Page 7 - HIMES_2007_

Performance Indicator Importance in MPA Management 607

closest judgments (Mardle et al., 2004). An example of one comparison between two of
the EIMR performance indicators is shown here:
Downloaded By: [University of Leeds] At: 13:49 18 January 2008
Generally, a mail-based survey approach is used to distribute the questionnaire.
However, based on suggestions given by Mardle et al. (2004), the questionnaire for
this study was presented in face-to-face interviews in order to prevent the loss of vital
interaction between participants and interviewers. All interviews were conducted in the
summer of 2004. A random sample of 53 respondents completed the questionnaire. A
representative working population, including members of the five stakeholder groups, was
randomly chosen to represent the general population of individuals that utilize the EIMR.
Questionnaires were presented to fishers at their fishing vessels, to MPA managers at their
offices, to researchers at local universities, and to local residents in bars and caf´ es in the
center of town on each island. Due to low population numbers, a census was attempted
for the populations of fishers, management officials and researchers. A random, stratified
sample was taken of local residents, by attempting to interview all residents that entered
bars and caf´ es that the interviewers sat in, and an attempt was made to interview an equal
number of men and women.

The Analysis of Performance Indicators Priority Preferences (Steps 3 and 4)
Priorities from the pairwise comparison scales used in the survey were derived for the
indicators in terms of their importance in achieving the overall goal (i.e., a successful
MPA). For each respondent, a pairwise comparison reciprocal matrix (A) of judgments was
constructed, a key and important feature of the AHP:

1 ···
 a 1 a 1 
 a 2 
 a n
 
 
 
 
 a 2 a 2 
 1 ··· 
a 1 a n
 
A = a ij = (1)
 
 · · 1 · 


 
 
 
 


a n a n ··· 1
 
a 1 a 2
where a i is the relative numerical preference (from –9 to 9) of performance indicator i.
Relative priorities were then derived for each of the defined alternatives from the pairwise
comparison reciprocal matrix by solving (Saaty, 1977; Wattage & Mardle, 2005):
n

a ij w j = λ max w i , ∀i(a ji = 1/a ij and a ij > 0) (2)
j=1
where a is an individual element of the preference matrix, i and j indicate the ith and jth
indicators, λ max is the largest eigenvalue, and the weights (w) are normalized appropriately,
n

w i = 1 (3)
i=1
The positive reciprocal matrix (A) and the set of equations (2) are solved using the
eigenvector method. The solution is normalized in this case as shown in equation (3).
Furthermore, an indication of respondent’s consistency in providing responses to each