Anda di halaman 1dari 8

ANNEX 5: PRINCIPAL COMPONENTS ANALYSIS AND THE WELLBEING INDEX

Defining an index of wellbeing There is no best practice approach for selecting variables which are indicators of socio-economic status or well-being1; however formative research can contribute to the identification of assets that are strong predictors of socio-economic position (Howe et al., 2008). Broadly speaking, socio-economic status involves many dimensions. In this study, variables were selected based on a literature review of analyses into the social dimensions of fisheries systems and indicators used in social impact assessments, broader evaluations of well-being and variables emerging as key differentiators between different case study locations. Variables most suitable for this type of analysis are those which have a wide range and are highly variable across the subjects surveyed in order to be most informative. Variables also needed to be relevant to all subjects, so while relatives working in the organisation was seen to be informative for some fleets (e.g. where crew were related to one or more other members) this variable was not necessarily relevant to fleets that were dominated by individual operators so the results would not be meaningful across the extent of the individuals and is therefore misleading. Variables which would be robust to time of analysis were also used, so rather than count assets which takes place in some studies, such as the ownership of specific navigation devices or other expensive technology, less time-dependent material variables (i.e. those whose very nature or utility is less likely to change over time) were used such as health and income. The use of a single proxy is unlikely to lead to reliable results, so the idea is to incorporate a number of such proxies or variables, covering each of the dimensions required. Key issues that arose in the case studies and particular areas with differential results were considered carefully in determining which variables to use in the well-being index. A key social feature that became apparent from the case studies was the different roles played by spouses in an environment dominated by ageing, male fishers. This role ranged from one highly active in the household fishing operations on an informal basis to being completely disassociated from the activities, undertaking an entirely different career path. While much of this was qualitative information based on qualitative methods, one quantitative relational variable which was recorded and could be included was the presence of the partner receiving benefits from the informal fishing related activities, so this was selected. Using household income in addition to individual income was also a way of including this aspect, incorporating the dimension of whether a household was entirely dependent on income from fishing or not. Another key variable that varied markedly depending on the case study region was income, so this was also included as a variable. As bounds were used, the midpoint of each of these was selected to provide a continuous variable. While difficult to measure, health was a key attribute that emerged from the respondent rakings of factors influencing their well-being and of their ranking of health was therefore also included as a variable in the analysis. Education did not emerge as a key factor in affecting entry into the fisheries segment, or employment within the fleet in any particular way so this was not used as a variable in the analysis. While the responses relating to gender were extremely interesting based on respondent views, the sample number was still too low to incorporate much diversity within this as a variable, so this was not used either. Instead, the subjective factors such as solidarity of the catching sub-sector, particularly influenced by the presence of local organisations such as FLAGs, cofradias and producer organisations proved to be very important to fishers. There was obviously a balance between which aspects were measured quantitatively and which issues emerged as important through all of the methods used which had to be struck in selecting variables. Based on this approach and the rationale described, the variables used were: Case study level variables Number of vessels in the case study location Proportion of people employed in fisheries in the case study location Unemployment rate in case study location Proportion of local people in case study area Material variables
1

MontgomeryM.R.,GragnolatiM.,BurkeK.A.&ParedesE.(2000)MeasuringLivingStandardswithProxyVariables.

Demography37,155174. HouwelingT.A.,KunstA.E.&MackenbachJ.P.(2003)Measuringhealthinequalityamongchildrenindeveopingcountries:does thechoiceoftheindicatorofeconomicstatusmatter?InternationalJournalforEquityinHealth2,8. HargreavesJ.R.,MorisonL.A.,GearJ.S.S.,KimJ.C.,MakhubeleM.B.,PorterJ.D.H.,WattsC.&PronykP.M.(2007)Assessing householdwealthinhealthstudiesindevelopingcountries:acomparisonofparticipatorywealthrankingandsurvey techniquesfromruralSouthAfrica.EmergingThemesinEpidemiology4,19.

