This exam consists of 5 questions, and you can earn 2 points per question (so 10 in total).
Always give a clear explanation to your answers and explain definitions/concepts when you
provide them. ‘Gambling’ is not rewarded: in case you explicitly write something that is false,
we can subtract points.
QUESTION 1
The following is the experimental design section of the paper by Andreoni, Nikiforakis and
Stoop (2017, working paper). Please read it carefully. All questions of this exam are based on
this paper, and for each question you will be given parts of the relevant text of that paper.
______________________________________________________________________________
3.1 Experimental treatments and procedures
In all treatments, a semi-transparent envelope was misdelivered in the mailbox of each
preidentified rich or poor household. All envelopes included an identical greeting card with the
following message written on the back:
“Dear Joost, here is €x for you. —Grandfather.”
The message to Joost was clearly visible from the back side of the envelope, as can be seen in
Figure 1, and hand-written for maximum authenticity. Joost is a real person, a research associate
of ours, living in a middle-class neighbourhood. His address was hand-written clearly on the
front side of the envelope which also contained a stamp and a post-mark. The latter made it seem
as if the envelope had been misdelivered by the mail company, when in fact it was
(mis)delivered by us.
A returned envelope is a pro-social/altruistic act as the action benefits Joost. To have a better
understanding of what motivates returns, we vary the contents of the envelopes across
treatments. Table 1 provides an overview of our experimental treatments. As can be seen, in two
treatments (BTC5, BTC20), in addition to the greeting card, we included a bank transfer card for
the amount of €5, or €20. A bank transfer card is an order to transfer money from one bank
account to another. It has no value for a third party, and it is impossible to go to a bank office to
cash it. Therefore, there is no monetary gain from keeping an envelope with a bank transfer card.
In the other two treatments (E5, E20), instead of bank transfer cards, a bank note of either €5 or
€20 was included.
Treatment
Content of the envelope
BTC5
Bank transfer card of €5
BTC20
Bank transfer card of €20
E5
Bank note of €5
E20
Bank note of €20
Table 1: Overview of the main treatments
Number of observations
W
X
Y
Z
3.2 Selecting poor and rich households
Each household in our sample was randomly assigned to one of the four experimental treatments.
Ideally, the selection of households would be based on their wealth. However, due to
(understandable) security and privacy concerns, CBS Netherlands was unable to provide us with
a list of households with their addresses and financial information in advance. They did agree to
provide us with household-level data on wealth, income, and other socio-economic variables
after data collection was completed. As a result, we needed to select poor and rich households
based on a measure other than wealth, but highly correlated with it. Our selection was based on
the value of the houses of the rich, and the rental prices of the houses of the poor.
Here is a graph that summarizes the results of the experiment:
______________________________________________________________________________
To summarize, the researchers make a selection of rich and poor households, and then randomly
assign these households to treatments. One of the researchers ‘misdelivers’ an envelope to a
household, and measure if the household returns the envelope. Returning an envelope is
interpreted as a pro-social act.
•
Consider the treatment E20. The return rate of the rich is 76% and that of the poor is
23%. Can we conclude from this fact that being rich causes a household to be (76/23 = )
3,30 times more likely to return an envelope that contains €20? If this is not causal,
would the true estimate be more, or less, than 3,30 times?
Answer Q1:
No, this is not causal. Being rich or poor is self-selected, and thus may be an over- or
underestimate. To see this, consider the equation about self-selection:
E[Y1i|Di = 1] − E[Y0i|Di = 1] + E[Y0i|Di = 1] − E[Y0i|Di = 0].
The observed effect is 3,30 times as much, but it may be that ‘potential altriusm’ differs between
rich and poor households. Any story goes, as long as it is well argued.
QUESTION 2
The following is the experimental design section of the paper by Andreoni, Nikiforakis and
Stoop (2017, working paper). Please read it carefully. All questions of this exam are based on
this paper, and for each question you will be given parts of the relevant text of that paper.
