Part B Flashcards
explain how a randomised controlled trial works
group of patients,
random assignment into treatment group or control group,
follow up to compare results
what bit is crucial in the randomised controlled trial (RCT)
the random assignment
what is the basic selection problem in food stamps
cannot just compare recipients with non-recipients as they are different,
income, household consumption
does the basic selection problem with food stamps (can’t just compare recipients with non-recipients) go away if you control for income and household composition
even if you just focus on eligible, not everyone receives food stamps (20% eligible don’t),
can’t just compare eligible recipients with eligible non-recipients because they are different, self-selected group selection on unobservable factors likely to be correlated with spending outcomes
what is internal validity
statistical inferences about causal effects are valid for the population being studied
what is external validity
statistical inferences about causal effects can be generalised from the population and setting studied to other populations and settings
what are some features of internal validity for whitmore study
fewer assumptions than DiD,
random allocation into treatment and control groups,
test that characteristics the same across the two groups
talk about external validity in terms of laboratory experiments
controlled environment to study behaviour,
may be unrepresentative population (students),
arguably artificial setting
talk about external validity in terms of field experiments (= randomised controlled trials)
natural setting and relevant population,
arguably limited scale (compared to genuine policy change),
issue if people know they are in an experiment (will they behave differently because they are in an experiment and will they expect the experiment to last)
talk about the methodology of Whitmore
randomised controlled trials (RCTs) run in San Diego (14 months) and Alabama (8 months) in early 1990s,
cash out experiment,
half eligible households received stamps as usual, half received cash,
then compares stamp infra-marginal (distorted) with check infra-marginal (distorted)
why are eligible recipients different with non-eligible households
recipients different,
low income households,
more likely to have children
why is it difficult to compare eligible recipients with eligible non-recipients
is a better strategy (than recipients to non-eligible),
self-selected group with selection on unobservable factors likely to be correlated with spending outcomes
what is the selection problem for eligible recipients vs eligible non-recipients
some selection process where everyone who’s eligible decides whether to apply for the benefit or not and this selection process is linked to the type of factors that are going to matter for how much they spend on food
what is exogeneity problem
treatment is determined by factors that do not affect the outcome of interest
p value in whitmore
probability that these outcomes will be generated under the null hypothesis,
probability that two are actually the same
how does Whitmore know who’s distorted and who’s infra-marginal
observe food stamp level (value of food stamps),
distorted if they spend less than or equal to the value of the food stamps plus $5 (to allow for rounding)
what is the other way of finding out what proportion are distorted (not observe level and if spend less distorted)
all based on how much food would they spend if given cash,
because random selection two groups should have same distribution of food spending so look in cash group and find percentage (say 25%) distorted, therefore know that bottom 25% in terms of food spending are also distorted
what percentage of distorted households do both methods of finding proportion of distorted find
both approaches give around 20% distorted and 80% infra-marginal
what two experiments does the abeler and marklein paper have
field experiment - german restaurant,
lab experiment - german students in computer lab at university
explain setup of both experiments in abeler and marklein
budget consists of cash endowment (own money in restaurant and given chance to earn money for lab),
also given additional “grant”,
interested in effect off giving a grant in cash compared to if grant given targeted at one of the goods,
level of targeted grant set such that consumers are infra-marginal (non-distortionary)
talk about the setup of the restaurant experiment in abeler and marklein
wine restaurant in Southern Germany,
goods=food and drinks,
additional grant= 8 euro voucher presented to guests on arrival to be spent on that day,
two treatments + baseline (three groups of restaurant goers) - cash treatment (gourmet voucher used to reduce overall bill), label treatment (gourmet beverage voucher for drinks), additional baseline (no voucher)
results of average spent for field experiment abeler and marklein *
label treatment increases spending on labelled good compared to cash treatment,
overall spending higher than under cash treatment,
consumers are infra-marginal, so difference between cash and label treatment are evidence of behavioural effect (because economic theory says they should act the same as they are both infra-marginal)
how is the treatment decided for the restaurant experiment abeler and marklein
it is done on a day to day basis
what are the strengths of the lab experiment compared to the field experiment for abeler and marklein
random allocation across individuals (not by day),
control over background factors,
control over experiment design
what is the design of the lab experiment for abeler and marklein
game involves allocation of budget (income and additional grant) between two goods,
random allocation of students into cash group (control) and label group (treatment),
student earns 50 points of initial income through a mundane task,
stage 1 allocation of income,
grant given as cash or as targeted benefit that can only be spent on one of the goods,
stage 2 allocation of income
