Contribution
This is the first test of the MAC-Q (LIT-496) by people outside the programme, and the first outside an English-speaking WEIRD sample. It does three things:
- tests the seven-factor structure against single-factor, two-factor and second-order alternatives, by CFA and ESEM;
- relates the seven subscales to two dimensions of political conservatism and to prosocial intentions;
- tests, in two experiments, whether priming resource scarcity lowers endorsement of the seven.
Key insight
The MAC-Q’s structure replicates where the original found it, on the Relevance items, and fails where the original found it weak. The authors’ practical conclusion is to use the Relevance subscale alone.
Assumptions
- Translation. A Turkish translation of the MAC-Q is assumed to preserve the item meanings. The translation procedure is not described in this manuscript.
- Samples.
- Study 1: undergraduates (n = 445, 369 women).
- Study 2: a convenience community sample collected by research assistants from friends and family (n = 576). It used a 5-point scale instead of the 0–100 slider (Table 1 note).
- Study 3: US MTurk (n = 921).
- The scarcity hypothesis. Scarcity lowers endorsement of all seven. This is the authors’ derivation via Roux et al. (2015), not a prediction stated in MAC’s own papers.
Key results
- Structure (Study 1, Table 2).
- Relevance, seven factors: χ²(168) = 378.4, CFI .949, TLI .936, RMSEA .053, SRMR .046.
- Relevance, second-order single factor: CFI .906, RMSEA .069.
- Relevance, second-order two factors: CFI .937.
- Relevance, first-order single factor: CFI .682.
- Judgement, seven factors: CFI .828, RMSEA .082.
- Full form: CFI .752.
- ESEM did not improve on CFA.
- The MFQ fits poorly in every version, “in line with previous work”.
- Latent correlations (Table 3).
- MAC-Q: fairness–property .874, reciprocity–fairness .808, family–group .800.
- MFQ: Authority–Sanctity .971.
- Relevance alphas .69–.89 in Study 1. In Study 2 they fell to .58–.68. Judgement alphas run as low as .38 for property.
- Ideology (Study 1, Table 4).
- Resistance to change correlates with family (.28), group (.20), heroism (.21) and deference (.44).
- Opposition to equality correlates with deference (.22), and negatively with fairness (−.18), property (−.15) and reciprocity (−.10).
- In Studies 2 and 3, right-wing or socially conservative orientation predicts a higher total score.
- Prosociality.
- In Study 1 the subscales jointly explain 5.1% of prosocial intentions. Only group is a unique predictor, and it is negative (β = −.17).
- In Studies 2 and 3, zero-order correlations in the whole sample are .10–.20 and .09–.34.
- Scarcity. No effect on prosociality or on any MAC-Q subscale, in Turkey (MANOVA p = .36) or in the preregistered US study (p = .30). Study 3 had no manipulation check, which the authors concede.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | The MAC-Q Relevance subscale has a valid seven-factor structure in Turkey | moderate | good CFA fit in one student sample; full form and Judgement fail |
| C2 | Only the Relevance subscale should be used | moderate | fit results here and in the original |
| C3 | MAC-Q relates to ideology as MAC and Sinn & Hayes predict | weak–moderate | correlations are mostly as expected; property’s negative tie to opposition to equality was not, and is explained post hoc |
| C4 | The results “supported the original proposition of MAC that all moral foundations are related to cooperation” | weak | rests on correlations with self-reported prosocial intentions of .1–.3, in which group is a negative unique predictor in Study 1 |
| C5 | Resource scarcity does not shift endorsement of the seven | weak (a null) | two null experiments; one manipulation check points the wrong way and the other study has none |
Concepts
- ESEM. Exploratory structural equation modelling: a CFA that allows cross-loadings, here approximated to zero by target rotation.
- Resistance to change / opposition to equality. The two components of conservatism in Jost et al.’s motivated-social-cognition account, measured with Sarıbay et al.’s Turkish scale.
Connections
- The MAC-Q (LIT-496). This is its independent replication. It matches the original’s Relevance and Judgement split, and it also shows the seven to be less distinct than the original’s framing suggests: a single higher-order factor fits acceptably, and three latent correlations are .80 or more.
- The sixty-society test (LIT-047, NOTE-034). It does not bear on the test’s weak diagnosticity. It has no comparison domain and no non-cooperative item. Its items are the MAC-Q’s, with the same virtue-term wording.
- Purity (LIT-101). Sanctity appears only through the MFQ, with poor reliability (α = .49 on Relevance). Its .97 latent correlation with Authority fits the “binding” reading of MFT, and also any reading on which purity concerns travel with deference. Neither of those is morality-as-cooperation’s account or LIT-101’s.
Bearing on the record
No ML practice, and the anthology does not hold it. For the morality-as-cooperation thread it is the citation for “the MAC-Q’s Relevance structure replicates outside the UK and US”. It is a replication in one student sample, and it is no evidence for MAC’s claim that morality is only about cooperation.
Limitations
- One non-representative sample for the structural test, mostly women undergraduates.
- Self-reported prosocial intentions as the measure of cooperation.
- Manipulation checks. The Study 2 pilot reports that the scarcity group scored lower on financial concern (2.31 against 2.63). It reads this as “perceived more scarcity”, which is backwards unless the scale was reverse-scored, and the text does not say so. Study 3 has no check.
- Translation. The translation procedure is not described in the manuscript read.
Open questions
- Does the seven-factor Relevance structure hold in a representative Turkish sample, and with measurement invariance tested across Turkey and the US?
- Are seven factors better than the second-order two-factor model (CFI .937) on any criterion beyond fit indices?
Corrections
- none to a seeded skim (there was no seed)
- Authors. The owner’s brief gave “Yilmaz, Harma, Doğruyol, Curry et al.”. The authors are Onurcan Yilmaz, Mehmet Harma and Burak Doğruyol only, per both Crossref and the manuscript. Curry is not an author.
- Venue. European Journal of Psychological Assessment 37(2):149–160, March 2021. The preprint was posted 20 October 2020.
- The original’s result as cited here. The paper summarises the original as showing that the full MAC-Q “provided a poor fit to data similar to MFQ” while the Relevance subscale had good validity. That is a fair reading of the original’s tables, and plainer than the original’s own summary.