______________________________________________________________________________
3.1 Experimental treatments and procedures
[…]
A returned envelope is a pro-social/altruistic act as the action benefits Joost. To have a better
understanding of what motivates returns, we vary the contents of the envelopes across
treatments. Table 1 provides an overview of our experimental treatments. As can be seen, in two
treatments (BTC5, BTC20), in addition to the greeting card, we included a bank transfer card for
the amount of €5, or €20. A bank transfer card is an order to transfer money from one bank
account to another. It has no value for a third party, and it is impossible to go to a bank office to
cash it. Therefore, there is no monetary gain from keeping an envelope with a bank transfer card.
In the other two treatments (E5, E20), instead of bank transfer cards, a bank note of either €5 or
€20 was included.
Treatment
Content of the envelope
BTC5
Bank transfer card of €5
BTC20
Bank transfer card of €20
E5
Bank note of €5
E20
Bank note of €20
Table 1: Overview of the main treatments
Number of observations
W
X
Y
Z
______________________________________________________________________________
•
As is described, the BTC treatments contain an item that has no value to the subject
household who participates in this experiment. Please argue why or why not the five
precepts by Smith hold.
Answer Q2:
The key is to see ‘time’ as a reward mechanism. It costs time to do the pro-social act of returning
an envelope, not money. Then, it holds that:
Nonsatiation: holds, more free time is valued over less
Saliency: if a subject makes a ‘better’ choice, he/she will have more of the reward (better is than
determined by the subject his/herself).
Dominance: mailing a letter takes some time, it is not entirely trivial. So, subjects should be
encouraged to think hard about it.
Privacy: people don’t know that they participate in an experiment, so this is satisfied.
Parallelism: no need to hold.
0.4 points per correctly described precept.
QUESTION 3
Please consider again the following part of the paper by Andreoni, Nikiforakis and Stoop (2017).
______________________________________________________________________________
3.1 Experimental treatments and procedures
In all treatments, a semi-transparent envelope was misdelivered in the mailbox of each
preidentified rich or poor household. All envelopes included an identical greeting card with the
following message written on the back:
“Dear Joost, here is €x for you. —Grandfather.”
The message to Joost was clearly visible from the back side of the envelope, as can be seen in
Figure 1, and hand-written for maximum authenticity. Joost is a real person, a research associate
of ours, living in a middle-class neighbourhood. His address was hand-written clearly on the
front side of the envelope which also contained a stamp and a post-mark. The latter made it seem
as if the envelope had been misdelivered by the mail company, when in fact it was
(mis)delivered by us.
______________________________________________________________________________
In reality, there never was a ‘Grandfather’ (the cards were made by the research team). Also, in
reality, there was no mailman that misdelivered the envelope. The households received the
envelopes on purpose (without them knowing, of course).
•
Is this design using deception? Argue why or why not you think this is the case.
Answer Q3:
This can be argued both ways.
Deception: these things are not real, there is no grandfather
No deception: there is no explicit lie. No reputation is damaged.
QUESTION 4
The following is the experimental design section of the paper by Andreoni, Nikiforakis and
Stoop (2017, working paper). Please read it carefully. All questions of this exam are based on
this paper, and for each question you will be given parts of the relevant text of that paper.
______________________________________________________________________________
3.3 Randomization
“Using the data obtained from CBS Netherlands, we can evaluate whether the randomization of
households into treatments was successful. This is essential in order to be able to safely interpret
the findings from the experiment. Table 2 reports by treatment the means of the variables that we
collected and will use later on to establish the robustness of our findings. The variable ‘HH Age’
in Table 2 is the average age of all adults in a household. ‘Benefits’ is a dummy with value one if
at least one adult member of the household receives unemployment benefits, social welfare,
social security, or disability insurance. ‘HH Size’ presents the average number of persons in a
household, including children. ‘Pension’ is a dummy with value one if at least one adult member
of the household receives pension. ‘HH Foreign’ is a dummy with value one if all adult members
of a household are non-Dutch. ‘Distance to Joost’ and ‘Distance to mailbox’ measure the
distance in kilometers from a house in our sample to Joost’s house and the nearest mailbox,
respectively.”
