Octopus cyanea, the day octopus, is short lived and grows very quickly. Rearing studies established that an animal passes from settlement to maturity and death in about a year (Van Heukelem, 1973), and age readings from wild animals confirmed a brief life span with rapid, variable growth (Herwig et al., 2012). An octopus left undisturbed for two or three months can therefore gain substantially in weight. In south-west Madagascar the species spawns through the year with seasonal peaks, and females mature at a large size (Raberinary and Benbow, 2012), so the timing of a closure relative to recruitment and spawning affects what it can achieve.

Benbow et al. (2014) examined the early closures and concluded that benefits were clearer when villages closed their sites at the same time, since a lone closure attracted fishers from the whole area on opening day. Oliver et al. (2015) analysed landings records for a long series of closures, compared the month after opening with a baseline period before closing, and used unclosed sites as controls. They reported increases in landings and in catch per fisher-day at closure sites that were not matched at the controls, and estimated that villages recovered more in income than they had given up. These studies depended on a landings monitoring system that had been running for years, which most communities lack.

Research on periodically harvested closures in the Pacific gives reason for care. Such closures can yield a pulse of catch, but the gain may be exhausted quickly if harvests at opening are intense, and the long-term result depends on the balance between the length of closure and the pressure applied afterwards (Cohen and Foale, 2013; Goetze et al., 2018). An evaluation must therefore look beyond the first day.

An impact is the difference between what happened and what would have happened otherwise, and the second quantity is never observed (Ferraro, 2009). The evaluation should begin by writing down the question in that form. For a village the relevant question is usually whether total octopus income over the whole period, including the closed months, was higher than it would have been with the ground left open. A secondary question is whether catch rates at the site stayed above their earlier level for more than a few tides after opening.

Closures differ in ways that determine their effect: the share of the village ground closed, the duration, the season, whether neighbouring villages closed at the same time, and whether the rules were kept. Poaching during the closed period and the number of outside fishers present on opening day should be recorded as carefully as catch. Institutional detail belongs here too, since the durability of community rules depends on clear boundaries, monitoring by the users and graduated sanctions (Ostrom, 1990).

Without a comparison, a rise in catch after opening cannot be separated from a good recruitment year or a seasonal upturn. The framework adopts the logic of before-after-control-impact designs developed for environmental assessment (Stewart-Oaten et al., 1986). Because any single control site may differ from the closure site by chance, several comparison sites are preferable to one (Underwood, 1992). They should have similar reef-flat habitat, be fished by similar gear and sell to the same buyers, and they should lie far enough away that fishers displaced by the closure do not simply move onto them, which would depress their catch rates and exaggerate the apparent benefit.

Data are collected at the landing point or the buying station by a trained resident. For each fishing trip the record holds the date, the fisher, the gear or method, the ground fished, the time spent, and the number and total weight of octopus; individual weights and sex are taken from a subsample. Recording starts at least two to three months before the closure, continues at comparison sites throughout, and runs for at least three months after opening. Days with no fishing are entered as such, since effort is half of the calculation. Buyers' purchase books provide an independent check on totals.

Raw catch per fisher-day is affected by much besides abundance. Gleaning is possible only on spring low tides, divers and gleaners differ in catching power, and experienced fishers catch more. Standard practice in fisheries science is to model catch rate as a function of such factors and to read the abundance signal from the remaining time effect (Maunder and Punt, 2004). For closure data a generalised linear mixed model with terms for tidal phase, month, method and site, and a random effect for the individual fisher, is a suitable starting point. Mean individual weight should be analysed separately from numbers caught, because a closure of a few months is expected to act mainly through growth.

The effect of the closure is estimated as the change in standardised catch rate at the closed site from before to after, less the corresponding change at the comparison sites. Three further calculations turn this into something a village can use. Total landings and value in the weeks after opening are compared with the baseline. The catch forgone during the closed months, estimated from baseline rates and from effort displaced to other grounds, is subtracted. And the decay of catch rates after opening is plotted, to show whether the gain lasted beyond the first tides.

Distribution deserves equal attention. Opening days favour those who can be present and who fish fastest, and divers may take a large share before gleaners reach the ground. The landings record allows the gain to be broken down by method, by sex of the fisher and by resident or visitor. A closure that raises the village total while shifting catch from women gleaners to visiting divers has not succeeded on the terms most communities would set.