Step 1: Identify bias types

• Undercoverage bias
• Response bias (includes question wording, social desirability)
• Nonresponse bias
• Voluntary response bias

Step 2: Analyze scenario (a)

Phone survey: landlines during business hours

Problem: Systematically excludes certain groups

• Young people (mostly use cell phones)
• Working people (not home during business hours)
• Lower income (may not have landlines)

Type: UNDERCOVERAGE BIAS

• Some groups in population have no chance of selection
• Sample not representative

Step 3: Analyze scenario (b)

Question: "Don't you agree the mayor is doing a terrible job?"

Problems:

• Leading/loaded question
• Suggests a "correct" answer
• Uses negative language ("terrible")
• Pressures respondent

Type: RESPONSE BIAS (Question Wording Bias)

• Question influences how people answer
• Doesn't measure true opinions

Step 4: Analyze scenario (c)

Online poll: strong opinions more likely to respond

Problem:

• People with extreme views participate more
• Moderate people skip it
• Not representative of population opinions

Type: VOLUNTARY RESPONSE BIAS (also called self-selection bias)

• Respondents choose to participate
• Those who respond differ from those who don't
• Overrepresents extreme views

Answer: a) Undercoverage bias (excludes cell phone users and working people) b) Response bias - question wording (leading question) c) Voluntary response bias (self-selection of strong opinions)

Step 1: Identify the paradox 90% think they're better than average Mathematically impossible: Only 50% can be above average So what's going on?

Step 2: Type of bias - Social Desirability Bias

Definition: People answer in ways that make them look good

• Want to present themselves positively
• Don't want to admit flaws
• Especially for socially valued traits

Step 3: Why driving ability triggers this bias

Good driving is socially valued:

• Nobody wants to admit being bad driver
• Being good driver = responsible, skilled, careful
• Admitting you're below average = admitting you're dangerous

Psychological factors:

• Self-serving bias (we view ourselves positively)
• Selective memory (remember our good driving, forget mistakes)
• Different standards (we judge ourselves by intentions, others by actions)

Step 4: How this manifests in surveys

What people think: "I'm a careful, skilled driver" What they say: "Better than average"

Even bad drivers think:

• "I'm careful" (even if slow)
• "Others are reckless" (go too fast)
• "I've never had accident" (been lucky)

Step 5: Other examples of social desirability bias

People overreport:

• Voting ("Did you vote?") - people say yes even if they didn't
• Charity donations - claim to donate more
• Exercise - claim to exercise more
• Healthy eating - claim better diet
• Reading - claim to read more books

People underreport:

• Illegal behavior
• Embarrassing habits
• Socially undesirable opinions
• Income (if seen as bragging)

Step 6: How to reduce social desirability bias

Strategies:

1. Anonymous surveys

   • No judgment possible
   • More honest responses

2. Indirect questioning

   • "How many of your friends..."
   • Less personal threat

3. Randomized response technique

   • Statistical method ensuring privacy
   • Can't identify individual responses

4. Behavioral measures instead of self-report

   • Observe actual behavior
   • Don't rely on what people say

5. Validate against objective data

   • Check survey responses against records
   • Driving: check actual accident rates

Step 7: The driving example specifically

Better measures than self-report:

• Actual accident rates
• Traffic violations
• Driving test scores
• Insurance company data

These would give more accurate picture than survey

Answer: Social desirability bias - people answer in ways that make them look good. Nobody wants to admit being a below-average driver, so people systematically overestimate their abilities. This psychological bias leads to impossible result (90% can't be above average). Common for socially valued traits like driving skill, voting, charity, healthy behavior.

Step 1: Identify the issue Response rate: 200/5,000 = 4% (very low!) Result: 80% dissatisfied

Step 2: The problem - Nonresponse Bias

Who responds to surveys?

• People with strong opinions
• People who are dissatisfied (more motivated)
• People who care deeply about the issue

Who doesn't respond?

• People who are satisfied (no complaints)
• People who are indifferent
• Busy people
• People who don't check email

Step 3: Why 80% is likely biased upward

Those who responded (200 students):

• Probably have complaints about dining
• Motivated by dissatisfaction
• Not representative of all 5,000

Those who didn't respond (4,800 students):

• Might be satisfied (no reason to complain)
• Might be neutral
• Don't care enough to respond

Result: Sample overrepresents dissatisfied students

Step 4: Cannot conclude 80% of all students dissatisfied

The 80% reflects:

• 80% of the 200 who chose to respond
• NOT 80% of all 5,000 students

True dissatisfaction rate unknown:

• Could be much lower
• Satisfied students less likely to respond
• Voluntary response bias

Step 5: Better survey design

To get accurate result:

1. Use random sample of students
2. Follow up with non-respondents
3. Offer incentives for participation
4. Make survey easy and quick
5. Use multiple contact methods
6. Aim for high response rate (>60-70%)

Step 6: Calculate scenario

Possible reality:

• 200 respondents: 160 dissatisfied (80%)
• 4,800 non-respondents: 960 dissatisfied (20%)
• Total: 1,120 / 5,000 = 22.4% actually dissatisfied

The 80% would be very misleading!

Answer: NO, cannot conclude 80% of all students are dissatisfied. Only 4% responded (200/5,000), creating severe nonresponse bias. Dissatisfied students are more motivated to respond, so the 80% likely overestimates true dissatisfaction. The 80% applies only to those who chose to respond, not to all students.

