Demographic variability and scales of agreement and disagreement over resource management restrictions

Conflicts over the imposition of restrictions in common-pool resources management institutions are expected to arise from variations in human values, perceptions of justice, and the disparate demographic scales of benefits and costs. We hypothesized and tested a series of propositions about how the demographic scale and context of common restrictions would influence coral reef fisheries of 4 African countries. We surveyed the preferences and perceived benefits of 1849 people in 89 fish landing sites for 6 common restrictions of increasing severity. Variability in perceived benefits within and between neighboring communities was evaluated to determine how perceptions changed with the severity of the proposed benefit/cost restriction scale, perceived benefits, disparities between beneficiaries, and national context. Within-community variability declined strongly (r2 > 0.90) as perceived benefit increased but was either weak or not significantly associated with the neighbor-community’s variation. Within-community variation was less than between-community variation and differed by nation. There was generally broader scale agreement on the benefits of weaker restrictions of minimum sizes of fish and allowable fishing gear and more disagreement on stronger restrictions on species, time, and space use. Reduced variability was strongly associated with less perceived disparity in the benefits received by local versus government beneficiaries. These findings indicate potential conflicts between neighboring communities for most, but particularly the strongest, restrictions. Consequently, the broadscale management benefits of strong restrictions will need to address between-community compliance and justice procedures. Demographic variability requires coordinating governance and management to account for restriction-specific variability in the perceptions of management benefits.

The between-community aspects of CPRM have been evaluated less frequently, possibly being more challenging to assess and therefore a marginalized area of CPRM research (Cox et al. 2010, Ostrom 2010, Koontz et al. 2015).Therefore, cultural and governance system differences that range from the individual to state are less considered, researched, and critically evaluated (Saunders 2014).Further, managing common-pool resources is difficult when resources are used by either many or sparsely distributed users in areas with unclear boundaries, where rule breakers are difficult to detect and sanction and between-group forums do not exist (Agrawal 2003, Cudney-Bueno andBasurto 2009).Further, the scarcity of resources will influence the emergence and need for conflict resolution as well as the feasibility of the implementing options (Hardin 1968, Odum 1982).
Countries with a strong rule of law having legislation, taxation, enforcement, and government penetration should ensure that rules are agreed on, shared, and enforced by stakeholders (Fukuyama 2011).These conditions are more likely where the values, wealth, and costs/benefits promote compliance and governance effectiveness (Inglehart and Welzel 2005, Hofstede et al. 2010, McClanahan and Rankin 2016).Therefore, social actors, contexts, and coordination from the individual to higher demographic and governance scales are expected to be critical to CPRM success.Critical reviews of CPRM conclude that investigators have not fully considered the influences of demographic heterogeneity, market penetration, economic development, and state policies (Agrawal 2003, Araral 2014).Conflicts and failures in CPRM can include a number of factors but are expected when individuals, communities, and their neighbors lack shared perceptions, forums, leaders, and agreements on restrictions and enforcement (Agrawal et al. 2013, Koontz et al. 2015).When neighbors do not share values, perceptions, and costs/benefits, they can fail to restrict some resource extraction behaviors, which can lead to matching only the most common and minimal restrictions (Agrawal et al. 2013).
Management restrictions differ in their costs and benefits and differentially accrue to professions based on the economic scales of the profession's dependency (Table 1).Resource managers, for example, are likely to see benefits across many economic sectors, i.e., fishing and tourism, and over longer periods of time than resource users, i.e., species preservation and spillover of fish from fisheries closures.Poor recognition of economic development and these differential scales of dependency can therefore lead to cost/ benefit context-policy mismatches and CPRM failures (Araral 2014, Daw et al. 2016, McClanahan and Abunge 2017).For example, because fishing impacts on vagile species are difficult to assess and more likely to negatively accrue beyond the local fishing Ecology and Society 23(4): 33 https://www.ecologyandsociety.org/vol23/iss4/art33/Table 1.Hypotheses of the scale of costs and benefits of six different fisheries management restrictions.Hypothesized gradient of perceived reduction in access to individual resource users for common fisheries management options.Organized along the hypothesized least to most restrictive to the individual, i.e., high to low cost/benefit, from left to right while acknowledging cost/benefit overlaps.restrictions on the capture of wide-ranging species.Managers, on the other hand, will consider the needs of vagile species on scales larger than local fish landing sites.Previous studies have supported this disparity in scale contention and identified a number of potential areas of management conflicts (McClanahan and Abunge 2016).Successful management needs to negotiate positions that integrate national with within-and between-community preferences.This requires understanding the pervasiveness of preferences and also boundaries of potential conflicts.To assist with this understanding, we evaluated differences in within-and between-community preferences for common-pool fisheries management options in four African nations.
We asked if neighboring communities differ in their preferences and if between-community preferences vary with restrictions and national contexts.The common fisheries restrictions examined were minimum size of fish, allowable gear, species restrictions, temporal closures, area-based management, and permanent closures (Table 1).Our previous study found contextual evidence for perceived disparity in benefits of restrictions based on profession (user vs. manager) and country but failed to examine the intercommunity perceptions (Table 2).Although previous findings supported the restriction-gradient assertion in the more developed countries of Kenya and Tanzania, they also uncovered untheorized variations between communities, professions, and nations (McClanahan and Abunge 2016).Resource managers' responses, for example, generally failed to acknowledge the proposed cost-benefit gradient and scaled most restriction benefits high and social disparity low, with some country-specific deviations (Table 3).Consequently, the differences between managers' and resource users' responses increased across the proposed restriction gradient, with some country-specific variation.We concluded that because managers' economic dependence is on a larger spatial and temporal scale, i.e., regional or national governments, they do not directly experience local costs and believe long-term and large-scale benefits will eventually accrue to local scales.Some of the observed differences between users and managers were hypothesized to result from national contexts of wealth, human development, and the histories of protected areas.
The question we addressed is whether perceptions about restrictions are uniform between neighboring communities when examining restriction-specific differences.In principle, restriction costs and benefits should be similar and appreciated broadly within a profession and nation.Perceptions may, however, be influenced by local contexts of shared experiences, isolation, neighborhood competitions, education, forums, benefit sharing, and specific wealth and governance contexts.Consequently, common socioeconomic factors such as residency, education, communications, wealth, age, and perceptions of fairness and justice are expected to influence these perceptions (Schroeder et al. 2003, Inglehart and Welzel 2005, Starmans et al. 2017).Therefore, a null hypothesis proposes no differences in perceptions of specific restrictions between neighboring communities.However, based on this discussion, we can predict Table 2. Summary of reported perceived net benefits of common fisheries restrictions of interviewed managers and resource users in the four studied countries Perceived differences (managers -resource users) between the benefits scaled by managers and resource users.Results are summarized from (McClanahan and Abunge 2016) based on interviews with ∼2100 stakeholders.+, perceived positive benefits; 0, perceived neutral benefits; -, perceived negative benefits.that differences between perceptions of restrictions of the neighboring community will (1) be higher than within-community variation; (2) increase with the spatial and temporal scale of the specific restriction severity and potential net benefits; (3) decline with perceptions of fairness, wealth, education, and residence time; and (4) vary by the interrelated factors of human population density, strength of the national economies, communication, governance strength, histories of protective management, and equity.These hypotheses guided our evaluation of variability within and between 89 reef-fishing communities in the countries of Kenya, Tanzania, Madagascar, and Mozambique.This information was then used to evaluate the spatial distribution of this variability and potential areas for conflict and cooperation over specific restrictions.

