POLI 110 Notes - RRR PDF

Preview

Summary

RRR...

Description

POLI 110 (2013 Winter Term)- Alan Jacobs POLI 110: Investigating Politics- What is Social Science? (Lecture 5-9/13)  

Ezra Klein’s tweet what is distinctive about science? Clicker question fundamental scientific inquiry?

A claim  A statement about the world  What is a claim based on?  Why should we believe the claim to be true?



Base for claims  Consider the claim: Foreign aid is beneficial  How do we know?

 Authority (A claim from authority): arguing that a claim is true because a person in position of authority says it is true

 Problems with authority as a basis for claims 1) Expertise is no guarantee being right 2) Even authorities may have an agenda 3) To easy to cherry-pick the authorities who agree with us 4) What do we do if authorities disagree?  Common sense (A claim from common sense): arguing that a claim is true because it’s what “everyone knows” or “just makes sense”.

 Why common sense can lead us astray: 1) Often based on superficial similarities between things (doesn’t account complexity of issue and the basis of two similar things may have very different inner characteristics) 2) Often based on analogy between very different situations 3) Often strong reasons why the opposite also makes sense (two face arguments which make the claim controversial)  Personal Claim (A claim from personal experience) : a claim based on one’s own personal (nonsystematic) observations or on one’s own reaction to an observation.

Sources of measurement bias:  Upward bias for countries with freer markets  Downward bias for countries with higher taxes, more regulation

POLI 110: Investigating Politics- Challenges of Measurement (Lecture 17-10/28) All measurements are with errors Random Measurement error vs. Measurement Bias  Small samples diff from answer and unique among diff groups  Larger samples more similar to each other and the answer 1) Random sampling (numbers jump around) 2) Sampling bias (tilted one way or the other)

POLI 110: Investigating Politics- Challenges of Measurement (Lecture 18-10/30) Too much to measure  For many measurement tasks, w cannot measure all instances of a phenomenon  We can only take measure of a subset  Population: the full set of cases (countries, individuals, wars, etc.) that we’re interested in learning about  Sample: Subset of the population that we actually take measures of  drawing a inference: from sample to population, things we have not observed from things we have observed  Examples of Sampling  We want to measure unemployment in Canada  Population: all members of labor force  Sample: 56,000 members of labor force  In march 2013, we want to know now people plan to vote in May 2013 BC election

POLI 110 (2013 Winter Term)- Alan Jacobs



Population: eligible and like BC voters Sample: 803 BC voting age citizens  We want to know whether British MPs tell the truth in legislative debates  Population: all statements made by British MPs Sample: 250 speeches made by British MPs  We fact-check these 250 speeches and we draw an inference from this sample measure about the population

Methods of sampling  How do we know that our sample is like our population?  How do we know if our sample is representative of our population  Most common strategy:  Random Sampling: selecting cases from the population in a manner that gives every case an equal probability of being chosen  Random Sample + Large numbers= high representatives= more certain inferences from sample population Sources of error in Sampling  Sampling error: the difference between the measure for a sample and the true value for the population  Random Sampling error: sampling error caused by random variation between samples  By pure change, one random sample of a population will be somewhat different from another random sample of the same population  And different from the population average  SOLUTION  Increase the size of your sample -Random noise “washes out” with larger numbers  Sampling bias: sampling error caused by a feature of the sampling procedure that makes some members of the population more likely to be sampled than others  Sampling bias sources: 1) Sampling frame =/ population  Sampling frame: group from which you draw a random sample  Example: election poll based on random sampling from phone book  Example: study public agenda based on random sample of New York Times article 2) Self-selection  Respondents often have control over whether they join your sample  Example: who decides to take a survey on environmental issues?  Example: Internet pools based on people clicking “your opinion counts” banners POLI 110: Investigating Politics- Challenges of Measurement (Lecture 19-11/01) Variables  A variable is a measurable property of a phenomenon that can potentially take on different values (those “values” may be numerical or categorical)  The variation may be across units or over time