Step 1: Analyze Question A

"Should taxes be increased to fund essential public services like schools and hospitals?"

Bias: Toward YES (supporting tax increase)

Why it's biased:

• Uses positive framing: "essential public services"
• Mentions sympathetic examples: "schools and hospitals"
• Implies taxes are necessary for good things
• No mention of downsides

How it influences:

• People don't want to oppose schools and hospitals
• Feels wrong to say no to "essential" services
• Guilt/social pressure to agree

Expected result: Overestimates support for tax increase

Step 2: Analyze Question B

"Should the government take more of your hard-earned money in taxes?"

Bias: Toward NO (opposing tax increase)

Why it's biased:

• Uses negative framing: "take" (implies theft)
• Emotional language: "your hard-earned money"
• Suggests government is taking what's yours
• No mention of benefits

How it influences:

• People resist having money "taken"
• "Hard-earned" makes it personal
• Government sounds greedy/unfair

Expected result: Overestimates opposition to tax increase

Step 3: Compare the two

Same policy question, opposite biases:

• Question A: Likely 60-70% support
• Question B: Likely 30-40% support
• Same people, different wording!

This shows power of question wording

Step 4: Neutral wording options

Option 1 (Simple): "Do you support or oppose increasing taxes?"

Pro: Very neutral Con: Might not give enough context

Option 2 (Balanced): "Do you support or oppose increasing taxes? Revenue would fund public services, but your take-home pay would decrease."

Pro: Mentions both sides Con: Which to mention first?

Option 3 (Best): "Do you support increasing taxes, oppose increasing taxes, or are you unsure?"

Pro: Neutral language, includes middle option Allows "no opinion" response

Option 4 (Even better - two questions): Q1: "What is your opinion on the current tax level: too high, about right, or too low?" Q2: "If taxes changed, which public services would you prioritize/cut?"

Separates questions, avoids loaded language

Step 5: General principles for neutral questions

DO: ✓ Use neutral language ✓ Avoid emotional words ✓ Present both sides if context needed ✓ Allow "unsure" option ✓ Keep it simple and clear

DON'T: ✗ Use loaded words ("take," "hard-earned," "essential") ✗ Suggest a correct answer ✗ Use only positive or negative framing ✗ Make assumptions ✗ Use double-barreled questions

Answer: Question A biases toward YES (positive framing: "essential services," "schools and hospitals"). Question B biases toward NO (negative framing: "take your hard-earned money"). Both are leading questions that will produce different results. Neutral wording: "Do you support or oppose increasing taxes?" - simple, balanced, no emotional language.

Step 1: Define undercoverage bias

Undercoverage: When some groups in the population have NO CHANCE or LOWER CHANCE of being in the sample

Result: Sample systematically unrepresentative Missing perspectives from excluded groups

Step 2: Example 1 - Literary Digest Poll 1936

Historical disaster:

• Magazine surveyed people from phone books and car registration lists
• Predicted Landon would beat Roosevelt for president
• Roosevelt won in landslide!

What went wrong:

• 1936: Only wealthy people had phones and cars
• Survey undercovered poor and middle-class voters
• These groups voted differently than wealthy
• Sample was systematically biased

Impact: Completely wrong prediction

Step 3: Example 2 - Online surveys about internet usage

Survey: "Complete this online survey about your internet habits"

Problem:

• Must have internet access to take survey
• Undercoverage: People without internet excluded
• These people have different internet habits (none!)

Impact on results:

• Overestimates internet usage
• Misses perspectives of disconnected populations
• Can't generalize to whole population

Who's excluded:

• Elderly without computers
• Poor without internet access
• Rural areas with limited connectivity

Step 4: Example 3 - Workplace satisfaction survey (business hours only)

Survey: Call employees during 9am-5pm

Problem:

• Misses night shift workers
• Misses part-time workers
• Misses field workers who aren't at desk

Impact:

• Night shift may have different satisfaction
• Part-timers may have different concerns
• Office-only perspective

Result: Biased view of employee satisfaction

Step 5: How undercoverage affects results

Systematically excludes groups → Biased estimates

Example effects:

• Missing young people → overestimate conservative views
• Missing poor people → underestimate financial struggles
• Missing minorities → miss diverse perspectives
• Missing rural people → urban-biased results

The direction of bias depends on how excluded groups differ

Step 6: Preventing undercoverage

Strategies:

1. Use comprehensive sampling frame

   • Lists that include whole population
   • Multiple lists if needed

2. Use multiple contact methods

   • Phone, email, mail, in-person
   • Reach different groups different ways

3. Stratified sampling

   • Ensure all subgroups included
   • Sample from each stratum

4. Adjust for known undercoverage

   • Weight responses to match population
   • Statistical correction (imperfect)

5. Know your sampling frame limitations

   • Be aware who's excluded
   • State limitations in conclusions

Answer: Undercoverage occurs when some population groups have no/low chance of selection. Examples: (1) 1936 Literary Digest used phone books, excluding poor voters who voted differently; (2) Online surveys exclude those without internet; (3) Daytime-only calls miss night shift workers. Effect: Sample systematically unrepresentative, leading to biased estimates that don't reflect the full population.