Regional overview
These nations are heavily reliant on natural resources and differ in their experiences with conservation and fisheries management (Hicks 2011).Eastern and southern African countries have variable economies but are generally poor and highly reliant on natural resources (Levin et al. 2018; Table 4, Fig. 1).The studied nations have different ethnic, colonial, and postcolonial histories, and their recent international trade and governance affiliations reflect these histories.GDP declines along the Kenya, Tanzania, Madagascar, and Mozambique sequence, as do the taxes spent as percentage of GDP.There was a legal and social movement toward decentralization of governance and resource management in the region that was at different stages of development and donor support and seldom fully implemented when we collected our data (Cinner et al. 2012).
Kenya has followed British-style parliamentary governance with a strong central government.Tanzania has had a strong central party with a prosocialist government that historically has controlled the economy and pricing, but these controls have declined over time.Tanzania has the lowest GINI coefficient among the countries, high rule of law, and low judicial effectiveness, and the dominant party and government has had a strong influence and a history of conflicts with coastal communities over resource management (Walley 2004, Benjaminsen and Bryceson 2012, Kamat 2014, Katikiro et al. 2015).
Madagascar has a French-style representative democracy and free-trade economy but has low tax revenue and spending and weak national governance, communications, and judicial penetration and effectiveness.Madagascar is poor and has a low population density, but some coastal regions are highly reliant on fisheries resources.Because national government penetration is low, international donors and trust funds often support a customary management system called dina (Cinner et al. 2009, Carrett 2013).
Mozambique has moved between socialist and more capitalist forms of government and economies with strong regional political divisions.There is a history of conflict in many Tanzanian fisheries and parks as well as low compliance (Horrill et al. 2000).Kenya has the largest percentage of well-protected reefs in closures, but gearrestricted reserves are poorly managed (McClanahan et al. 2005).

