Lecture_3 Flashcards
(20 cards)
What Exactly is Descriptive Research?
- Research that simply describes, but does not directly link outcomes to particular causes
- Information is typically useful although not causal
- Example: “95 % of online consumers are satisfied with their online shopping experiences”
- Even when linking different variables, the research remains descriptive (although the statistical
methods are not anymore), e.g., “Men earn more money than women”, “conservatives less
intelligent than liberals” …. - Note: descriptive research design and descriptive statistics are two different things
- The majority of empirical data comes from descriptive studies
The Difference Between Descriptive Research Designs and Statistics
Descriptive research focuses on summarizing sample characteristics without establishing causal relationships, whereas causal research aims to determine cause-and-effect relationships between variables.
Why Use Descriptive Research?
- To describe the characteristics of relevant groups, such as consumers, salespeople, organizations, or market areas.
- To estimate the percentage of units in a specified population exhibiting a certain behavior.
- To determine the perceptions of product characteristics, firm regulations, state laws, etc.
- To understand the attitudes or the behavior of people, students, employees.
- To determine the degree to which variables are associated.
- To make specific predictions.
Example for a Descriptive Study
Zurich ranks second place in Quality of Living City Ranking 2023
Important Sources for Descriptive Data
Secondary
Primary
External
Internal
Secondary
External
Internal
External: Census data
Publicly available statistics
Published studies
Journals and newspapers
Industry association reports
Panel data
Data from external social media
Crowdsourcing platform …
Internal: Internal statistics
Financial accounting
Cost accounting
Sales data
Customer database
Clickstream data
Data from firm-hosted
community …
Primary (focus of this course)
External
Internal
External: Surveys to customers,
shareholders, partners,
the general public …
Internal: Employee surveys
Salesmen surveys …
Cross-Sectional Designs
Single Cross-Sectional Designs
* Sample: One distinct group of respondents
* Data Collection: Occurs once from this group
* Purpose: Offers a snapshot of a specific group at a particular point in time
* Example: Surveying employees’ job satisfaction in a company in 2023
Multiple Cross-Sectional Designs
* Sample: Several distinct groups of respondents
* Data Collection: Occurs once from each group, often at different times
* Purpose: Compare and contrast different groups at different times without repeated measures on the same group. It’s like taking
multiple snapshots
* Example: Surveying employees’ job satisfaction in the same company in 2023, 2025, and 2027 using different employee samples each time
Commonality: Both designs give insights into a specific time point, without tracking changes in specific individuals over time
Cohorts & Longitudinal Designs
Definitions
* Cohort: A group experiencing a shared event or characteristic in a specific timeframe
* Example: Individuals who entered the workforce in 2020
* Longitudinal Design: Research methodology collecting data on the same subjects repeatedly over time
* Example: Surveying a group of 100 people in 2020 about their job satisfaction, then re-surveying the same group in 2022, 2024, and 2026 to track changes
Relationship
* Cohort Analysis are a type of longitudinal study
* While cohort specifies whom you’re studying, longitudinal describes how you’re studying them
Panel Designs
A panel is…
* … a survey of individuals, households, companies etc. to obtain data on a single subject at regular
intervals over a longer period, using the same sample and carried out using the same methods each
time.
* … most important in FMCG (=Fast-Moving Consumer Goods) industries:
* Household panels
* Retail panels
Panel Designs – Problems of Panels
Panel mortality
Selection effects
Panel (participation) effects
Information collected is predetermined
Designing a Questionnaire
- Introduce the study: inform participants about who is conducting the study, the purpose of the study
and alleviate any concerns potential respondents might have (time, confidentiality, “correctness” of
answers, etc.) - Introductory & screening questions: easy-to-answer (often factual) questions related to the main
subject; often screening questions for sample quotation (disqualifying certain respondents from
participating) - Sensitive & related questions: questions about sensitive issues, personal opinions and attitudes,
personality related questions; often blocked into question categories, e.g., demographic data. - End the study: thank the participant for their participation, leave contact information for further
inquires
Research Example: Social Desirability
Study on Privacy Concerns and Loyalty Programs
The Bradley Effect
- Definition: An observed discrepancy between voter opinion polls and election outcomes for African-American candidates
- Origin: 1982 California governor’s race - Tom Bradley, an African-American, was ahead in polls but lost the actual election
- Possible Explanation: White voters may say they support a black candidate due to social desirability, but vote differently in private
- Broader Implication: Reflects challenges in accurately gauging public opinion on sensitive or racially-charged matters
- Skepticism: Some researchers question the validity of the Bradley Effect, suggesting other factors might explain discrepancies
6 rules to help you develop better questions:
Rule 1. Avoid complexity. Use simple, audience-specific language if possible.
Rule 2. Avoid leading and loaded questions. Use neutral questions.
Rule 3. Avoid ambiguity. Be as specific and precise as possible.
Rule 4. Avoid double-barreled questions. Ask about one topic at a time.
Rule 5. Avoid making assumptions. Ask, don’t assume.
Rule 6. Avoid burdensome questions. Use ‘top-of-mind questions’.
Considering the Question Placement (Reihenfolge)
- Apply question sequencing: First easy, non-private, and non-confidential questions and then difficult,
private, and confidential ones to reduce non-response, and bias. - Be aware of question interdependencies.
How to Increase Response Rates
- Multiple contacts (mail & phone)
- Why important (strong appeals)
- Credible sponsor or affiliation
- Anonymity, confidentiality
- Personalization
- Incentives (monetary or nonmonetary)
- Survey length & design
Beware of a possible non-response bias!
Nonresponse Bias
Definition: Nonresponse is the absence of a reply from certain survey participants, leading to potential data distortion
Challenges
* Nonresponse effects on data are often ambiguous
* Difficult to predict when nonresponse is biased
Studies Highlight
* Keeter et al. (2006) & Kohut et al. (2012): Low response can be as accurate as high response
* Groves (2006): Emphasizes caution in interpreting results
Strategies to Address
* Analyze how nonrespondents differ from respondents
* Use external data sources for comparison
Credibility
* High response = Reduced potential for bias
* Low response = Increased scrutiny and potential credibility issues
AI, Bots, and Reliable Surveys
Bots and Survey Responses
* Problem: Automated bots can flood a survey with fabricated responses
* Implication: Data becomes unreliable, and results may be skewed
* Solution: Use CAPTCHAs, respondent authentication, or IP filtering
AI-generated Text
* Problem: Advanced AI can generate human-like responses, potentially filling out surveys inauthentically
* Implication: Distorted insights and conclusions
* Solution: Ensure unique respondent verification and track response patterns for inconsistencies
Sampling Bias with Online Platforms
* Problem: Users on platforms such as MTurk might not be representative of the broader population
* Implication: Results may not be generalizable
* Solution: Use a mix of platforms and methods to gather a more diverse sample
Summing Up
- Descriptive research can help you describe a population.
- When collecting data from a representative sample, you can make inferences for the whole population and generalize your findings.
- Questions are in a standardized format and careful consideration should be taken to word them properly (avoid misinterpretations).
- Always pretest your survey to ensure it works correctly and apply measures to increase response rates (+ decrease response bias).