POLI 110 (2013 Winter Term)- Alan Jacobs     

Income (numerical) Whether a count is at war (categorical) Level of public support for a prime minister (numerical) Regime type (categorical) Ethnicity (categorical)

Causes and effects as variable  Causal claim: democracy increases economic growth  Cause= democracy  LEVEL OF democracy  Effect=higher economic growth  RATE OF economic growth  Independent variable= level of democracy= the cause  Dependent variable=rate of economic growth= the effect  higher level of democracy leads to a high rate of economic growth  

Independent variable: the variable capturing the suspected cause in a causal clam Dependent variable: the variable capturing the suspected effect in causal claim



Causal claim: centralized political authority leads a country to develop a generous welfare state Independent variable=degree of centralization of political authority Dependent variable= level of welfare-state generosity

       

Causal claim: a greater level of national wealth causes the population to have a higher level of happiness Independent variable: level of national wealth Dependent variable: level of happiness Causal Claim: Civil war is cause by a lack of material resources Independent variable: Level of material resources Dependent variable: Presence /absence of civil war

POLI 110: Investigating Politics- Testing Causal theories: comparison and correlation (Lecture 19-11/04) Causal claims  The conservative party of Canada won the 2011 election because the Liberal party had had a major scandal  Concentrated political authority caused Sweden to develop a generous welfare state From causal claim to empirical test  An empirical prediction: a statement about what we should expect to observe if the claim is true  An empirical prediction connect claims to evidence  Tells us what kind of evidence would support or refute a claim  Allows us to test the claim

POLI 110 (2013 Winter Term)- Alan Jacobs   

 

So: what should we see if X causes Y? To say “C is a cause of E” means: if c had not happened, then would not have happened “The conservative party of Canada won the 2011 election because the Liberal party had had a major scandal” Implied counterfactual… if the liberal party had not suffered major political scandal then the conservative party would not have won the 2011 election Less of Xless of Y Claim X affects Y -Implies the counterfactual…If claim is true, then in a case with same level of X and some level of Y  If the level of X had not been diff, then the level of Y would have been diff

The fundamental problem  Fundamental problem of causal inference  We never get to observe a case under the counterfactual condition; we can only observe each case under one (the factual) condition  We observe Canada in the last decade with the liberal party suffering scandal  Would the conservative have won if the Liberals had not had a scandal?  We will never get to observe this  We only observe every case with the level of X and Y that it actually had  What should we observe if X caused Y? A partial solution: the comparative method  Back to that definition of causal effect: X had an effect on Y in a case means:  If X had taken on a different value in that case, then Y would have taken on a diff value  We can’t re-run history for that case, BUT:  Maybe we can find another case that is a lot like this case, but where X has a different value A Case  A single instance of a phenomenon under investigation. A case has an outcome it takes on one value on a dependent variable  E.g. an election (why do some parities win and some lose elections?  E.g. An individual surveyed  One country’s welfare state (why do some countries have bigger welfare states than others?)  A war between two countries (Why do wars happen?) The comparative method  The comparative method tests an empirical prediction  If the XY theory is correct, then  Prediction to test: -If we observe two cases to be the same in all relevant respects except for the value of X, then we should observe that the two cases differ in the value of Y -AKA. The method of difference Method of Difference

POLI 110 (2013 Winter Term)- Alan Jacobs   

Puts differently… Theory: C causes E Hypothesis: if we observe two cases like Case 1 Case 2 A is present A is present B is present B is present C is present C is ABSENT Then… E should be present here E should be absent here

The comparative methodMethod of differenceWelfare states  Start with Canada  Concentration of authority (at federal level) high X  National health insurance: present Y  We don’t get to re-run Canadian history with a different level of concentration of authority BUT:  Maybe we can find another country that is similar to Canada except for the degree of the concentrated authority  Empirical prediction: if we find a second country that is similar to Canada in all relevant respects EXCEPT with low concentration of authority, then we should no NHI      

