2
Effect:
- Tunisian uprising

Triggers vs. Structural Causes


Social scientists are usually interested in causal patterns/regularities, which
means…
- Causes that generally contribute to a given effect.
o Revolutions do not generally require vegetable peddlers to set
themselves on fire.
o Vegetable peddlers setting themselves on fire don’t generally
make revolutions more likely.
Structural causes = portable knowledge.

Individual Choice/Qualities vs. Structural Causes


- Social interaction is a sum of individual choices, but:
- Individual choices are often produced by structural factors.
o Constraints, incentives, resources, grievances.
- Individuals often selected for by structural factors.
Why are there so few women in politics?
Is it because individual women simply do not enter politics at the same rate
as men?
… well, yes, but is this really the cause?

Trump vs. Clinton


Structural Causes
- Disaffection with government and “the system”
- Conditions for Anti-Establishment
So, does individual choice matter in Politics?
Of course, But… Individual choice and qualities are often a symptom of
broader political forces.

3
- E.g. Donald Trump’s election holds roots in the structural causes of
discontent with the government and the encouragement of an anti-
establishment government.

Midterm:
Topics 1 to 5 on the syllabus.
Midterm content:
- Bases of claims
o Non-scientific
o Scientific
- Features of science
- Scientific method
- Types of questions and claims
o Normative/Prescriptive
o Descriptive
o Causal
- Types of causal claims
o Probabilistic
o Deterministic
o Multiple Causation
o Conjunctural Causation
o General
o Specific
- Triggers vs. Structural claims

Structural Causes vs. Individual Choice


Want to understand why things occur/why patterns exist?
What structural factors condition these individual choices?

Causal Logic

4
Life Expectancy
- Life expectancy correlates with geographical location and wealth of the
country  But why else?
- System of government being a possible answer to this.
o Democracy and Health: Democracy appears to be good for your
health, has a causal effect on how long you live.
o Democracy is associated with high life expectancy, low infant
mortality, less maternal death.

Democracy and Health


Democracy is therefore correlated with good health, but is the relationship a
causal one?
Is democracy a cause of a healthier population?
Two ways to look at this:
- Control for other explanations
- Develop (and later test) a causal logic
Democracy therefore seems to have a causal influence on public health when
we control for other factors, including income.
But… this still isn’t good enough, we also want to understand: why and how.
- Why does democracy lead to a population?
- And how? How does this look like?

Causal Claims
Causal claims assert that some factor C, has an effect on factor E.
Democracy leads to better health.
But this is not enough to explain how or why E happens as a result of C 
Why or How does democracy lead to better health?
- More democracy (C)  Voting more easily bringing about removal of
government (S1)  government officials needing to please large
numbers of citizens to stay in office (S2)  Governments spending

5
greater resources on improving citizen health (S3)  Better health of
citizens.
-  means that this cause “leads to” this effect.
- One component is that each step logically follows the one before it, this
is what makes up causal logic.

Causal Logic Definition


A causal logic or mechanism is a set of statements about how or why a cause
produces an effect.
- More common term in advanced political science research is causal
mechanism.
A causal logic usually involves a causal chain that logically connects the
cause C to the effect E: C  S1  S2  S3  …  E.
A causal logic always includes a set of assumptions:
- What do we need to believe about the world to believe the steps in the
causal logic are true?
- What do we need to believe in order to believe that S1 does infact lead
to S2?

Assumptions in Causal Logic


Voting more easily bringing about removal of government (S1) 
Government officials needing to please larger numbers of citizens to stay in
office (S2).
The assumption is that: politicians want to remain in office.
- If we thought politicians didn’t care about remaining in office, this step
would not work.
Government officials needing to please large numbers of citizens to stay in
office (S2)  Governments spending greater resources on improving citizen
health (S3).
The assumption: citizens vote based on health-related issues.

6
- If citizens only vote based on the economy, or who has the best
campaign ads, this step would not work.

Producing a Causal Logic


1. Make sure causal logic starts with a cause, and ends with the effect.
2. Make sure each step is something that logically causes the step after it.
a. S1 “leads to” S2.
3. A causal logic is not a chronological list of events that happened in a
specific place, it’s a general logic of how one thing causes another in
the world.
Not introducing different variables here to produce the outcome, just stating
why one thing causes the other.

