Fundamental Paradox of Social Science

The Core Principle

The more specific and localized your investigation, the broader and more generalizable your claims can be.

Deep engagement with a particular case, context, or phenomenon often yields insights that travel further than broad surveys or aggregated data.

The Paradox

Common intuition:

  • Large sample, many cases → generalizable
  • Small sample, one case → not generalizable

But often:

  • Large sample → superficial patterns, context-stripped, atheoretical
  • Deep case → reveals mechanisms, identifies scope conditions, theory-building

Why It Works

Mechanisms Over Patterns

Specific investigation reveals how things work, not just that they correlate. Mechanisms generalize better than correlations.

Context as Data

Close study shows which features of context matter. This tells you where else the finding applies (scope conditions).

Theory Development

You can’t build theory from aggregated data alone. You need thick description, process tracing, understanding of local meaning. Then generalize the theory.

Rich Variation

A single case over time, or one setting with internal variation, can reveal more causal structure than many cases measured once.

Examples

Ethnography

Deep ethnography in one community reveals mechanisms of social trust. The mechanisms (reciprocity, reputation, sanctioning) apply broadly, even if specific practices don’t.

Case Studies

Study of one organization’s innovation process reveals general principles about how new ideas diffuse, get adopted, or fail—more than survey of 1000 organizations asking “did you innovate?”

Clinical Research

Detailed study of a few patients with rare mutations reveals fundamental biology that explains common diseases.

Experimental Iteration

Running many conditions with small N in each (deep exploration of parameter space) beats one condition with large N (just testing if an effect exists).

Application to Research

Study Design

Don’t default to “bigger sample = better.” Ask: Do I need breadth (establish prevalence) or depth (understand mechanism)?

When to Go Deep

  • Building theory (need to see process, not just outcome)
  • Identifying mechanisms (need fine-grained observation)
  • Finding scope conditions (need to vary context systematically)
  • Explaining anomalies (need thick description)

When to Go Broad

  • Estimating prevalence
  • Testing scope of established theory
  • Finding rare cases
  • Demonstrating robustness across contexts

Combining Both

Ideal: Deep case study → develop theory → broad test → refine with more cases. Iterate.

Limitations

Not a License for Small N

The paradox doesn’t mean “anything goes with one case.” You need:

  • Theoretical framework to interpret the case
  • Clear reasoning about what generalizes (mechanisms) vs. what doesn’t (particulars)
  • Systematic within-case analysis, not cherry-picking

Generalization Still Requires Argument

You must explain why this case reveals something general. What features are shared? What mechanisms travel?

Risk of Overextension

Easy to mistake local particulars for universal mechanisms. Need external validity checks.

vs. Inventor’s Paradox:

  • Inventor’s: Solving general problems helps with specific solutions
  • Social Science: Studying specific cases helps make general claims
  • Different phases: problem-solving vs. inference

vs. Case Study Method: This is the justification for case study methods, but it’s broader—applies to experimental depth, ethnography, clinical research, any intensive study.

Connection to My Work

This principle shapes:

  • Research design: Prefer iterative deep exploration over single large-N study
  • Modeling: Fit models deeply to one context to understand mechanisms, then test breadth
  • Data collection: Rich measurement in fewer cases rather than sparse in many
  • Theory building: Use cases to develop theory, not just test pre-specified hypotheses

Examples:

  • Rather than survey many trilingual speakers once, track fewer speakers intensively across contexts
  • Rather than one big dataset, multiple smaller datasets with deeper measurement
  • Rather than test one hypothesis broadly, explore mechanism space deeply

Key Sources

  • Flyvbjerg, B. (2006). “Five Misunderstandings About Case-Study Research”
  • Ragin, C. (1987). The Comparative Method
  • Mitchell, J. C. (1983). “Case and Situation Analysis”
  • Geertz, C. (1973). “Thick Description”
  • Gerring, J. (2004). “What Is a Case Study and What Is It Good For?”