3.- RESULTADOS DEL MÉTODO DELPHI (BARRERAS AL PNAN)
3.3. Grupo de Ejecutores
The geographic area selected for the survey was the West Midlands Region of the UK.
The survey area corresponded to the postcode areas supported by the regional development agency (AWM). The area was considered ideal for the research project due to its economic size and diversity of business.
The West Midlands has a population of 5,366,700 (9% of the GB total), with 197,592 registered firms employing 2,511,300 staff (Sutherland 2008). It has a diverse economy based on both urban and rural enterprises. Manufacturing is still important to the region employing 285,500 people and generating 15% of the regions total GVA (Gross Value Added) but still a significant decline from the manufacturing sector’s 33% recorded in 1989 (Medland 2011). In the same report, Medland stated that the West Midlands is found to have the highest proportion (14.5%) of working people with no qualifications in the UK. The West Midlands has been blighted for decades in what Worrall (2007) describes as ‘low-skill equilibrium’ but also found that surviving firms had been able to compensate for lack of internal knowledge and resources by using external partners to survive and change.
The economy of the region has indeed undergone significant change over the past twenty years. The biggest change being the growth of the service sector, where 49% of employees are now employed in a wide range of service businesses, including banking, insurance, financial services, property and business services, health care, social work and
education. The service sector provides over half of the region’s GVA (£49.1bn) with the largest sector being property and business services (£17.8bn). The region’s capital city is Birmingham with a population of approximately 1 million people (ONS 2008). The West Midlands central location in the UK means that it has good transport links to other parts of the country and excellent direct air connections from Birmingham Airport to 180 destinations in Europe, North America, Asia and the Middle East, carrying 9.5m passengers in 2008 (Medland 2011).
FIGURE 4.4
Map of the West Midlands with Postcodes
The West Midlands region shown in Figure 4.4 (source: The Post Office) is a land locked area of 13,000 square kilometres. It is often described as an area of contrasts. The region includes the densely populated conurbations of Birmingham and Coventry, surrounded by rural and often remote countryside stretching from the Welsh border to the Peak District in the North, across to its border with the East Midlands.
The sample frame was defined as senior employees or directors of firms within the region who were actively engaged in business networks and networking activities. The target sample was defined as being directors and executives of firms who were members of a business network, networking group, professional association or professional institution and therefore had a good knowledge of business networks and networking.
The sample frame was designed to identify respondents at firm/actor level, representing commercially active businesses in the West Midlands, as defined by the government funded regional development agency (Advantage West Midlands) in accordance with the sample frame guidelines suggested by (Alreck and Settle 1995). Recognising that it is difficult to obtain ‘a perfect sample’, considerable attention was paid to making the sample frame relevant to the target firms in the survey, to ensure compliance with the sample frame criteria, to obtain a range of responses representing the geographic, demographic and economic diversity in the region.
A high degree of reliability and validity in the sample is a prerequisite for a robust survey, free from bias and random error. The most common test for reliability is one of
‘repeatability’ where the distribution of data can be repeated between samples being surveyed in the same way. To be considered reliable, a sample must be free from random error. By conducting pre-survey interviews to check the relevance and accuracy of the research assumptions, greater confidence can be attributed to the final survey sample (Sekaran 1992). To be valid, the sample must be free from extraneous factors that can influence the results in a particular direction (Alreck and Settle 1995). Anything that introduces a degree of systematic bias to the sample may result in the results being less valid. Bias may inadvertently be introduced at any stage in the survey process and any factors that would change the probability of a qualifying respondent being ruled out should be avoided (Alreck and Settle 1995).
Another potential cause of bias in this type of survey is common method bias (CMB) or common method variance (CMV) as the effect is more commonly known (Doty and Glick 1998). Method bias can be a problem if it results in measurement error and therefore affects the validity of empirical results and associated conclusions. CMV is defined as a variance attributable to the measurement method rather than the individual constructs under consideration (Podsakoff et al. 2003). Offering a detailed explanation, Podsakoff (2003, p.879) state; “Based on theoretical considerations, in a hypothesized relationship between two constructs, it might be expected that measures of one might be correlated with the other, however, if they share common methods, those methods may exert a systematic effect on the observed correlation between the measures”. However, given the different nature and likelihood of CMV in the literature, it is not clear whether applying a post-hoc statistical technique to further justify researched findings is
appropriate (Richardson et al. 2009). Although possible statistical tests for CMV vary in method and outcome, the consensus for researchers is to follow good measurement practice by implementing procedural remedies related to questionnaire and item design and to control for method bias by; (a) considering the source for predictor and criterion variables, (b) assessing whether predictor and criterion variables can measured in different contexts, (c) identify whether the source of the method bias can be identified, and (d) whether the method bias can be measured (Podsakoff et al. 2003).
Podsakoff et al. (2003) catalogued the advantages and disadvantages associated with methods for assessing and controlling for CMV/CMB. Among the various methods suggested (e.g. Harman’s single factor test) those based on confirmatory factor analysis tend to be the most rigorous (Podsakoff et al. 2003). Following the recommendation Harman (1967) all the measures used in this research were collected using the same questionnaire. All the variables were entered into an un-rotated principal components analysis, as reported in Chapter 6. In this technique, if a single factor emerges from the analysis, or one factor amounts for most of the covariance in the scores, common method variance may be present. In this study, the results of the analysis reported later in Chapter 6 indicate nineteen items with eigenvalues greater than 1 and that no single factor amounted for more than 33% of the covariance. The results indicate that CMV, though probably present in the data to some degree, does not affect the results in this research.
4.6.1 Sample Size
Determining the sample size is critical to the degree of confidence required in the survey (Salant and Dillman 1994). There is a direct relationship between sample size and sample reliability (Alreck and Settle 1995). It is generally agreed that the larger the sample size, the greater the reliability of the survey, with the incidence of sampling error reduced (Bryman and Cramer 2005). It is obviously not practical to survey the entire population, in this case approximately 200,000 registered firms in the region of the West Midlands, so a suitable sample size had to be determined with a reasonable confidence level (Render and Stair 1990). The subsequent sample size which met the sample frame criteria was 3013, or approximately 1.5% of the 200,000 of firms in the region, which was therefore considered representative of the registered firms in the West Midlands.
Confidence level is defined as the probability that a value in the population is within a specific numeric range from the corresponding value calculated from the sample commensurate with the likely standard error (SE) and confidence interval (CI).
For this survey, a sample of 3013 firms located in the geographic region of the West Midlands, were identified from data supplied by different sources. Firms were selected from database listings and developed in collaboration with Advantage West Midlands.
Organisations giving permission to use their membership data included the regional Chambers of Commerce, plus data obtained from a number of established business networking groups in the West Midlands, including Business Network International, NRG Networks, 4 Networking, Birmingham Forward, Telford Business Partnership,
FineST (Stoke on Trent), Business Referrals Xchange, Coventry First, Success (Lichfield) and Alliance 4 Black Country.
The sample framework required that the selected firms should have knowledge of business to business networks and to participate in networking activities. By using data supplied by the various networking groups across the region, it could be reasonably assured that respondents would qualify by meeting the sample frame criteria. By focusing attention on respondents who are seen to be the key ‘actors’ representing their firms in a network, it can be argued that these individuals, being influential, enhance the effectiveness the network and will therefore add knowledge to the study (Cross and Prusak 2002). The identification of key informants and the issue key informant competence (Phillips 1981), has been addressed in the survey design by ensuring informants were at director or senior executive level identified by job title, years of service, membership of networking organisations and by personal networking experience.