Multiple Causal Logics


The number of causal logics between any two variables can be infinite.
There are numerous potential causal logics for every cause and effect.
- C  S1  S2  S3 …  E
- C  T1  T2  T3 …  E
- C  U1  U2  U3 …  E
As social scientists, we often are interested in which causal mechanism has
the most “influence” or “weight”.

Causal Claim:
Fewer parties in an election leads to a lower voter turnout.
Two possible causal logics:
1. Mobilization
2. Policy Representation

Mobilization
Fewer parties (C) 

7
Fewer party organizations investing in citizen mobilization (S1) 
Fewer citizens contacted by party organization (S2) 
Fewer citizens feel pressured to vote (S3) 
Lower voter turnout (E) 

Policy Representation
Fewer parties (C) 
A narrower range of policy positions on the ballot (S1) 
More citizens not identifying with party policy platforms (S2) 
Fewer citizens feeling motivated to vote (S3) 
Lower voter turnout (E) 

If we make a normative commitment to increasing voter turnout, then these


causal mechanisms have different policy implications.
- Mobilization
- Provide funding for more parties?
- Find more efficient ways of contacting citizens?
- Policy representation
- Reform the electoral system to encourage more parties?
How much “weight” we assign to each causal mechanism is important.

Why Causal Logics Matter


If we know the cause of an effect, why do we care about the logic?
Wealth and Civil War
Probabilistic causal claim – Greater national wealth (C) reduces the
likelihood of civil war (E).
Cause: Greater National Wealth (C).

8
1. Richer states can spend more on counter-insurgency operations (S1) 
quick suppression of rebellion (S2)
2. Wealth reduces discontent (T1)  less rebellion (T2)
3. Young men having better economic alternatives to fighting (U1) 
rebel groups can’t recruit (U2)
Effect: Less Civil War (E)

Three causal mechanisms


C - wealth
Mechanism 1 – counter-insurgency capacity
Mechanism 2 – reduced discontent
Mechanism 3 – recruitment problems
E – less civil war

So why does knowing the causal mechanism matter for better understanding
the relationship between wealth and civil war?
Policy makes cant really manipulate a country’s wealth,
BUT…
… They can implement policies that address steps in a causal logic.

Unintended Causal Logics


Causal claim – A lower individual welfare entitlement in a town leads to a
reduced percentage of people on welfare.
Causal Mechanism 1: Employment
- Lower individual welfare entitlement (C)  People in town have an
incentive to go look for work (S1)  Fewer unemployed people (S2)
 Reduced percentage of people on welfare in the town (E)
Desirable consequence – people in town having an incentive to find
employment.

9
Causal Mechanism 2: Substitution
- Lower individual welfare entitlement (C)  People move to
neighboring town with higher welfare payments (S1)  Reduced
percentage of people on welfare in the town (E).
Unintended consequence – people move to a neighboring town with higher
welfare payments: this makes this causal mechanism be a policy failure as it
doesn’t actually help the cause, just takes people away from it.

Why Causal Logics Matter


Improving Understanding
- Helps to better understand specific cases
- Portable knowledge: generalizable to other areas as well
Testing Causal Claims
- All convincing causal claims have clear causal mechanisms
- Testing individual steps is a more rigorous

Why do political leaders create large marine protected areas:


Possible causes:
Biodiversity richness
Minimal commercial fishing in the region
Personal support for reserve from a spouse
Legal authority
Environmental groups’ lobbying efforts

Cause: Environmental groups’ lobbying efforts


S1: Increased lobbying efforts affecting the attitude of the general public
S2: Growing public support for large marine protected areas
S3: Politicians and political leaders wanting to keep the voters content

10
Effect: Political leaders creating large marine protected areas

Environmental groups’ lobbying efforts (C)  Increased lobbying efforts


affecting the attitude of the general public (S1)  Growing public support for
large marine protected areas (S2)  Politicians and political leaders wanting
to keep the voters content (S3)  Political leaders creating large marine
protected areas (E).

Assignment 3 – Submit on turnitin!!!!

February 28, 2017


Fuzzy Language vs. Clear Concepts
Claim: “political success depends on how well the economy is doing”. This is
a poor causal statement.
Concepts in this claim need to be defined  conceptualize cause and effect
here.
- One possibility: Political success is the ability of politicians to stay in
office. The economy doing well is an improvement in the standard of
living across society.
- Other possibilities: Political success could be achieving goals, staying
in office, inclusivity, job creation…etc. The economy doing well could
be GDP (economic growth).