Annual individual income (numerical value) Health (score out of 10 provided Annual household income (numerical value) Education level Proportion of income from fishing Hold another job Age Gender Nationality Benefits received

Relational variables Partner benefits Father worked in fisheries Mother worked in fisheries Involvement in a fisher association Size of household Spouse worked in organization Relatives in the organisation

Subjective variables Job satisfaction Quality of life Whether he/she would consider changing job Solidarity of catch sub-sector Solidarity of community Would consider changing job Have actively looked for another job

There were no rejected variables as this is an approach in which the variables are selected a priori based on the researchers decision as to which elements to include. Nevertheless, the PCA was re-run with different subsets of variables to test the robustness of the method used and the variables selected and the results proved very similar. Data requirements Data to inform a PCA must be numerical so categorical variables were not included in this approach if they could not be quantified. Raw data on each of these variables were obtained for every individual from the respondent questionnaires conducted on culmination of the focus groups in each area. Calculation of the index required there to be no missing data, so where the full set of variables were not available for an individual, an average value was used. Method The idea for constructing a socio-economic index is then to aggregate these multiple variables into a single index which best represented the set of information. This reduces the data from a complex, multidimensional frame to a single dimension which is easier to interpret. The general method of doing this is to assign weights to each of the observed variables and summing the total. In this study the material, relational and subjective wellbeing variables were reduced to a single index of well-being using a Principal Components Analysis (PCA). For data sets with many variables the variance of some axes may be great, whereas others may be small, such that they can be ignored. This is known as reducing the dimensionality of the data set such that one might start with thirty original variables but might end with only one or two meaningful axes. The formal name for this approach of rotating data such that each successive axis displays a decreasing amount of variance is known as PCA. In particular a PCA allows us to identify the principal directions in which the data vary by transforming a set of correlated variables into a set of uncorrelated components. The first principal component is selected as the linear index of all the variables that

captures the largest amount of information common to all of the variables which may then be used as the index2. This approach allows the determination of the most appropriate weightings for each variable to derive an index which captures maximum variation. Calculation A PCA is a statistical procedure concerned with elucidating the covariance structure of a set of variables. It is a method that projects a dataset to a new coordinate system by determining the eigenvectors and eigenvalues of a matrix. It involves a calculation of a covariance matrix of a dataset to minimize the redundancy and maximize the variance. Given a data matrix with p variables and n samples, the data are first centred on the means of each variable to ensure the cloud of data is centre on the origin of the principal components without affecting the spatial relationships of the data or the variances among the variables. The first principal component (Y1) is given by the linear combination of the variables X1,X2,, Xp

Y1 = a11X1 + a12 X2 + + a1pXp

The first principal component is calculated such that it accounts for the greatest possible variance in the data set. One could make the variance of Y1 as large as possible by choosing large values for the weights a11, a12a1p so to prevent this weights are calculated with the constraint that their sum of squares is 1.

a112 + a122 +

The second principal component is calculated in the same way, with the condition that it is uncorrelated with (i.e. perpendicular to) the first principal component and that it accounts for the next highest variance. This continues until a total of p principal components have been calculated, equal to the original number of variables. The weights aij are the elements of an eigenvector of the covariance matrix of the original data (also called factor loadings). The eigenvalues are the variance explained by each principal component and are constrained to decrease from the first principal component to the last. Results In the analysis of the variables described above, the resulting first principal component explained 28% of the variability in the data. The factors loadings associated with the variables indicate which are the most important in terms of distinguishing between different levels of well-being and so which variables the index is most sensitive to. Those with the highest factor loadings are those most highly correlated with the first principal component, or the best single-dimensional descriptor of the dataset. As the data have been scaled and centred, the resulting principal component and index of values based on this component are all relative values enabling comparisons, however their absolute values do not mean anything. Many of the subjective and relation variables were highly correlated with other variables and these had the highest factor loadings, such as quality of life, job satisfaction and solidarity of the catch sub-sector and the community (Table 1), whereas the variables which were less important in the index, but still contributed to the distinction, included income and partner benefits received.

