Most-Different Systems Design¶

Comparison design — instantiates Structured Comparative Case Design

Compares cases that differ in almost every way yet share the same outcome, so the one condition they all hold in common becomes the candidate cause.

Most-Different Systems Design (MDSD) is the mirror image of its sibling. It gathers cases that are as unlike one another as possible — different sizes, settings, histories, cultures — but that nonetheless converged on the same outcome. The logic is again subtractive, but run in reverse: every condition on which the cases differ is demoted, because it cannot explain an outcome they all share; the cause must lie among the few conditions they hold in common despite their diversity. It is the case-based expression of Mill's method of agreement. What makes it THIS design is the inverted bet — difference is the control, commonality is the suspect — and the payoff it chases: a factor that survives across wildly different contexts is a robust one.

Example¶

Four countries cut road-traffic deaths sharply over a decade. They could hardly be more different: a wealthy Nordic state and a middle-income Southeast-Asian one, a sprawling federation and a dense city-state, with different road cultures, vehicle fleets, and legal systems. An unstructured look would drown in the contrasts. A most-different comparison ignores them on purpose and asks what these dissimilar successes share. Nearly nothing — except that each rolled out systematic automated speed enforcement paired with lowered urban limits, in the spirit of a Vision Zero strategy. Because that factor is almost the only thing four otherwise-unalike countries hold in common, it becomes the candidate cause, and a strong one precisely because it held across such different terrain. The design cannot prove it single-handed, but it has found a factor worth tracing inside each case.

How it works¶

The core artifact is a commonality matrix: conditions down the rows, the diverse cases across the columns, most cells marked differing and struck out as controls, the search fixed on the rare rows marked shared. The design succeeds to the degree it can drive the shared set toward one. Its leverage comes from diversity rather than similarity: the more different the cases, the more impressive — and the more generalizable — a surviving commonality becomes, because rival explanations tied to any one context are eliminated by the cases that lack that context.

Tuning parameters¶

Diversity target — how unalike the cases must be. Greater diversity strengthens any surviving commonality but makes measurement equivalence harder to guarantee.
Outcome strictness — how identical the shared outcome must be to count as "the same." Loosening it admits more cases but blurs what is being explained.
Commonality threshold — whether you demand a single shared condition or accept a small shared cluster carried forward.
Rival coverage — how deliberately you seed the case set with cases that lack each competing explanation, so diversity actually eliminates rivals rather than merely decorating the table.

When it helps, and when it misleads¶

Its strength is reach: a factor that persists across cases sharing almost nothing else is a strong bet for a real, portable cause, and MDSD is the design that manufactures that kind of evidence. Its central failure mode is that "maximally different" cases are never fully different — they usually share more than one thing, so the method can indict a shared factor that is merely a common correlate rather than the cause.^[1] It is also weak against equifinality, where different causes produce the same outcome in different cases. The classic misuse is assembling diverse cases that happen to share your favoured factor and presenting the agreement as discovery. The discipline that guards against it is to choose cases that genuinely vary on the rival explanations, and to confirm the surviving commonality by tracing it inside each case.

How it implements the components¶

comparison_logic_choice — commits the study to the method-of-agreement branch: control by difference, explain by residual commonality.
most_different_commonality_matrix — its central product, the tabulation that strikes out the many differing conditions and isolates the rare shared one.

It does not build the mirror most_similar_difference_matrix — that is Most-Similar Systems Design; it does not verify that a variable means the same thing across such varied cases (that is Measurement Equivalence Audit); and it does not itself bound how far the finding generalizes — that is Sensitivity to Case-Set Analysis.

Instantiates: Structured Comparative Case Design — MDSD is the second of the two contrast logics the design can adopt.
Consumes: Case Universe Sampling Frame supplies the bounded pool from which maximally diverse cases are drawn.
Sibling mechanisms: Most-Similar Systems Design · Measurement Equivalence Audit · Matched Case Pairing Protocol · Within-Case Process Tracing · Deviant Case Follow-Up Protocol · Replication Case Sampling Cycle · Sensitivity to Case-Set Analysis · Rival Explanation Elimination Table · Case Selection Bias Audit · Case Universe Sampling Frame · Comparative Case Review Panel · Comparative Historical Timeline · Configurational Comparison Truth Table · Counterfactual Contrast Memo · Cross-Case Evidence Matrix Tool

Notes¶

Because MDSD deliberately maximizes diversity, it is the design most exposed to measurement non-equivalence: a variable that quietly means different things in a rich country and a poor one can manufacture a false commonality. That exposure is exactly why it should be paired with Measurement Equivalence Audit before any shared row is trusted.

References¶

[1] Mill's method of agreement — if instances sharing an outcome have only one antecedent in common, that antecedent is the cause. Mill himself noted it is weaker than the method of difference, because two "different" instances may share several antecedents, only one of which is causal. ↩