Study sites and design
Fishing villages were visited between 2008 and 2012 with the intention of evaluating their management preferences (McClanahan et al. 2008(McClanahan et al. , 2012(McClanahan et al. , 2014)).In most cases, we selected a group of three to five landing sites near or associated with some specific management systems, such as an MPA or a proposed management system.Fishers were often strongly associated with a fishing ground and landing site, but this may largely be because of limited access or affordability of transportation rather than a sense of ownership.Fishers often walk or use bicycles to reach landing sites, and this restricts their movements.Fishers most often used body-and wind-powered boats, and engine-powered boats were uncommon.
Most sampled neighboring villages were close to each other, but some were far enough apart that they would seldom interact.Consequently, we set a maximum distance of 40 km for evaluating neighboring landing site interactions.Therefore, a total of 89 landing sites were sampled, which resulted in 60 pairwise nearestneighbor comparisons for sites that were <40 km apart.The mean linear distance for these 60 comparisons was 8.99 ± 8.48 km (± standard deviation), which we consider an appropriate scale for potentially interacting at the scale of the fishing grounds.With emerging efforts to promote local control and comanagement, there were increasing numbers of local management groups formed and associations with these groups.Although similar in many ways, they assumed different names in the four western Indian Ocean (WIO) countries of study.For example, in Kenya and Tanzania, the fisher organizations were known as Beach Management Units, and in Mozambique and Madagascar, they were called Conselho Comunitario de Pesca and Village Fisheries  14) 10.5 ± 5.0 (4) 11.7 ± 11.0 (3) † National wealth rank is extracted from World Bank wealth ranking of nations for either 2011 or 2015 when 2011 data not available.‡ Access to knowledge and information is sourced from Social Progress Index.§ Fisheries information extracted from (Burke et al. 2011, Teh et al. 2013).Artisanal fishery is fishing mostly done by men and is for both food and income, whereas subsistence fishery is for food and is dominated by women in most countries (Pauly 2007).
Committees, respectively.Legal jurisdictions were often being discussed and established but less seldom implemented or strongly enforced, which made it difficult to evaluate communities by their variable adherence to legal or informal justice procedures.

Field interviews
Community-level variables were the means and variances resulting from face-to-face interviews of 10 to 30 individuals per landing site.A total of 1849 interviews were completed in the 89 landing sites.Individual resource users were chosen through a structured and randomized process whereby landing site leaders were consulted and a list of resource users and their primary gear was obtained.From the proportion of people using each fishing gear, specific interviewees were chosen randomly to obtain a stratified (gear type) random design.During the project-planning meeting, the landing site leader provided the names of all fishers by gear type, and researchers targeted a minimum of 30% of total fishers per site and by gear.Proportional sampling with respect to number of fishers and gear type was followed in most sites except in marina and reef where one gear type was dominant.
A standard questionnaire was used in all the fish landing sites and reported in a previous publication (McClanahan and Abunge 2016).The questionnaire assessed the user's perception of six basic management restrictions and who benefits from them.
Interviewees were asked about each of the specific management restrictions and how they were perceived in terms of their benefits and sustainability.Each restriction was addressed separately (Appendix 1).The respondents rated their support, and we coded them from −2 to +2 values, the lowest and highest values being that they totally disagreed or agreed, respectively, with the statement about sustainability."Don't know" responses were not included in the analyses.These coded responses of −2 to +2 were normalized to a positive 1 to 5 scale by adding 3 to all values for the purposes of statistical analysis.https://www.ecologyandsociety.org/vol23/iss4/art33/Questions about who would benefit the most from each of the 6 management restrictions and how the fisheries could be improved followed.Beneficiaries included individual resource users, their communities, and the national government, and respondents marked a point on a continuous 1 to 10 scale to estimate the benefit of each restriction, 10 being the maximum benefit (Appendixes 1 and 2).Differences between the estimated benefits to each beneficiary were calculated and explored to determine a best measure of social disparity.The average difference between national government and community and self-benefits was chosen, and the following formula used.Disparity = (benefit government − benefit community) + (benefit government − benefit self) / 2 Socioeconomic characteristics of the fishing communities that might influence these responses were measured by standard socioeconomic questions reported in previous studies (Appendix 2; McClanahan et al. 2009aMcClanahan et al. , 2012)).These included the 10 variables, of which 8 were demographic and economic including number of jobs, years in occupation, residency, education, age, expenditures, material style of life (MSL; = physical capital), and household size.Two variables measured involvement in community organizations as membership in either fishing or conservation groups.MSL questions on house type and possessions were based on a principle component analysis (PCA) of the presence or absence of household items (Cinner and Pollnac 2004).Community averages were based on those individuals that answered all questions.

Data analyses
Responses to the previous questions were evaluated for withinand between-community differences in variations.Prior to statistical tests, response metrics were tested for normality using the Shapiro-Wilk W test. Transformations were explored; log 10 transformations were used for between-community data, and logit transformations for coefficient of variation (COV; %) data.Both within-and between-community data sets did not meet the assumptions of normality, but within-community data weakly satisfied these assumptions; hence, nonparametric methods were used to test for significance.
The pairwise physical distances between communities were calculated using ArcGIS software (version 10.2.2; ESRI Inc., CA, USA).We took the GPS coordinates with handheld GPS receivers while visiting the landing sites.These coordinates were used to make a shape file using the World Mercator 1984 projection coordinate reference system.From this shape file, pairwise distances were calculated, obtained in ascending order for each site pair, and the shortest paired distance was used to establish the nearest neighbor.Tests for clustering in the spatial variation of the mean perceived benefits used the Moran's I index for withincommunity variation and between-community distances.
The COV (% = standard deviation / mean perceived benefit × 100) of the normalized (adding a constant to eliminate negative numbers) response to benefits for each and all restrictions combined at each landing site was the within-community metric of variation.Between-community variation was calculated as the absolute difference between the same restriction responses for nearest-neighbor communities.This distance was used to evaluate the variation between communities for each fisheries restriction.
National within-community estimates were based on means of means and, therefore, were appropriate for parametric statistics.Intercommunity values were not means of means.Therefore, Wilcoxon/Kruskal-Wallis nonparametric tests and post hoc Dunn's test with Bonferroni adjustments for multiple comparisons were used.Transforming and normalizing the intraand intercommunity data allowed testing for significance of the interactions between nations and restrictions.Maps of the withinand between-community variation used z scores of the within and nearest-neighbor distances to visualize the spatial distribution of variability among the six management restrictions, variability being a proxy for community agreement and disagreement.
Within-and between-community variations were also tested for associations with the 10 socioeconomic variables (Appendix 3) using a stepwise forward logistic regression procedure for each country and all countries combined.A PCA of the six management restrictions was done to ordinate the within-and between-community variation of all communities in all countries along these six axes of restriction variation.Key country-level economic descriptor vectors were included as supplementary variables to examine the relationships with the variability in each restriction.All statistics were performed using JMP version 13.0 statistical software (Sall et al. 2001) and R (R Development Core Team 2015) using the FactomineR package (Josse and Husson 2016) and visualized using the factoextra package (Kassambara and Mundt 2016).

