Advanced Orga 2

- Use advanced econometrics

- Do experiments

- include all variables that have an effect of independet variable

if members are not randomly asssigned to treatment group and control group, there could be differences in the group, which could be a variable that correlates with the independet variable. we would therefore not now, if an experiment outcome is happening because of the variable we want to test or the difference between the group. --> characteristics that cannot be controlled are distributed independently from treatment. 

randomization is important to have as similar group as possible and the only change between the treatment group and the control group should be the variable we want to test.

subjects alter their behaviour in experiments because they know they are being observed

Subjects act the way they perceive the experimenter wants them to behave.

- subjects know they take part in experiment 

hawthrone effect: they act different because they are observed

experimenter demand effect: subject at aligned with what the perceived desire of the researcher is

selection bias in the decision to take part in experiment: more risk seeking people who like challenges. if these characteristics have effect on experiment, the treatment will be overestimated.

Informed consent in the us.

in europe rules are more diverse

• changing peoples behaviour, without significantly changing incentives.
• without paying a lot of money and without forbidding something
• nudges must be easy and cheap to go around

z.B Früchte auf Augenhöhe und ungesunde sachen weiter oben im Regal.

Is an actual debate.

Interfere with people's liberty and autonomy for their own good.

--> goverment startet to use it for social good

research question: do academic field research on nudging scale? can we expect a similar impact when done in big implementations, as in small implementations (like academic papers)

they compared the academic papers from the meta analysis with the US Goverment Nudging data.

Method: 

1. exclude all lab study

2. exclude all study that have no control group

Result:
much smaller effect for government (in big implementations)
but still --> its not at a big cost.
its is still scalable, and significant.
only 8% increase (compared to 33.4%)

Why difference in result:
Publication bias (estimation of this is 60% of the difference): Academic incentives to publish good results: maybe only paper get published which found large impact on nudging
In Person contact: Some nudges might need in person contact. Possible in small sample, not possible in big governmental implementations.
Area: Researches focus on environmental topics, whereas goverments focus on tax compliance. Could be difference in the area
Difference in size: Sample size academia (500) and goverment (10000)

--> f.e gym membership

--> absence management: how can you reintroduce people who where sick for long?

Research Question: how can gym attendance be increased? (low cost)
can a reminder help?

Done with students, sending reminders each week.

Results:

no significant effect of treatment to control treatment.

change between pre treatment and during treatment. --> in both groups: control group and treatment group --> seems to have something to do with the time of the year. if that effect would have only been in treatment group, we would have thought reminder email has a negative effect on going to the gym.

--> maybe splitt sample into two groups: people that already go often in the gym, and people who dont go often. because on people already go often, the reminder might not help.

what is statistical significance is not necessary economically siginifcant and vice versa.

After looking at the groups of high gym attendance and low gym attendance, there is a significant effect for people with low attendance.

High attendance people might already go as often as it fits them and reminder thus has no effect.

Result: 

Reminders lead to an increase in gym attendance for low attendance users.

There is no effect for high attendance users.

Habit formation: No habit formation. No post treatment significance

Difference in difference: Calculated the difference of the difference between control and treatment group

• So are women less likely to choose competition than men? --> Can a default help to overcome this gender difference?
• natural field experiment. researchers are the employers.
• Participants: 477 freelancers (304 men and 173 women), who were unaware of the experiment.
• Participants completed a data entry job (scrape data about different companies).
• 3 Phases: 
  • 1st: Test phase: 5 min they enter data. they get 5chf + 0.25 per correct data entry
  • 2nd: Same as 1st phase. But You can choose between standart or an advanced task: if you choose advanced, its a competition, you get a high paying job in phase 3 (7.5chf and 1chf per correct entry) but only if you are among the best 25% in phase 2. the standart, you can work anyway in phase 3 but for the normal wage.
  • 3rd: better payed job for those wo won in advanced. and normal job for the ones that chose standart.
• Treatment variation: Opt-out: this treatment has the advanced (competitive) decision set as default. 
  Opt-in: this treatment has the standart (non-competitive) decision set us default. 
  Clicking the button changes the decision.

- Loss aversion when the default serves as a reference point.

- Default interpreted as a recommendation or social norm.

- Less cognitive effort when sticking to the default.

- Avoid switching costs (e.g., time).

- lazyness

- loss aversion: we dont want to change the default, because we feel like we lose the default. f.e we get one gift, and we first see the apple, then we are asked if we rather have a chocolate bar, we will not change. and vice versa.

• Results: Opt-in: 72% of men chose competition. 57% women choose competition. --> significance between gender
  Opt-out: 72% men, 67% women. --> no siginificance anymore between gender
  • In the phase 2, the people choosing competition, had the same performance than the people choosing standart (-> rules out nervousness)
  • Men and women are equally good
  • --> Women and men are equally likely to competete if default is on competition
• Reflexion: Why did they include phase 1? 
  • Because they want to test the ability of the people. Because they want to check if there is a gender difference in the task. That would not be good for the experiment, because the gender that is worse at performing the task might more probable choose the non competitive task. --> But no genderdifference found.
  • To make people used to the task, so that its easier for them to choose later.
• Replicated in the lab to find out if difference in nervousness:
  • Women said in the lab, that they were more nervous than men. But no stress effect on performance.
  • They made it in the lab, because in the natural field experiment, they could not ask, because people where not aware its an experiment.
• How could this be used for real life:
  • Everyone gets automatically in the pool for the top job
  • --> but maybe management position needs a competitive person, stress level could be also important

Research questions:How do workers react to different fixed wages? How should firms set their wages?

• One of the first experiments in this way.
• Task anyway needed to be done
• participants seat at the computer and enter data about book (title, author ect), participants can take break. they are always alone doing the experiment (so they don't compare each other)
• every 90 min somebody came and gave new books and in the back counted the amount of books done.

Gitft exchange doesnt seem to work in long term. 
but it can still mean that it makes sense to pay more as an employer (fe there could be intrinsic motivation, efficiency wage theory ect)