Contribution
Nine in ten constitutions carry emergency provisions, but whether governments keep to them had not been measured across countries. The paper builds the first cross-national list of states of emergency (SOEs) coded for lawfulness, splits unlawfulness into declaration, conduct during the emergency and prolongation, and asks what goes with it. It identifies “115 out of 853 analyzed SOEs as unconstitutional” (p. 3).
Key insight
The cause matters more than the constitution. Natural disasters are exogenous; political turmoil is “highly endogenous to government behavior” (p. 8) and threatens the government’s survival, so an executive has more reason to overstep after turmoil. Nearly all unlawful emergencies follow turmoil, and the authors read that as executives acting “not to further some common good but, rather, to safeguard and possibly extend their own power” (p. 25).
Assumptions
- Universe. Hafner-Burton et al. (2011), updated to 2017, restricted to national-level SOEs that could be confirmed: 853 cases, 1949–2017 (pp. 12–13).
- Coding of unlawfulness (pp. 13–14). Factiva and Lexis/Nexis searches for “state of emergency” with “unlawful”; Freedom House and other NGO reports; and SOE length against the constitutional maximum, coded unlawful when no lawful extension was found. Evaluations by courts, law professors, NGOs and journalists all count, so the cases “could also be called “suspected to be unlawful.”” (p. 14). Only three of the 115 breach statute rather than the constitution (p. 13, fn. 8).
- Under-reporting acknowledged (pp. 14–16). English-language, rich-world press; press freedom; digitization over time (“the vast majority of unlawful states of emergency took place post 1990”, read as an artefact); long-standing SOEs in Jordan, Syria and Zimbabwe with no report of unlawfulness.
- Regime type from Cheibub et al. as updated (p. 17); emergency constitution from the authors’ Index of Emergency Powers, cost and benefit components (p. 19).
- Estimation. Simple logits, because “there is amazingly little systematic structure in the data” (p. 20); endogeneity of turmoil conceded, and a selection correction judged infeasible (pp. 20–21).
Key results
- Table 1 (p. 15). Democracies: 271 SOEs after natural disasters (8 unlawful), 241 after domestic turmoil (41). Autocracies: 68 (6) and 245 (56).
- Phases (Table 2, p. 16; p. 22). Declaration 19, conduct 65, prolongation 37 in Table 2 and 39 in the text, overlapping. 97 unlawful SOEs followed turmoil, 14 natural disasters, five man-made events (p. 22).
- Duration (p. 22). Of lawful SOEs, 63 per cent ended exactly at their maximum, 15 per cent earlier, 22 per cent were lawfully extended. “Lawful SOEs last 115 days on average while the average duration of unlawful SOEs is 321 days” (p < .01).
- Full sample (Table 4, p. 23). Democracy −0.849 (s.e. 0.228), log GDP per capita −0.275; presidential 0.547 and bicameral 0.398 (both p < .10), second-chamber approval 0.593 (p < .05). Pseudo-R² 0.047–0.070.
- Domestic events only (Table 5). Democracy −0.52 to −0.92; bicameral 0.532, Cost INEP 2.688 and second-chamber approval 0.700 significant; investment price −0.479. Recession −0.409 and the conditions index 0.277 are not significant in the logits. Pseudo-R² 0.013–0.039.
- Democracies only (Table 6). Government price level −1.184, executive respect for the constitution −0.622, Benefit INEP −1.801 (p < .10) and second-chamber approval 0.724; nationalist/populist 0.708 (s.e. 0.688), not significant.
- Simple differences (Table A1). Among domestic events, recession years show 30 per cent unlawful against 10 (p < .01), and constitutions with more grounds for emergency 25 against 16 (p < .02).
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | About one in seven national SOEs, 1949–2017, was reported as unlawful (115 of 853), probably an undercount | moderate: coding by press and NGO reports | pp. 13–16 |
| C2 | Unlawful SOEs follow political turmoil far more often than natural disasters | strong as a description | Table 1; p. 22 |
| C3 | Unlawful SOEs last much longer: 321 days against 115 on average | moderate | p. 22; partly built into the coding of prolongation |
| C4 | Democracies are less likely than autocracies to run an unlawful SOE | moderate | Tables 4–5 |
| C5 | Bicameral systems and second-chamber approval go with more unlawful SOEs | weak to moderate | Tables 4–6; p < .10 to < .01 |
| C6 | Recession and more grounds for emergency raise the risk after domestic events | weak: simple differences only | Table A1; not significant in Table 5 |
| C7 | Nationalist or populist executives in democracies are far more likely to run one | not supported by the tables | Table 6 coefficient insignificant; p. 25 |
| C8 | Executives who overstep do so to safeguard their power, not for the common good | weak: an inference from C2 | p. 25 |
| C9 | Emergency measures sometimes survive the SOE’s end | conjecture; no data | p. 6 fn. 2; p. 25 |
Concepts
- De jure/de facto gap: government behaviour under an SOE that departs from the constitution’s emergency provisions (abstract).