Spatial distribution of mean perceived benefits
Mapping of the mean perceptions of the benefits of all restrictions combined indicates that Kenya and Tanzania had the highest variation or broadscale disagreements about benefits (Fig. 1).For example, in northern Kenya, adjacent communities were among the most and least positive about restrictions.Communities near the Mafia and Mtwara MPAs in Tanzania were among the least positive clusters as were some communities near the Dar es Salaam Reserve.In Madagascar and Mozambique, there were many communities with intermediate, but none with low, scaling.

Comparing within-and between-community variation
When evaluating all restrictions combined, all factors of nation and restriction were statistically significant for within-and between-community variation (Table 5).When tested for differences between nation and restriction type, withincommunity variation displayed larger differences than betweencommunity variation.Nation was the largest source of withincommunity variation, followed by restriction and, finally, their interaction.Differences in between-community variation differed significantly by nation, restriction, and their interaction, but all factors were weak (i.e., F < 3.5).

Patterns of within-community variation
Within-community variation declined strongly with an increase in the mean perceptions of restriction benefits for all four countries (Fig. 2a).Kenya, Tanzania, and Mozambique all had strong relationships (R² > 0.90) between mean benefits and within-community variation.Madagascar had a poorer fit (R² = 0.47), but, because few communities scaled benefits lowly, the range of responses was low.
Table 5. Tests of differences in pooled restriction variation by a two-factor ANOVA based on transformed data.The strengths of variation and possible interactions between within-and betweencommunity variation were tested.Within-community variation compared individuals within a community.Between-community variation used nearest-neighbor communities, and results are presented as the percentage of the maximum possible distance.Test of significance used log-transformed and logit-percenttransformed data for between community and within community, respectively.COV, coefficient of variation; SEM, standard error of the mean.Within-community COV was 64% for Kenya, 48% for Tanzania, and ∼22% for both Madagascar and Mozambique (Fig. 3).Within-community variations in countries were distinguished primarily by four restrictions, namely, species selection, closed seasons, protected areas, and closed areas (Fig. 4a).Variations were primarily positively associated with the country descriptors of per capita income, mobile phone ownership, and advanced education (Table 4).The smallest variation was associated with size and gear restrictions, which were weakly negatively associated with a nation's GINI coefficient and government expenditures as percentage of GDP.Differences between nations and restrictions were influenced by the proposed restriction strength (Table 6).For example, gear restrictions had the lowest variation and did not vary by country.Minimum size of fish had the second lowest variation, and only Tanzania differed by having higher variation than the other nations.Two response groups were produced by protected areas, seasonal closures, closed areas, and species restrictions, namely, a high-variation group of Kenya and Tanzania and a lowvariation group of Madagascar and Mozambique.
Perceptions of benefits and fairness, or the perceived disparity between benefits to government versus those accrued to individuals and communities, were stronger than the 10 socioeconomic indicators in predicting within-community variation.All variables combined predicted 53% of the variance (Table 7).Variation declined strongly as the perceived benefit of the restriction and the total number of jobs of the respondent's household increased.Membership in fishing organizations was significant but weaker, and education and wealth, i.e., expenditures and MSL, did not reduce variation.Residence time was also predicted to be negatively associated with community variation but positively associated in Mozambique and not significant in the other countries.