Canada US Liberal political culture Liberal political culture Public support of NHI Public support of NHI Dominance of private insurance till 1960s Dominance of private insurance till 1960s Moderate trade union strength Moderate trade union strength X (concentration of political authority)= High X (concentration of political Authority)= LOW Canada: Y=NHI present US: Y=NHI absent The Comparative method  Theory lack of material resources causes civil war  Comparison: find two countries with similar in all relevant respects except X: level of material resources  Test: difference in civil war (Y)?  Theory: Democracy leads to faster economic growth

POLI 110: Investigating Politics- Testing Causal theories: comparison and correlation (Lecture 20-11/06) Case 1: UBC Values Case 2: U of T Values High/middle/low ranking of High ranking High ranking university

POLI 110 (2013 Winter Term)- Alan Jacobs X1: local quality of life X2: acceptance rate X3: provincial level of funding

High Low High

High Low Low

POLI 110: Investigating Politics- Testing Causal theories: comparison and correlation (Lecture 21-11/08) Midterm Revision Notes: -Conjunctural: depends on a combination of causes (requires both…and…) -Sufficient and Necessary condition It would not necessary because bad economy is the only way to cause govt defeat No govt can lose unless there is depression (then this would be necessary condition) Cause always produces an effect (re-write: a govt will always lose the election if the economy is in depression= No government can win an election if the economy is in a depression) -Do not rely on a “keyword” but need to think about the meaning behind the logic -Conjunctural: and together -Necessary and sufficient (diff) Testing theories by looking for correlations  Correlation is a clue to causation  Comparative method -Relies on correlation of cause and effect More generally, -One empirical prediction of most causal theories is that X and Y should be correlated if the theory is true  Correlation  A relationship across cases between the values that two variable (X and Y) take on  A positive correlation: cases with high values of X also have higher values of Y  A negative correlation: cases with higher values of X have lower values of Y  Can look for correlations across large numbers of cases  We may except X and Y to be correlated -Across units -Over time      

Theory: a higher rate of pre-election income growth (X) increases the vote share received by the incumbent party in a presidential election (Y) What should we expect to see if the theory is true? A: Positive correlation between X and Y B: Negative correlation between X and Y Empirical prediction: if theory is true, we should observe a positive correlation across elections between the rate of income growth and the incumbent party’s vote share Elections with higher x should have higher Y

POLI 110 (2013 Winter Term)- Alan Jacobs   

Theory: a higher level of inequality (X) in a country reduces the degree to which citizens participate n public life (Y) If theory is true, we should observe a negative correlation across countries between the level of inequality and the level of participation in public life Countries with higher x should have lower Y

From theory to evidence 1. Identify a phenomenon you want to understand  Why does this outcome happen?  Murder rate=effect=dependent variable (Y) 2. Develop a causal theory Suspected cause=x=number of guns -More gunseasier to use deafly force in disputemore use of deadly force 3. Ask yourself: how would I know if more guns caused a higher murder rate?  Make an empirical prediction  I predict a positive correlation between X and Y across countries 4. Now, go and look for predicted correlation Take measurements of X (number of guns and Y (murder rate) in a set of countries Is x positively correlated with Y across countries?

POLI 110: Investigating Politics- Testing Causal theories: The trouble with correlations (Lecture 21-11/13)  Eating Chocolate (X) makes you more likely to win a Nobel prize (Y)  Spurious Correlation: when two variable are correlated but that correlation is not a result of a causal relationship between those two variables Spurious Correlation  One common source of spuriousness Z Z Causes

Causes

X

Y Correlation

Not Causal

 

Does foreign aid hurt economic growth? Peacekeepers (X) and length of Civil War (Y)

POLI 110: Investigating Politics- Testing Causal theories: The trouble with correlations (Lecture 22-11/18)

POLI 110 (2013 Winter Term)- Alan Jacobs

GDP (Z) Z Causes more Degree of democracy (x)