Table 1: Factors loadings of the first principal component Variable Factor loading (eigenvector of the covariance matrix)

2 FilmerD.&PritchettL.H.(2001)Estimatingwealtheffectswithoutexpendituredataortears:anapplicationtoeducational
enrollmentsinstatesofIndia.Demography38,115132. MckenzieD.J.(2003)Measuringinequalitywithassetindicators.JournalofPopulationEconomics18,229260. VyasS.&KumaranayakeL.(2006)Constructingsocioeconomicstatusindices:howtouseprincipalcomponentsanalysis.Health PolicyPlan21,459468. CordovaA.(2008)Methodologicalnote:measuringrelativewealthusinghouseholdassetindicators.In:M.A.Seligson&E. Zechmeister(Eds)AmericasBarometerInsights.6:VandetbiltUniversity,9.

quality.of.life job.satisfaction Solidarity.of.catch.sub.sector HH.income Solidarity.of.community Health Personal.income Partner.benefits mother.in.fisheries Involvement.fisher.association HH.size Education spouse.child.ever.worked.in.org. Prop.income.fishing No.vessels.CS Age Benefits Prop.employed.in.fisheries.CS father.in.fisheries Gender Origin relatives.in.organisation Prop.unemployment.CS looked.other.job consider.changing.job Prop.local.CS Other.job Discussion - exploring an index of wellbeing

0.34 0.34 0.30 0.29 0.29 0.29 0.19 0.16 0.12 0.09 0.04 0.03 0.02 0.02 0.01 -0.007 -0.011 -0.013 -0.015 -0.022 -0.061 -0.097 -0.165 -0.204 -0.245 -0.287 -0.336

The difference in wellbeing index scores by case study site are provided in Figure 1. These scores are based on the centred and scaled covariance matrix so the absolute values are not important, but the relative values provide an indication of the difference between fishers in the different regions.

Wellbeing score

-6

-4

-2

Sicily

Shetland Region

Galicia

Figure 1. Difference in well-being index scores among case study sites A key observation from these results is the relatively lower wellbeing of the Sicilian fishers compared with other regions. This may be partly due to the low ranking of subjective factors such as job satisfaction and quality of life, which were ranked particularly low by respondents in Sicily (Figure 2). This was also true for the opinions of Sicilian fishers regarding community and catching subsector solidarity (Figure 3) and they were the only respondents to state that they would consider changing jobs in the future. This could be because these are actually lower than in the other areas, or it might be due to a cultural attitude in which situations are perceived to be worse than they are that is reflected across Sicily or Italy as a whole. This is unlikely, given that Italy has been clustered as relatively high happiness levels (46.4) in the 2012 Happy Planet Index3, higher than Spain (44.1) and close to that of France (46.5) and the UK (47.9). In addition to these low ranked factors, the lower estimated wellbeing of the Sicilian fishers could also be partly due to the relatively low annual income obtained by Sicilian workers (mostly <= EUR 10,000) compared with those in Shetland and Galicia where no one receives less than EUR 10,000, indeed some fishers in Whalsay receive over EUR 60,000 annually. While conditions in Sicily seem relatively negative, including the fact that they were the only fishers who admitted to considering or actively looking for other jobs, the reason that fishers would not consider changing occupation in Galicia might not just reflect positive associations with the fisheries sector but might instead be indicative of the poor economic situation in all areas of the economy and the current lack of opportunities elsewhere. According to the most recent survey by the Sociological Research Center (Centro de Investigaciones Sociolgicas-CIS, Opinion patterns Report) confidence in potential economic improvement is still very low across Spain. Nevertheless, the results of the index remain very similar even when different combinations of variables are used. The results are also confirmed by the qualitative results from the detailed case study reports, suggesting that the index may provide a fairly good indicator in situations where a detailed report cannot be completed (see Section 6).

