4B Flashcards
What is the direction of a causal process?
- Causal effect in expected direction
- Spurious correlation
- Reverserd causal relation
- Bidirectional causal relation
- Causal network
Causal effect in expected direction
The obsered correlation is in line with what is expected
Spurious correlation
There is a statistical correlation between issue attitudes and left-right identification due to some third variable, even though there is no causal link between both
Statistical solution for spurious correlation
This possibility can be tested with a multiple regression model with control variables
Reversed causal relation
The observed correlation could be in the opposite direction
Bidirectional causal relation
The causal effect could also run in both directions simultaneosly
Statistic solution for bidirectional causal relation
With a panel study (observation on same cases at multiple time points) this can be examined with lagged effects
Panel study with lagged effects
Lagged effect
The delayed response of a dependent variable to a change in an independent variable (or variables).
- '’Are citizens’ issue attitudes predicted by what their left-right identification was at an earlier moment?
- Is citizens’ left-right identification predicted by what their issue attitudes where at an earlier moment?
Causal network
The independent variables could be a part of a causal network with many other factors that all influence each other.
- positive feedback loop: a causal network that reinforces itself
- negative feedback loop: a causal network taht cancels itself out
statistic solution: network models
These type of processes can be studied with network models, usually in combination with longitudinal data and/or computer simulations
These types of models are very common in economy and increasingly popular in psychology
Causal networks are rarely used in political science
Five reasons why correlation may exist
- X causes Y
- X and Y are both caused by a third variable Z
- Y causes X
- X causes Y and Y causes X
- X and Y are both part of a much larger causal network
The absense of a correlation does not imply the absense of a causal relation either
What are the steps of a causal process?
A causal process can be either direct or indirect
An indirect effect runs via something else
The causal chain:
A > B > C > D > E
For almost any effect you can think of some intermediate step
causal chain
Falling dominos are an example of a causal chain
The falling of stone A is a cause of the falling of stone B
But the falling of stone A is also a cause of the falling of stone C
Counterfactual: If stone A would not have fallen, than stone C would not have fallen
The effect of stone A on stone C is indirect because it runs via stone B
moderation
the association between two variables depends on another variable
three types of moderation
- moderator determines if there is an association or not
- moderator determines the magnitude of the association
- moderator determines the direction of the association
on what leel does a process take place?
individual levels (citizens)
aggregate level (electoral districts, cities)
ecological fallacy
conclusions on the individual level, based on analysis at the aggregate level
individualistic fallacy
conclusions on the aggregate level based on analysis at the individual level