Patterns of between-community variation
Overall, the four countries had largely similar betweencommunity variation (Table 8).Nevertheless, specific restrictions produced more between-country variation, and these were variably associated with country-level socioeconomic associations (Fig. 4b).Madagascar's normalized between-community variation (COV, %) was about 20 percentage points higher than the other countries and influenced by the specific restriction.Tanzania also had a high spread in between-community differences for specific restrictions, with particularly high differences for protected and closed area restrictions.This contrasts with Madagascar where species and minimum size at capture restrictions were the most variable.In strong contrast to within-community variation, between-community variation was not associated with the mean perceived benefits of restrictions in Kenya, Madagascar, and Mozambique (Fig. 2b).There was, however, a wide spread and a weak relationship in the perceived benefits in Tanzania (R² = 0.27).
Between-community had low variation in minimum size and gear restrictions that did not differ between the four countries (Table 8).Species selection restrictions had the highest variation with no differences between countries.High strength restrictions of closed seasons, protected areas, and closed areas had high variation and statistical differences between nations.Madagascar had low variation, and Mozambique was not different from the higher variation in Kenya and Tanzania.Consequently, when comparing all restrictions, Tanzania and Madagascar had significant differences between restrictions, whereas Kenya and Mozambique did not.Closed seasons in Madagascar had the lowest betweencommunity variation of all restrictions and nations.
Perceptions of benefits, fairness, and the 10 socioeconomic indicators were weakly associated with between-community variation (R² < 0.13; Table 7).Variation increased with the perceived disparity of the benefits and years in the fishing occupation.Relationships with perceptions of benefits and disparity, jobs, and occupation metrics were weak and of similar strength.Variation declined as the total number of jobs and the perceived benefits of restrictions increased.In Kenya, variation was not associated with perceived benefits and years in occupation.In Mozambique, variation was not significantly associated with any of the studied factors.

Locations of potential conflict and cooperation
There was significant spatial clustering of mean perceived benefits when all restrictions were combined and for some specific restrictions (Table 9a).Kenya, Madagascar, and Mozambique had significant but low clustering strengths (Moran's I ∼0.45-0.50),whereas Tanzania had stronger clustering (Moran's I = 0.82).In Tanzania, the clustering was significant for minimum size, gear restrictions, species selection, and closed seasons but not for protected areas and closures.Madagascar showed the same by-restriction clustering patterns as Tanzania, but the scaling of benefits of protected areas was also significantly clustered.In contrast, Kenya only showed significant spatial clustering associated with protected areas and closure restrictions, and Mozambique only for closures.Within-community variation also displayed significant spatial clustering for all restrictions combined and followed similar patterns to mean benefit values for specific restrictions and by country (Table 9b).
Between-community variation in some specific restrictions displayed weaker spatial clustering (Table 9c); Kenya had none, and Tanzania had only between-community spatial clustering for gear restrictions.Mozambique and Madagascar had the most between-community clustering and significant spatial clustering for all restrictions combined, as well as clustering of minimum size, species, and closed and protected area restrictions.Scatter plots and regression statistics indicate that despite the abovementioned clustering, there were no significant linear relationships between the physical distance between communities and the differences in the scaling of benefits within nations (Fig. 5).
Maps of the distribution of the within-and between-community variation in restriction preferences identify a number of potential areas of agreement and disagreement (Appendix 4).For example, preferences for the minimum size of fish exhibits high withincommunity agreement but with potential neighbor conflict areas in northern and southern Kenya and Tanzania, as well as northeast Madagascar (Fig. 6).Similarly, gear displayed high within-community agreement.However, where there were areas of disagreement, they aligned with size-restriction preferences.Northern Mozambique also had potential conflict areas in gear preferences.There was less within-community agreement on species selection, and some high-agreement communities had potential disagreements with neighbors in the border regions of Kenya and Tanzania, as well as southwest Madagascar (Fig. 7).
Closed season had a moderate potential for neighbor-community disagreements in northern Kenya, southern Tanzania, and northern Mozambique.Protected areas had the most potential for disagreement ranging from northern Kenya to northern Mozambique and also southwest Madagascar.Closed area responses were similar to protected areas but with less potential for disagreement in southwest Madagascar.

DISCUSSION
Perceptions of management restriction benefits are variable in this region and influenced by the national context, type of restriction, and a number of interrelated demographic characteristics.We confirmed that within-community variation declined strongly with perceived benefit and was considerably less than between-community variation.This unsurprising finding supports the homophily proposition that perceptions are more similar when people stay and interact together (McPherson et al. 2001, Ostrom 2010).More surprising was that within-community similarity did differ with the restriction type, suggesting limits to the similarities of cohabitant views when faced with potentially lower net individual benefits.Consequently, conflicts are possible over specific restrictions and may reflect each individual's views of the costs/benefits of specific restrictions, even within homogenous communities.Moreover, within-community variability was lowest in the low population density, poorer, and weak national governance countries of Madagascar and Mozambique.These two countries also had less restriction experience in terms of the ubiquity, ages, and effective national governance and compliance of protected areas and fisheries closures.Context and experiences combined with isolation, poverty, lower advanced education, reduced access to communication technologies, and fewer interactions with media and government were likely to promote stronger local norms and lower community variability (Table 4, Fig. 4a).