http://www.gfmag.com/tools/global-database/ne-data/11940-happiest-countries.html#axzz2b6Yml7LV

Figure 2: Respondent raking of subjective variables

Figure 3: Respondent ranking of solidarity The results from the index are quite interesting as merely considering fisher incomes alone, one would not expect there to be such a dramatic difference in the wellbeing of fishers in Galicia and Sicily. While incomes in Shetland are much higher as mentioned, those in Galicia and Sicily are much more comparable, yet the subjective variables indicate that fishers feel very different in these two regions. This difference is illustrated in Figure 4 which shows that the community solidarity in Galicia is much higher than in Sicily, and is even higher than in Shetland. This may be due to the lack of collective organisation of the fishers in Sicily, where they operate on a far more individual basis, concerned by the needs of their own family. In Galicia there is a much more collective approach to the fisheries and organisations such as the cofradias and FLAGs are highly active. This is reflected in the results indicating that the level of wellbeing is higher for fishers who were members of a fisher association such as a Cofradia or other support group (Figure 5). However, this difference was not significant.

raw _dat$Solidarity.of.community

10

Italy

Scotland

Spain

Figure 4. Solidarity levels of the communities in each region

well being score

-6

-4

-2

0 Involvement in fisher association

Figure 5: Wellbeing score of fishers who are and are not members of a fisher association The wellbeing analysis was an attempt to combine a number of variables to provide a single index that could be used to track socioeconomic performance (a key priority for future data collection). The analysis was also able to highlight important individual variables that contribute to wellbeing that can also guide the prioritisation of indicators for the DC-MAP. It was also clear that there are variables and indicators related to the wider community that it would be difficult to collect through the DCF but that this study has suggested are fairly easily accessed and that can be used to examine the dependence on fisheries and the degree of heterogeneity between the sector and the wider community. The viewpoints of respondents about their quality of life, job satisfaction and other subjective factors correspond well to the index of wellbeing developed in this study as they place slightly lower importance on income compared with more subjective factors, highlighting that these subjective factors are at least as important a consideration in evaluating the social wellbeing of fishing communities. Factors which were ranked more variably (such as diversity of household income streams) were not include in the wellbeing index as these only corresponded to increased wellbeing in some regions, not universally.

An aggregate index such as this one is useful in looking at the wellbeing of people without only considering income. It highlights differences in wellbeing where the economic situation may be similar, allowing identification of some of the differentiating factors. While Spain is currently undergoing an economic downturn, the wellbeing of fishers was ranked relatively high compared with Sicily where the economic outlook is also not particularly bright. This is due to a number of the other variables included in the index, such as the high social cohesion in the areas surveyed. A further step would be to investigate why respondents felt positive about these subjective aspects and which institutional, regulatory or other factors contributed to these. While the variables selected are considered robust, the index is representing the variables selected, and it is important to note there may be other variables that might impact socio-economic status and wellbeing. This was an exploratory study and based on the situation in the four selected regions and particular locations within them. The sample size was particularly small for some locations such, as Shetland, where communities are small and few fishers were willing to participate. Nevertheless, these results do provide a useful starting point for examining the utility of the approach and the effects of including subjective as well as material elements in assessing fisheries performance. This method could be further developed, tested and adapted to incorporate a wider range of variables and/or contexts. Comparing the results using a larger set of variables will provide an additional test the robustness of the measure. While this index was calculated based on the set of data for variables obtained from the case study results, these same variables could be expected to provide a useful description of the socio-economic status of fishers in other areas. A statistically derived index of wellbeing provides a useful starting point for investigating what contributes to wellbeing and of the relative levels across different communities. Although this study has only focussed on four regions, within these there has been a substantial amount of diversity in the role fishing plays in livelihoods in each location and how the various issues emerging influence wellbeing.

Anda mungkin juga menyukai