Week 10
Group 8
David Francis Newell, Moshweshwe King Mokoena, Sean Kevin McAvoy, Jack Nguyen
1. Draw a causal diagram, as in class. Include an outcome variable, a treatment variable, an observed confounding variable (control variable), an unobserved confounding variable, an instrumental variable, and a bad control variable. (Jack Nguyen)
2. If I want to estimate the (linear) causal relationship between the treatment variable and outcome variable under the assumption that there are no unobserved confounders, what linear regression model should I estimate? (David Francis Newell)
If there are no unobserved confounding variables, then a standard simple linear regression model can be used to estimate the causal relationship between the treatment and the outcome. One model should be done using the treatment and outcome as the variables, then another should be done without using the treatment, ‘the control’. This will determine the causal effect of the treatment on the outcome.
3. Suppose I am concerned that there are unobserved confounders that affect both the treatement and the outcome, and as such I cannot assume that the relationship between them is causal. How should I go about identifying the causal effect? (David Francis Newell)
In the case that there are concerns about an unobserved confounder affecting the outcome, a multivariable regression model can be used. In this case you can use the exposure, as well as selected measured variables as explanatory variables, and this will allow you to measure the causation between the outcome and the treatment. Provided that the adjustment for the confounding variables was sufficient, this will be provide an unbiased estimate of the effect of the treatment on the outcome.
4.
Describe the two properties that an instrumental variable must have. (Moshweshwe King Mokoena)
It must be related to the explanatory variable and uncorrelated with the errors, it has no direct effect with the outcome variable other than through the treatment variable
What is the exclusion restriction? (Moshweshwe King Mokoena)
Exclusion restrictions are observed variables that impact treatment assignment, but not the outcome of interest conditional on treatment assignment. Basically, if one is interested in the effect of a “treatment variable” on an outcome variable, and the treatment is not exogenously assigned, then one may perform causal inference by exploiting the presence of variables that causally affect the treatment status but do not have a direct causal effect on the outcome. Exclusion restrictions are identifying restrictions, so they can not be tested. This means that empirical results critically depend on the validity of the exclusion restriction.
What is the relationship between an instrumental variable and a natural experiment? (Sean Kevin McAvoy)
A Natural Experiment is an experiment where the conditions are determined by nature, i.e. outside the control of the investigators. As such if X and Y are the variables then in a natural environment these variables with have an array of variation due to a number of different factors making it difficult to gauge the relationship between X and Y. An Instrumental Variable Z is a variable that can be useful in determining the relationship between X and Y though through one aspect only, not whole relationship just a good indicator of a major relationship. Z is to have not direct impact on Y and only affects Y through changes in X. This allows for the investigators to see the relationship between X and Y through changes in X by Z and its indirect affect on Y and account for the natural variations that exist in a Natural Experiment.
5. PIH
A Test of the Permanent Income Hypothesis on Czech Voucher Privatisation (Sean Kevin McAvoy)
The data used in this experiment was a survey done by the researchers in the Czech Republic after a 3 year period. The Natural Experiment is the consumption levels experienced by the Czech people across the time frame from the early 90s.
The exclusion restriction in this experiment is consumption. The vouchers would have no direct impact on permanent income and would only change it through its affects on consumption. The exclusion restriction in this experiment does make sense as testing to see if the PIH holds then an increase in a temporary payment should have no impact on consumption and providing evidence for PIH.
The results were that with the windfall from the vouchers there was little to no rise in consumption across the Czech Republic. This provides evidence for the PIH as only changes in permanent income would alter the consumption levels.
(Moshweshwe King Mokoena)
The researchers have used the natural experiment associated with the Dutch postcode lottery (PCL) to study the own and social effects of a temporary, unexpected income shock equal to about eight months of income on households’ consumption behavior and self-reported happiness. The data used was from questionnaires sent out to lottery winners and a sample of non-winners in the postcode lottery.
The exclusion restriction, Given the inherent randomness in the prize draws and absent externalities between winning post codes and non-winning postcodes, participants in non-winning postcodes constitute a valid counterfactual for participants in winning postcodes. This allows the researchers to test for the effects of unexpected, temporary income shocks (both cash and in kind) on winning households’ consumption and happiness under quite general conditions. Similarly, under the above conditions nonparticipants in non-winning postcodes constitute a valid counterfactual for nonparticipants in winning postcodes. This allows for a clean test for social effects of income shocks on non-participating households. Their findings.
For PCL participants in winning and non-winning codes the researchers found statistically significant post-lottery differences in expenditures on food away from home, other monthly items, and total monthly expenditures. They also found that participants in winning codes were 4.5 times as likely to initiate major exterior home renovations during this period and spent over \(euro\) 500 more on durables than participants in non-winning codes. This finding is consistent with a permanent income model in which households adjust the timing of their durables purchases to smooth consumption.
A test of the permanent Income Hypothesis on Alasken Households, Hsieh (2003) (Jack Nguyen)
The data is of Alaskan households from the consumer expenditure survey. The natural experiment is tax refund. They are anticipated as the tax payers has to calculate it themselves when filing the tax return.
The exclusion restriction is dividend payments. These payments are from oil royalties redistributed by Alaska’s Permanent Fund.This exclusion is satisfying as all residents receive this payment regularly and doesn’t change current consumption.
They find that only 28% of the tax refund is spent. Households that show sensitivity to tax refunds do not show such sensitivity to the dividends from Alaska’s Permanent Fund. This supports PIH as the tax refund is seen as temporary shock and consumption doesn’t change.
Fiscal
Evidence on the Fiscal Multiplier from a Quasi Experiment (Sean Kevin McAvoy)
The data used in this experiment was output and public spending investment across a 10 year period across the provinces of Italy. The Natural experiment is the output and spending that happened across Italian provinces across the 10 year period.
The exclusion variable is government spending as only affects through spending would have an impact on the output. This exclusion makes sense as the fiscal multiplier is being tested so any instruments can only affect Y through changes in spending.
The results found were that there was a government multiplier as such when spending changed there was greater impact on output. Any changes that could affect spending could then have the capacity to cause greater harm to output.