Sources of variation
Greater individuality within communities is expected where people are less isolated and have more wealth, education, occupational diversity, and information available through either https://www.ecologyandsociety.org/vol23/iss4/art33/Fig. 6.
Ecology and Society 23(4): 33 https://www.ecologyandsociety.org/vol23/iss4/art33/Table 8.Comparisons of between-community variation in management preferences by country (a, mean distance, + standard deviation; b, coefficients of variation, %) based on Kruskal-Wallis test and post hoc Dunn's test with Bonferroni adjustment for multiple comparisons.Numbers (1, 2, 3) distinguish significance between countries, and captial letters (A, B, C) distinguish management restrictions within countries.Values followed with the same numbers in parentheses are not significantly different from each other for comparison of nations.Values followed with the same letters are not significantly different from each other for comparison of management restrictions.Bolded values are the significant P values.Countries are arranged from most to least wealthy.Management options are arranged from least to most restrictive.the media or government.For example, a large study of trending media topics found that both personal and mass media trends followed wealth distributions, with some barriers and clustering created by differences in language and culture (Carrascosa et al. 2015).Despite the potential for greater information distribution in wealthy countries, spatial clustering of preferences was present in all countries.Within-community clustering was strongest when pooling all restrictions, but weaker patterns were observed for specific restrictions.Tanzania and Madagascar had the most spatial clustering for the low cost/benefit restrictions, such as size, gear, species, and seasonal restrictions, whereas Kenya had the most clustering for the high cost/benefit restrictions of closed and protected areas.Mozambique was more spatially uniform for specific restrictions and only clustered significantly when comparing the pooled restriction responses.Consequently, spatial variability and clustering of within-community preferences was restriction specific in some cases and, therefore, difficult to attribute solely to country-level statistics.This suggests nationallevel responses for specific restrictions within nations, possibly similar to the restraining effects of language and culture on the spread of media.
Intercommunity variation was high overall and exhibited small differences between nations, with somewhat higher variation in Madagascar than the three other countries.Consequently, the relative differences or inter/intracommunity ratios were higher in Madagascar and Mozambique.Consequently, individuals from a country with small within-community variation might sense stronger differences in restriction preferences when interacting with other communities in the nation.However, there was also higher within-community spatial clustering in Madagascar and Mozambique for the high individual cost/benefit restrictions, making conflicting perceptions less likely for some restrictions among neighboring communities.Therefore, less conflict would be expected in countries like Madagascar for restrictions that are distributed over limited spatial scales.On the other hand, Kenya and Tanzania had the potential for stronger conflicts within and between communities for some restrictions, but this depended on the restriction type.
Both scales of community variability increased with our proposed cost/benefit gradient, but more conservatively, there were two management restriction groupings: two low individual cost/ benefit restrictions of size and gear, and four higher individual cost/benefit restrictions of species selection, closed seasons, closed areas, and MPAs.Weaker differences for betweencommunity variation in size, gear, and species selection restrictions among nations support the proposition that perceptions are shared because of broadly acknowledged individual cost/benefit considerations.However, variation in preferences within communities must play some role in this variability because between-was higher than within-community variation.
Variation in the perceived benefits declined strongly as perceived benefits increased for within-but less so for between-community comparisons.This indicates that variations in costs/benefits described in Table 1 are not universal between neighbors.Tanzania was the one exception where between-community variation declined as perceived benefits increased, but the relationship was weak.Tanzania also fit our proposed cost/benefit scale best by having the greatest spread in restriction responses.Consequently, cost/benefit responses were spatially limited and could fail to be confirmed because of the competitive or polarizing effects of communications and interactions between neighboring communities.More communication is expected to reduce conflicts but can also produce polarization if neighbors are seen as potential defectors or competitors or as receiving unequal benefits from restrictions (Sugiarto et al. 2017).All mechanisms could work simultaneously and produce complex and patchy neighborhood relationships.Future research would benefit from studying the amount, type, and consequences of betweencommunity communication for achieving resource management goals in different national contexts.
Our findings support general patterns proposed by our hypotheses but also indicate contextual complexity and the importance of the nation, restrictions, and community interactions in predicting perceptions.Modeled behaviors suggest that local management benefits can be reduced when adherence levels between neighbors are not complementary (Agrawal et al. 2013).Consequently, high between-community variability in the perceived benefits for some restrictions could challenge successful compliance outcomes for many restrictions.One would predict better cooperative outcomes when communities are small or isolated from noncompliant neighbors, or when members adhere to and clearly benefit from enforcement (Powers and Lehmann 2013).Community size and market integration have therefore been good at predicting the status of fisheries and CPRM in the absence of effective enforcement or livelihood alternatives (Cinner 2005, McClanahan et al. 2006, Cinner et al. 2016).
Diagnosing and developing successful policies and guidelines for CPRM will need to consider these contextual issues.
Key persons or regional or national entities may be required to communicate and ensure high compliance outside of small, isolated, and homogenous communities, particularly for restrictions with greater national than individual and community benefits.If stakeholders are susceptible to corruption and favor the interests of small over large groups, then the potential to increase productivity and achieve large-scale benefits will be challenged (Hardin 1968, Sundström 2015).Not preferentially benefiting smaller groups that typically design and enforce management is considered a fairer system for designing rules https://www.ecologyandsociety.org/vol23/iss4/art33/(Rawls 2001).Given that management designs and enforcement were done by national governments in Kenya and Tanzania, experienced fishers expressed cynicism about restrictions with large-scale benefits.This was especially common in Tanzania where conflicts with government have been commonly reported (Walley 2004, Benjaminsen andBryceson 2012).Although many comanagement actions focus on strengthening local community governance, developing regional governance and intercommunity forums may be more important in nations with moderate capacity (Wright et al. 2016a).
Fairness in the perceptions of benefits was the strongest predictor of benefits and variation, especially in the wealthier countries and within, but also between, communities.Consequently, ensuring procedures of fairness and the distribution of benefits are options likely to increase compliance.Fairness can be achieved by attending to the four elements of justice: procedural, distributional, retributive, and restorative.Incorporating justice mechanisms into natural resource management should reduce the chances of marginalizing people historically excluded from justice (Schroeder et al. 2003, Giakoumi et al. 2018).For example, Kenyan stakeholders' perceptions of fisheries closures were shown to change from negative to neutral through transparent sharing of group benefit information (Cinner and McClanahan 2015).Reducing the costs of transactions is a key ingredient in many successful management and economic activities that should arise with increasing justice elements and the consequent trust that arises from fair and reliable transactions (Fukuyama 1995, Stern 2008).
Promoting jobs, permanence, and community organizations associated with positive views and lower variation in perceived benefits are among frequently proposed actions to improve perceptions and compliance in community and conservation development (Wright et al. 2016b).Nevertheless, our findings, like others, suggest they are weakly associated with proenvironmental behavior, perceived benefits of restrictions, low community variation, and considerably weaker than perceived benefits and fairness (Cinner 2009, 2014, McClanahan and Abunge 2016).Consequently, these options are likely be contextually important, but efforts to achieve procedural and distributional justice should be a priority for CPRM.Wealth and education did not predict positive perceptions toward restrictions and, therefore, are not likely to increase perceptions of benefits in our context.Consequently, a person's profession and sense of justice may be more important than wealth in predicting perceptions (Hicks et al. 2013, McClanahan and Abunge 2016, Turner et al. 2016).These two and other variables evaluated can often correlate and be potentially confounding when attributing causes of perceptions (Fig. 4).Multiple and hidden variables are a common problem when pooling people's values and behavioral responses into large or societal-level demographic analyses (Inglehart andWelzel 2005, Klöckner 2013).Consequently, the national socioeconomic context hypotheses we tested would require larger country sampling to better understand their roles in influencing perceptions of restrictions.