Causes More

Life Expectancy (Y) Positive Correlation

May not be causal

The basic problem: why spurious is so common  So many of the causes social scientists are interested in “cluster” together  Low income and low education  Democracy and high national wealth  Democratic institutions and liberal political culture  Crisis and aid  Our X’s will often be correlated with Z’s that may also be the cause  If we suspect that a third variable (Z) is generating a spurious correlation between X and Y:  Include Z in the analysis  Examine the correlation between X and Y for cases with the same value of Z ”Controlling for” Z

GDP Causes more Causes More Degree of democracy (x)

Life Expectancy (Y) Positive Correlation

POLI 110 (2013 Winter Term)- Alan Jacobs

May not be causal

Intervening variable  A variable through which X influences Y  If XiY: I is an intervening variable  Intervening variable do NOT yield spurious correlations  X still effects Y Antecedent variable  A variable that influences X  If AXY: A is an antecedent variable  Antecedent variable also do NOT yield spurious correlations as long as A does not affect Y through a pathway that doesn’t include X  X still affects Y POLI 110: Investigating Politics- Testing Causal theories: The trouble with correlations (Lecture 22-11/20)  Do more guns cause more murders?  Reverse causation # Of guns (x)

# Of gun murders (Y)

Maybe Y causes X



Does clean government increase growth? Maybe X causes Y

Level of corruption (x)

Rate of econ growth (Y)

POLI 110 (2013 Winter Term)- Alan Jacobs

Maybe Y causes X

   

So many of the causes and effects social scientists are interested in mutually reinforce on another: Low income and low education Democracy and high national wealth Democratic institutions and liberals political culture Our X’s will often both cause and be caused by our Y’s A perfect correlation Mur der Rate

Guns per 100 people

The problems of randomness  Even a perfectly random process will sometimes produce recognizable patternsapparent correlations  A monkey tying randomly for long enough will eventually produce hamlet  Some of the correlations we see in the world are random, not real  How can we tell the difference? Statistics: What’s it good for?  Statistics helps us avoid being fooled by change  Uses probability theory to tell us how likely it is that correlation happened by change  Computes how closely correlated the variable are  When x is higher how often is Y higher or lower

POLI 110 (2013 Winter Term)- Alan Jacobs  Takes into account how many cases we have   Pattern in LOTS od cases less likely to be random  Law of large numbers  Tells us the probability that the correlation we see is “real” or random Statistical significance  An indicator of how likely it is that the correlation we observe is due purely to chance. A higher degree of statistical significance means it is less likely it is that the result derived from chance.

POLI 110: Investigating Politics- Testing Causal theories: The trouble with correlations (Lecture 23-11/22) More than 10% chance is usually considered not significant POLI 110: Investigating Politics- Beyond correlation: Process Tracing in Case Studies (Lecture 23-11/25) Reviewing Correlations  XY  Look for predicted correlation between X and Y across cases  If we find the right correlation, a clue that X affects Y  But then ask: Is the correlation real or random?  If real, could it be spurious?  If so, look for correlation again controlling Z  If not spurious, could it be a result of reverse causation?  Evidence that the causal logic played out in the case: process tracing POLI 110: Investigating Politics- Beyond correlation: Process Tracing in Case Studies (Lecture 23-11/27) What causes ethnic wars?  Theory: Group hatred  X1: Members of each group hate and fear other group   Threatening and hateful discourse and symbols stir up passions on both sides  o Evidence: Nimieri (northern leader o Imposes sharia on whole country o Begins dressing in Arab clothing o Declares Islamism the core of his political program o Groups in the south Saw abolition of Sharia as top priority Talked of “Arab” northerners as slaveholders 

Provides violent acts o Southern liberation army rebels against imposition of Sharia

POLI 110 (2013 Winter Term)- Alan Jacobs  

Fear and hatred lead to spiral of violence  All our war (Y)

  

Theory: Elite manipulation X2: Government officials from group A face risk of losing power Provoke violence between groups to shore up support from Group A members o Nimieri provokes violence BUT this weakens his powers  Loses key allies in South Gives up control over barks and army to Muslim Brotherhood Leads to his ouster Members of Group A don’t want war, but fear attack from Group b o Presents evidence that:  Northerners did not fear attach from south  Sought to cultur...