The
Rich
Treatment Disposable Wealth
HH
BeneHH
Pension
HH
Distance Distance
Income
Age
fits
Size
Foreign
Joost
Mailbox
BTC5
87.641
3.451.325 55,63
0,04
2,76
0,47
0
4,13
0,32
BTC20
87.312
1.689.802 56,73
0,13
2,8
0,40
0
3,95
0,37
E5
84.616
2.749.497 56,81
0,07
2,73
0,44
0
4,16
0,35
E20
78.587
2.095.890 56,57
0,07
2,78
0,40
0
4,39
0,35
Average
84.539
2.496.629 56,44
0,08
2,77
0,43
0
4,16
0,35
Treatment Disposable
Income
BTC5
19.208
BTC20
18.924
E5
20.689
E20
17.276
Average
19.024
Wealth
26.32
15.655
53.747
13.223
27.236
HH
Age
42,23
45,41
43,29
38,88
42,45
Benefits
0,31
0,31
0,31
0,40
0,33
The
Poor
HH
Size
1,33
1,53
1,40
1,47
1,43
Pension
0,11
0,20
0,18
0,07
0,14
HH
Distance Distance
Foreign
Joost
Mailbox
0,13
3,55
0,24
0,18
3,53
0,25
0,13
3,52
0,25
0,16
3,52
0,25
0,15
3,53
0,25
Table 2: Observable characteristics of rich and poor households; entries are means
______________________________________________________________________________
This part of the paper looks at if observable characteristics of households are more or less equal
across the four treatments.
a) What test can be used to see if ‘Disposable income’ is randomized successfully within the
rich? State the null hypothesis, the alternative hypothesis and statistical test. (1 point)
b) Suppose you are forced to apply the Fisher Exact test using the variable ‘HH Age’ (you
can do whatever you want with this variable for the Fisher Exact test). What hypothesis
can you test with it? State the null hypothesis and alternative hypothesis (1 point)
Answer Q4:
a) Kruskall Wallis test. H0: disposable income is the same in all four treatments, H1: at
least one treatment differs from the rest.
b) Make a new variable, a dummy variable that has a value for 1 if someone is older than
50. Then, one can test the effects of age, for example: H0: people older than 50 have
similar return rates as people younger than 50. H1: people older than 50 do not have
similar return rates as people younger than 50.
QUESTION 5
The following is the experimental design section of the paper by Andreoni, Nikiforakis and
Stoop (2017, working paper). Please read it carefully. All questions of this exam are based on
this paper, and for each question you will be given parts of the relevant text of that paper.
______________________________________________________________________________
3.1 Experimental treatments and procedures
[…]
A returned envelope is a pro-social/altruistic act as the action benefits Joost. To have a better
understanding of what motivates returns, we vary the contents of the envelopes across
treatments. Table 1 provides an overview of our experimental treatments. As can be seen, in two
treatments (BTC5, BTC20), in addition to the greeting card, we included a bank transfer card for
the amount of €5, or €20. A bank transfer card is an order to transfer money from one bank
account to another. It has no value for a third party, and it is impossible to go to a bank office to
cash it. Therefore, there is no monetary gain from keeping an envelope with a bank transfer card.
In the other two treatments (E5, E20), instead of bank transfer cards, a bank note of either €5 or
€20 was included.
Treatment
Content of the envelope
Number of observations
BTC5
Bank transfer card of €5
W
BTC20
Bank transfer card of €20
X
E5
Bank note of €5
Y
E20
Bank note of €20
Z
Table 1: Overview of the main treatments
______________________________________________________________________________
•
With what ratio should the researcher allocate observations to the E5 and E20 treatment?
You can make assumptions where you see fit, as long as you argue why you make the
assumption.
Please provide this ratio four times:
o the poor: comparing E5 to E20,
o the rich: comparing E5 to E20,
o E5: comparing the poor to the rich,
o E20: comparing the poor to the rich.
Answer Q5:
The student should use the equation n0/n1 = sigma0/sigma1*[p1/p0]^ 1/2, and assume that the
T-test is used.
Then, one could argue that the sigma’s are the same in all treatments. The difference in
allocations between E5 and E20 should be 4^1/2 more observations in the E5 treatment.
o
o
o
o
the poor: comparing E5 to E20, 4^1/2 more observations in the E5 treatment
the rich: comparing E5 to E20, 4^1/2 more observations in the E5 treatment
E5: comparing the poor to the rich, same number of observations
E20: comparing the poor to the rich. same number of observations