Practical applications
Our results have a number of practical applications for CPRM.Additionally, perceptions of a community closure changed as stakeholders experienced the lack of anticipated negative consequence once closures were in place (Cinner and McClanahan 2015).
Dependent on local social-ecological conditions and occupational options, perceptions are expected to align or polarize in adjacent communities (McClanahan et al. 2008(McClanahan et al. , 2012)).Consequently, polling of perceptions as we have done should help to identify, plan, and monitor the conflicts most expected to emerge in particular settings and over time.In Madagascar, for example, the preferred regulations were seasonal closure where we found high scaling of benefits and low variation.As would be predicted, there was a recent rapid proliferation of seasonal octopus (Octopus cyanea) closures in Madagascar (Rocliffe et al. 2014).Although generally successful, there was evidence for intercommunity breaching of closures that could reduce and negate benefits (Oliver et al. 2015).Consequently, low variation and clustering of community preferences has not fully eliminated breaches.Thus, imperfect compliance outcomes further emphasize the need for effective intercommunity and comanagement relationships, even for the least variable communities and restrictions.These examples beg the question of whether our results arise from recent experiences or from cultural predispositions to specific restrictions.In the case of Madagascar, many surveyed communities did not have seasonal closures, and therefore, positive views were unlikely to result from local experiences alone.Both predisposition and experiences are likely to play a role, and historical studies are required to better understand their influences on perceptions.
Conflicts over closures and other gear management systems in Kenya and Tanzania are likely to arise from negative experiences, disparate cost/benefit outcomes, and the availability of other occupational opportunities.To address conflicts, targeted communication is a starting point, especially for restrictions for costs/benefits that accrue on large time and spatial scales (Mbaru and Barnes 2017).Past perceptions of unfair practices and corrupt applications of policing could, however, be the source of this variability (Mulder et al. 2006).Regardless of the underlying https://www.ecologyandsociety.org/vol23/iss4/art33/causes of these perceptions, our results suggest that increasing the fairness of processes, reducing corruption, sharing benefits, and matching social and ecological scales for management are actions expected to lead to greater success.Given the effort to decentralize fisheries management in this region and globally (Cinner et al. 2012, Davis andRuddle 2012), matching community costs and benefits with ecological variability is badly needed.The benefits of coordinated government bodies are not likely to be achieved by wholesale movements from national to federal governance, but rather by creating a compound republic of governance (Wagner 2005).
Can our polling approach lead to better management?It should increase the chances for clearly understanding management preferences and inclusive decision making, as well increasing the four justice procedures (Schroeder et al. 2003).Perceptions of equitable processes are at the core of many political conflicts, and polling and using perceptions to craft management should increase perceptions of fairness (Starmans et al. 2017).Nevertheless, polling is only one among other procedural justice options that can be considered.Additionally, agreement and enforcement become more challenging when they involve neighboring communities less involved in, or committed to, local procedures.Even if fishing communities are connected through coordination of government leaders and hierarchies, they are not always meaningfully engaged in critical channels of communications, which can lead to failure in adopting important decisions (Barnes et al. 2016).Additionally, a neighbor's resource is an easy target without consequences if sanctions are only applied to local stakeholders.If there is a lack of agreement or if restrictions are only locally enforceable, then betweenneighboring-community dilemmas and minimum-restriction matching may ensue (Agrawal et al. 2013).
Local decisions will often fail to protect the full spectrum of nature's diversity if local rather than larger scale restrictions are always the basis for decisions (Weeks et al. 2010).Many species and ecosystem processes require space and resources over scales larger than most fishing communities (Green et al. 2015, Isbell et al. 2017).In the WIO, we found neighboring communities were often <10 km apart, and large-bodied and migratory species are unlikely to benefit from local management decisions unless they are shared broadly.There are some broadly shared restrictions, such as minimum size and gear restrictions, that can prevent the demise of fisheries resources (McClanahan et al. 2015).In the absence of shared decisions, procedures, and enforcement, the local polling and implementation of favored restrictions approach could favor only the profitability of economically important species with rapid growth and limited movements.Our study exposes contextual issues around demographic variability and the problems that local and overlapping governance institutions will need to address to better manage social-ecological dynamics.If demographic heterogeneity, economic developments, communication, state policies, and histories of justice and trust are not fully considered and crafted into management systems, the sustainably of CPRM will be challenged.
Responses to this article can be read online at: http://www.ecologyandsociety.org/issues/responses.php/10544