Concepts and Measures


“Improvements in the average standard of living make it more likely that
politicians will be able to stay in office”
Concept – Politicians’ ability to stay in office.
Measures – Time in office, Number of elections won.
Concept 2 – Improvements in average standard of living.
Measures – GDP per capita, unemployment, income equality.

11
BUT we usually want to be more specific than this when we want to define
measures.

Definitions
Concept – an abstraction used to describe characteristics of a group,
phenomenon, or individual based on a set of criteria or qualities.
Measure – a procedure for determining whether or to what degree a concept
applies to specific cases based on observation of those cases.

Politicians’ ability to stay in office


- Count the total number of times a politician was successfully reelected
after already holding office.
Public approval of Justin Trudeau
- Call 1000 randomly sampled Canadians and ask “What do you think of
PM Justin Trudeau?”

“Yardsticks”
Concept: Improvement in average living standards.
Measure: National income.
But, we need a yardstick.
a. Growth in national income.
b. Growth in income per capita.
c. Growth in per capita income compared to prices.

Types of Measures
Absolute vs. Relative
Absolute:
- Measures that use measurement units.
- Do not include comparisons of cases with one another.

12
- Example being… Dollar amounts, number of occurrences.
Relative:
- Measures that consist of a comparison of cases with one another.

Categorical vs. Continuous


Categorical:
- Measures that place cases into categories based on whether
characteristics are present or absent.
Continuous:
- Measures that place cases along a spectrum from “more” to “less”
based on the degree to which characteristics are present.

Case Study – Post-Racial Politics


“A New, ‘Rost-Racial’ Political Era in America” – NPR.
“Post-Race: Is Obama the End of Black Politics?” – The New York Times.
“Obama effect’ on race politics: Hope, little change” – USA today.

Types of Categorical Measures


Binary:
- A measure of an attribute or behavior which allows only two possible
categorizations. Also known as a dichotomous measure.
- Examples  Whether a person voted.
Multichotomous:
- Placing attributes, properties, or behaviors into a pre-defined list of
more than two categories.
- There are two types of mulitchotomous measures…
o Nominal: a mulitchotomous measure in which the categories do
not have an intrinsic or natural order.

13
 Examples – Language: English, French, Mandarin, Punjabi.
Location: Downtown, Kitsilano, Point Grey.
o Ordinal: a mulitchotomous measure in which the categories do
have a intrinsic or natural order.
 Example – Social Class: Lower class, Middle class, Upper
class. Political ideology: Very liberal, Liberal, centre…etc.

Continuous:
Continuous vs. Ordinal measures.
- With a continuous measure there is an equal distance between
consecutive levels or categories.
o Example: with age, the distance in time between 1 year and 2
years is the same as the distance in time between 4 years and 5
years.
- This is different from ordinal.

March 14th, 2017


Sampling Error
- The difference between the measure for a sample and the true value for
the population.
- Very similar to how measurement error works.
- Two kinds:
o Sampling Bias - caused by a feature of the sampling procedure
that makes some members of the population more likely to be
sampled than the others.
o Random Sampling Error – caused by random variations between
samples.
Sources of Sampling Bias
1. Sampling frame ≠ Population
a. Bias that arises when the group from which you draw a random
sample differs from the population in a way relevant to what you
are studying.

14
b. Example – asking only people wearing short sleeves what their
favorite temperature is.
2. Self-selection
a. Bias that arises when respondents decide if they want to join your
sample or not.
b. Example – voluntary survey on effectiveness of government
environmental policy  will likely over represent
environmentalists’ views.

Random Sampling Error


- Sampling error that is caused by random variation between samples.
By pure chance, a random sample of a population will somewhat differ
from…
1. Another random sample from the same population.
2. The true value from the population.

Causal Theory
A set of general claims about the cause of the effect of a class of
phenomenon.
A causal theory must include…
- An independent variable (IV)
- A dependent variable (DV)
- A causal statement about what effect the IV has on the DV
- A causal logic that explains how the IV affects the DV

Variables
A variable is a measurable property of a phenomenon that can potentially
take on different values.
The values may be numerical or categorical.

15
The variation may be across units or over time.
Categorical…
- Ethnicity
- Currently at War
- Regime Type
Numerical…
- Income
- Level of Public Support
- Casualties

Types of Variables
Independent variable (x)  A variable that (we think) produces a change in a
dependent variable.
Dependent variable (y)  A variable whose value (we think) depends on the
value of some independent variable(s).
Examples of these in causal claims:
- Stronger anti-terror legislation reduces the likelihood of a terrorist
attack.
- Cutting corporate tax rates leads to lower youth unemployment.

