Lab Experiments

Question 1

Pros and Cons of Lab experiments

Answer 1

Pros:

• Randomisation
• Control for external/confounding factors

Cons:

• No natural environment/social element
• Hard to Generalise (External Validity)
• More costly (recruiting, skilled people needed…)

Question 2

Necessary conditions for causation

Answer 2

1. Correlation
2. Time order: first independent variable variation, and then dependent variable.
3. Non-Spuriousness.
   Spurious = false, not genuine, when changes in both variables are actually cause by a third.

Question 3

Effect of Randomisation on identification of Causality

Answer 3

Without Randomisation
Difference in expected outcomes = ATE + Selection bias

with Randomisation:
Expected outcome given Treatment = Expected outcome given Control
hence,
Difference in expected outcomes = ATE

Question 4

Definition of Internal Validity

Answer 4

Truthfulness of an assertion that A causes B.

Question 5

Threats to Internal Validity in Lab Experiments

Answer 5

1. Maturation Effect: changes due to naturally occurring internal processes.
2. History Effect: Exogenous external shocks or changes over long periods of time.

Remedies:

Treatment Identification
1. Hawthorne Effect: subjects modify their behaviour in response to their awareness of being observed.

Remedies:

2. Demand Effect: changes in behaviour by subjects due to cues about what constitutes appropriate behaviour.
   (reading instructions, participants realising treatment)

Remedies:

3. Placebo Effect: Effect on the subject is due on the subject’s own belief that the treatment will have an effect.

Remedies:

4. Expectancies on Experimental Staff: experimenter has positive expectations on the treatment and will induce a positive effect on the treated.

Remedies:

5. Testing effect. Performing the test twice improves your score, not because of the change over time.

Remedies:

6. Regression to the Mean. If the first observation in extreme, the second will likely be more moderate.

Remedies:

7. Order Effect. the order in which the treatments are delivered can be confounded with the effect itself.

Boredom/Fatigue, Demand effect, Carryover effect (answers to current item might depend on previous answers).

Remedies:

8. Operation confound. It occurs when the measure measures something else then intended.

Remedies:

9. Instrumental Decay.