Fig. 1 .
Fig. 1.Map showing the distribution of fishing communities in the four countries.Sites were clustered based on the similarity of their responses to the perceived benefits of management restrictions.The nine clusters were then scaled by their mean scaling of the perceived benefits of the six management restrictions from highest in green to lowest in red.

Fig. 2 .
Fig. 2. Relationship between the scaling of the perceived benefits at landing sites and (a) within-community variation and (b) nearest-neighbor between-community distances for all six restrictions pooled.COV, coefficent of variation.

Fig. 3 .
Fig. 3. Within-community coefficient of variation (COV, %; means + standard deviation) of each landing site in the four studied countries based on pooling all six management restrictions.The mean for the whole country for all landing sites is also presented.Countries and sites are arranged from north to south.

Fig. 4 .
Fig. 4. Principal component analyses (PCA) of the fishing communities in the four nations as distributed along the (a) within-community and (b) between-community variation of the six management restrictions.Country socioeconomic variables are overlain as supplementary variables to the PCA analysis.

Fig. 5 .
Fig. 5. Plotting of management restriction preference and physical distances by country for all the management restrictions.Preference distances are based on betweencommunity differences on their rating of the management restriction (1 to 5 adjusted from −2 to +2 by adding 3).

Table 3 .
Perceived differences (managers -resource users) between the net benefits scaled by managers and resource users.Results are summarized from (McClanahan and Abunge 2016) based on interviews with ∼2100 stakeholders.++, large difference in perceived benefit/disparity; +, small difference in perceived benefit/disparity; 0, no or negative difference in perceived benefit/disparity.

Table 4 .
Coastal human population density, GDP income, inequality, and governance indicators in 2011 in the four studied countries.GDP is in U.S. dollars.Governance is a 0 to 100 scale for various metrics.Table organized from most to least wealthy people and governments.GINI index (a measure of economic inequality) and governance indicators data extracted from World Bank and Index of Economic Freedom-The Heritage Foundation.Marine protected information taken fromWells et al. (2007)and Wildlife

Table 6 .
Test of within-community variation results (coefficient of variation, % mean + standard deviation) of Kruskal-Wallis test and post hoc Dunn's test with Bonferroni adjustment for multiple comparisons.Numbers (1, 2, 3) distinguish significance between countries, and capital letters (A, B, C) distinguish management restrictions within countries.Values followed with the same numbers in parentheses are not significantly different from each other for comparison of nations.Values followed with the same letters are not significantly different from each other for comparison of management restrictions.Bolded values are the significant P values.Countries are listed from most to least wealthy.Management restrictions are arranged from least to most access restrictions.Regional mean 26.67 ± 18.66 21.65 ± 17.47 47.93 ± 21.38 44.44 ± 23.96 33.86 ± 25.28 45.54 ± 26.56 36.68 ± 15.64

Table 7 .
Results of stepwise regression analysis evaluating socioeconomic variables associated with between-and within-community variations.The variables are arranged from strongest to weakest association.Results presented are (a) all countries using pooled data, (b) Kenya, (c) Tanzania, (d) Mozambique, and (e) Madagascar.Signs of the t ratio indicate the direction of association with the variation.

Table 9 .
Moran's I statistic testing for spatial associations for (a) mean management restrictions responses and the variation attributable to (b) within-community and (c) between-community distances.Tests determine if the distributions in space are different from the null hypotheses of no significant spatial organization.Bolded numbers represent significant P values for each management restriction in each of the four western Indian Ocean countries: Kenya, Tanzania, Madagascar, and Mozambique.