- Three phases (pp. 5–6): unlawful declaration (no event, a ground not in the constitution, the wrong actor, formal conditions unmet); unlawful conduct during the SOE; unlawful prolongation. A fourth, measures that outlive the SOE, is named and not measured: “we have no data on this aspect of unlawful emergency behavior” (p. 6, fn. 2).
- Index of Emergency Powers (INEP): cost components (who declares, who approves, how many grounds) and benefit components (power to dissolve parliament, suspend rights, expropriate, censor), each 0–1 (p. 19).
Connections
It builds on the authors’ earlier work: harder declaration means fewer SOEs (2018a); more emergency benefits go with more disaster deaths (2021b); SOEs after terrorist attacks go with more attacks (2020). It cites Lührmann and Rooney (2020), by the author of LIT-1485, for the finding that governments declaring an SOE move towards autocracy more often (pp. 4, 7, 25).
Bearing on the record
- CLAIM-445 (protective regimes become self-maintaining). Cited for
the gap between emergency law and emergency practice. The reading
supports one strand of the claim, and less of it than the citation
implies.
- Supports, the strand of prolongation. Unlawful emergencies last nearly three times as long, and 66 per cent of unlawful SOEs run past their constitutional expiry against 22 per cent of lawful ones lawfully extended (p. 22). Emergencies that escape their legal limits persist. The authors’ hypothesis 5, that an executive prolongs an SOE to postpone review of what it did during it (p. 11), is the claim’s epistemic and political feedbacks in one: the emergency defers the judgement of itself. It was not tested.
- Supports the claim’s own caution. Nothing here measures whether the threat had declined. Natural disasters, where the threat is exogenous, rarely yield unlawful SOEs; political turmoil, partly the government’s own doing, usually does. That is a correlation between endogenous threat and unlawfulness, not the residue the claim’s test asks for.
- Does not support the persistence of emergency measures beyond the emergency, the regime becoming self-maintaining in the full sense. The authors “suspect that at least in some cases the measures implemented during an SOE survive its end and become standard practice. To ascertain whether this really is the case, more fine-grained data are needed” (p. 25).
- The word. The paper uses “unlawful” and “unconstitutional” interchangeably (p. 13, fn. 8); the exchange’s “unconstitutional” is the paper’s title. Either is short for “suspected to be unlawful”.
- Filed as THEORY-911.
- CLAIM-446. Only by analogy: the prolongation hypothesis has the shape of a defence that defers the evidence against it.
- No instruction for machine-learning practice; nothing for the anthology.
Limitations
- A working paper, not the version of record. The JLS article may differ; only the abstract was compared.
- Coding by report. Unlawfulness is what the press and NGOs said, or a duration past the constitutional maximum with no extension found; the authors call it subjective and expect undercounting.
- Correlational and weak. Pseudo-R² at most 0.07; the authors say they “cannot claim that our findings are fully causal” (p. 3).
- Inconsistencies.
- The odds ratio from Table 4’s democracy coefficient (−0.849) is about 0.43, not “about 75 %” (p. 23).
- The conclusion’s nationalist or populist executives “far more likely” (p. 25) rests on an insignificant coefficient (0.708, s.e. 0.688).
- Recession and the number of emergency grounds, in the abstract, are significant only as simple differences (Table A1), not in the logits.
- Prolongation cases are 37 in Table 2 and 39 in the text.
- “exactly two thirds of all unlawful states of emergency remain in place too long” (p. 25) is the 66 per cent past expiry, not the 34 per cent coded unlawful for prolongation.
Open questions
- Do measures adopted under an SOE survive it? The authors’ fourth phase, unmeasured.
- Is prolongation driven by fear of ex post review (hypothesis 5)? A comparison of SOEs with and without independent ex post judicial review of emergency conduct would test it.
- Does the JLS version repair the inconsistencies?