Online fundraising and ask avoidance Flashcards
What is ask avoidance and how is it preserved?
People are willing to make an effort to not donate to avoid instances where self image may be threatened.
It is preserved when there are alternative methods of paying for an item e.g. mail instead of online.
What is the paper we will look at here?
What is the paper by Adena et al (2020) about?
The paper is about the role of self image in charitable donations.
Context: unavoidable online request for donation at the time of checkout when ordering opera tickets.
Bottom line : customers buy fewer tickets online in the following season.
What is the experimental design of this experiment?
An opera house in Germany introduced an online fundraising tool for a period of approximately 3 months
we have 3 treatments in 3 subsequent time periods
1) Baseline( normal you are asked to donate ( 10/20/50/100) if not you continue before purchase)
2) Push up grid donations - ( higher donation grids after 28 days ( 20/50/100/200) suggested amounts ) - you can continue if you don’t want to do it.
3) Force non-donors to tick a box - ‘Do you wish to make a donation and you have ‘i have already donated and no thank you’- this is after 33 days, you have to do this step you cannot skip.
REMEMBER EVERYTHING IS BEFORE CHECKOUT.
What is the summary finding and why is the findings what they are?
They find that the last intervention has a strong intervention, more people are donating and donating larger amounts.
This is because in the first 2 stages they can self deceive themselves that they didn’t notice the donation requests , thus don’t perceive themselves as a bad person. Whereas in the 3rd stage you need to confront yourself that your not actually donating ( it becomes salient), so to avoid not feeling bad about not donating( preserve a positive self image, people donate more.
Why do people give to charity?
Intrinsic reasons (altruism)
Extrinsic reasons ( tax breaks)
Image reasons: others perception and self perception.
What us a Quasi-Experiment?
Quasi-experimental design attempts to establish a cause-and-effect relationship by using criteria other than randomisation.
How is the Experiment in the paper by Adena et al a quasi experiment?
The assignment of subjects are not quite random here.
So they do the baseline treatment for a while ( so anyone that purchases tickets during that period is in the baseline group), then they do the higher donation grid for a period and everyone who purchased tickets there is in the higher grid. Finally, the same for ticket box treatment, depending on the period.
Why are customers who buy more than one treatment period excluded from the analysis ?
To prevent spillover effects, you don’t want people’s behaviour to be anchored due to already having a treatment.
Interpret this.
T1: baseline treatment, most people donate the minimum amounts 10/20.
T2: higher grids. the amount of donations decrease and people donate the minimum amount which is 20.
T3: Number of donations increase again but people are picking the minimum amount 20, so overall revenue increases
I will now look at the results, firstly by looking at does an individual donate,
So dependent variable here is if the individual donates or not. its a 0 1 variable.
The results are relative to the second treatment( higher grids, so the base is higher grid treatment, NOW INTREPRET IT.
Explain why we have different Logit ( columns
HINT LOOKING AT LAST COLUMN WITH THE DIFFERENT CONTROLS HELPS.
1) Relative to the high grid treatment, significantly more people donate in the low grid and T3 ( so when you go from low grid to high grid, you get a significant decrease in the amount of people that donate and when you go from high grid to t3, you get significantly more people donating.
2) its the same regressions but with different control variables, so checking whether adding control variables makes a difference, checking if results are robust.
Looking at results again but this time we will look at how much do they donate= dependent variable
Again we have different columns with different control variables for robustness, now interpret it?
Also everything is relative to high grid treatment.
Given that coefficients are positive and significant, people donated more 9( with the expectation of column 3 where control affected it) then more people donated with lower grids than higher grids and also T3. So change in website design had a significant effect on the amount of money donated.
Why is time control a reasonable control to have to check the robustness of results?
Why is ticket controls at t a reasonable control?
If you book tickets in advance, most likely to deviate but if closer to the date then donate more/less ( not really sure)
current prices and demand might affect about tickets brought extra.
Why do we have to be cautious with quasi experiments and how does Adena rectify this problem?
Because of the non-random assignment, we could argue there was a difference in those who purchased ticket in the first period vs second and third period rather than different treatment ( e.g. different types of people might buy a ticket in jan compared to feb, affecting our results)
To reinforce robustness of results, they repeat the analysis using only data for x days before and after changes( e.g. 4 and 5 days) to see if we get same results, as its hard to argue that its because of different types of people.
For lower grids the coefficient is significant and above 0, hence they donate and they donate more than t2.
Same for T3
So even if you use smaller or bigger windows around the date you change treatment, you still get the same results, thus unlikely that its due to different people purchasing at different periods.
What is the whole point of table 3 ( clue interaction effects)?
They add interaction terms ( past treatment dummy and past customer dummy but find no interaction effects.( column 1 and IV)
Maybe a point you could say that they tested for interaction effects of festive ticket buyers with the T3 ( none sold in t1, hence only T3) and found that festive buyers respond less to t3, although main coefficient is significant and increases in magnitude.
Now we want to look at the following year, are people going to act differently in terms of purchasing tickets the following year ( is there evidence of ask-avoidance)
Dependent variable is the number of tickets the individual buys in the next season.
Interpret these results
Compared to high grids there is no significant effect of being in low grid but there is a significant effect and negative of being in T3. So if you were forced to tick a box, either no thank you or donated already, if you are in that treatment, then you are significantly less likely to purchase tickets online the following year.
Now we want to look at how much money is spent online after a year
Those in the lower grids spend signficantly less money online in comparison to high grid treatment.
The ones that were in t3, we see is mostly negative and significant that people spend less money online after a year if in t3.
Interpret the long term effects on all tickets?
Dependent variable is =number of tickets that are brought in any way. ( HINT LOOKING AT LAST COLUMN WITH THE DIFFERENT CONTROLS HELPS.)
The coefficient on T3 is not statiscially significant so this means that those people next year who buy tickets in T3 dont buy significantly fewer tickets in comparison to T2. They switch from buying tickets online to other ways, such as box office. Note everywhere else there are significant effects of less tickets purchased overall despite less control variables.
Same can be said with T1.- not statiscially significant so means that those people next year who buy tickets in T1 don’t buy fewer tickets in comparison to t2, they switch.