March 21, 2017


Case Study: Corruption
Correlation – what we found in both tests in our case study is that there is a
correlation between X and Y. Correlations are a clue to causation.
- Definition of correlation: a relationship across cases between the values
that two variables (X and Y) take on.

Two Types of Correlation


- Positive correlation: cases with higher values of X have also higher
values of Y.

16
- Negative correlation: cases with higher values of X have lower values
of Y.
Why do we use correlations…
1. The fundamental problem of causal inference.
2. We cannot directly observe causation.

Testing Using Correlations


Hypothesis: The higher the level of income (X), the higher the carbon
footprint (Y).
Step 1. Taking measurements of X and Y.
Step 2. Plot them on a scatter plot.
Step 3. Draw a “line of best fit”.

Correlations can vary in three important ways:


1. Direction – positive or negative.
2. Strength – stronger or weaker.
3. Significance – more or less significant.

Direction: Positive Correlation


Causal statement – Greater equality among citizens leads to an increase in
voter turnout.
Causal logic – more citizens feel empowered and involved in the political
process and so they turn out to vote.
Hypothesis – the higher the level of equality in society, the higher the level of
turnout.

Direction: Negative Correlation


Causal statement – Negative campaigning leads to a decrease in voter
turnout.

17
Causal logic – Voters dislike politicians generally and feel like it doesn’t
matter who they vote for.
Hypothesis – the higher the level of negative campaigning, the larger the
decrease in voter turnout.

Strength: Positive correlation


A change in X leads to a large change in Y  steep line.
Weak correlation indicates that a change in X leads to a small change in Y 
flatter line.

Strength: Negative correlation


A change in X leads to a large negative change in Y (steeper line).
A weak correlation indicates that a change in X leads to a small negative
change in Y.

Significance
How closely the data points are fit to the line of best fit.
- Low statistical significance indicates that the dots are fairly spaced out
and far from the line of best fit.
- Higher statistical significance indicates that the line fits the plot points
very well.
- Correlations can have differing significances yet have the same
direction and strength.
- Statistical significance is a measure of the likelihood that a correlation
is due to random chance.

Correlation Test
Causal statement – high pre-election income growth leads to an incumbent
receiving a higher share of the vote in an election.

18
Hypothesis – the higher pre-election income growth, the higher the
incumbent vote share.

The Scientific Method


1. Formulate a question.
2. Propose a claim.
3. Test the claim.
a. Transparent procedures.
b. Systematic use of evidence.
c. Test against alternatives.
4. State conclusions.
a. Acknowledge uncertainty.

March 23, 2017


Spurious Correlation
When two variables are correlated, but that correlation is not the result of a
causal relationship between those two variables.
One common source of spurious correlation is when a third variable (Z)
causes both X and Y to vary:
 Z causes both X and Y, thus there is a correlation between X and Y.

Causal statement – high levels of democracy leads to an increase in average


life expectancy.
 GDP (Z) causes more of both the ‘level of democracy’ (X) and ‘life
expectancy’ (Y).
o There is a positive correlation between X and Y, but does not
indicate if it is causal or not.

Case Study: UN Peacekeeping Operations


Causal claim – The more UN peacekeepers in a country leads to a shorter
civil war.

19
Hypothesis – The higher the number of peacekeepers, the shorter the civil
war.
 Intensity of conflict (Z) causes more of both # Of UN Peacekeepers
(Y), Length of Conflict (X).
o Negative correlation because there are more outside factors to
consider.
So what is the relationship between a hypothesis and a causal claim?
A correct hypothesis is a clue that the causal claim may be accurate.
But a single hypothesis is rarely sufficient to convince us that a causal claim
is accurate.

Case Study: US postal charges and temperature rising.


Spurious Correlation and Time
One common source of spurious correlation: Passage of time (Z).

Spurious correlation is very common in the social sciences.


So many of the cuases that social sciences are interested in cluster together:
 Humanitarian crisis and foreign aid.
 Low income and low education.
 Democracy and wealth.
 Ethnic tension and poverty.
X variables will often be correlated with Z variables that may also be the
cause.

Solution to Spurious Correlation


To include Z in the analysis  not just rely on X, Y correlations to answer
questions.
Ways of doing this are…
 The comparative method.

20
 Multivariative Correlations (Large-N).
The comparative method  test whether causal relationship (X and Y) is still
accurate adding in the Z variable and keeping it constant.
The multivariative correlations  control for “Z”.

Third Variables
But, third variables do not always undermine a causal claim.
Intervening variable – a variable through which X influences Y.
 Intervening variables do not yield spurious correlations.
 They would be included in your causal logic.
Causal claim – a higher GDP (X) leads to a higher average life expectancy
(Y).
Antecedent variable – a variable that influences X.
 A  X  Y.
 Antecedent variables do not yield spurious correlations, as long as A
does not also affect Y through a pathway that doesn’t include X.
Problem when: A  X  Y … AND … A  Y.

The three types of third variables


Spurious Correlation – Z: undermines the causal claim.
Intervening Variable – I: does not undermine the causal claim.
Antecedent Variable – A: usually does not undermine the causal claim.

Reverse Causation
Causal claim – A greater number of guns (X) leads to a greater number
homicides (Y).
Causal logic (abbreviated) –

21
Reverse causal claim – A greater number of gun homicides leads to a greater
number of guns.
Causal logic – more gun homicides leads people to want to be able to protect
themselves which leads them to buy more guns.
This is reversing the positions of X and Y  different causal variable and
different dependent variable from original causal statement.
A reverse causation is when the direction of the causal relationship between
X and Y is the opposite of what we expected.
 Often undermines causal claim, but not always. Sometimes causation
does work both ways.

Reverse Causation
Also very common in social sciences.
Which causes which?
 Low income and low education.
 Democracy and wealth.
 Ethnic tension and poverty.
They could work both ways.
Idea here is that many of the causes and effects that we are interested in
mutually reinforce each other  causation often works in more than one
direction.

Randomness
Guns per Capita vs. Homicide by Firearm Rate (Developed Countries
excluding South Africa).
 Possible that this correlation is due to random chance.
 If we were to randomly draw points on a scatter plot there is a chance
they would look like this.
 Could be a sheer chance that more violent countries have more guns.

22
Even a perfectly random process will sometimes produce recognizable
patterns and apparent correlations.
The infinite monkey theorem  a monkey typing randomly for an infinite
amount of time will at one point write the word Hamlet.

Solution to Randomness
The solution: Statistics
 Uses probability theory to measure how likely it is a correlation
happened by sheer chance and not that they are causally related.
 Computes how closely correlated the variables are.
 More cases lead to more certain result.

Statistical Significance
An indicator of how likely it is that the correlation we observe is due purely
to chance.
Low levels of statistical significance are more likely due to chance, while
high levels of statistical significance are probably not due to chance.

P-Values
An indicator of the degree of statistical significance of a correlation.
Tells us how likely it is that we would have gotten this data pattern if there
really was no systematic correlation.
A low p-value indicates greater statistical significance.
Interpreting p-values…
 P-value is a value between 0 and 1.
 Most common threshold for “statistical significance”.
 The lower the p-value makes the chance less likely that the correlation
is due to randomness.

23
Real vs. Random Correlations
Question – When are correlations likely to be “real” and not random?
1. When the correlation is stronger.
2. When the correlation holds across more cases.
a. Law of large numbers.
b. Randomness has a smaller effect in a larger sample.

Testing with correlations


We essentially need to ask these three questions:
1. Spuriousness – Is the correlation because X and Y are both cause by a
third variable, Z?
2. Reverse causation – Does X cause Y, or might Y actually cause X?
3. Randomness – How likely is it that the correlation is due to chance?

Assignment 5
1.General causal claim – Economic recession leads to need for economic
policy change.
b.independent variable – economic recession
c. need for economic policy change

2.The correlation is that very often in the aftermath of an economic recession,


the country or countries involved bring forth new economic policies in order
to curb the current recession, or prevent future recessions from occurring.
Economic recession and the need for new economic policy vary together in
comparative studies of correlation.

3.The correlation may be spurious as there could be another variable


influencing the correlation. This variable could be public discontent for
governments in times of recession. Governments most of the time want to act
in favor of the population in order to keep their place in power. Thus, public

24
discontent with governments in time of recession may be a factor that would
influence governments to bring forth new economic policies.

b. The spuriousness of the correlation indicates that a correlation between the


two variables does not equal causation, therefore suggesting that my original
causal claim is not valid. This is as the correlation is not a result of the causal
relationship between economic recession and need for economic policy, but
rather another distinct variable.

25

You might also like