IMPLEMENTACIÓN DE LA SOLUCIÓN
8.4 PUNTOS CRÍTICOS
In this section, the author lists the limitations that arose from methodological issues, operational feasibility and time constraints.
1 The author was unable to control for the intangibility o f assets because all potential proxies (R&D expenditure, R&D to asset ratio and advertising expenditure) for this construct required the collection o f firm-sensitive information. The collection o f this data was operationally unfeasible.
2 Measures o f bank credit availability, corporate governance and loan characteristics were self-assessed. Self-assessment raises issues about the reliability of these measures.
3 Minimum sample size for regression:
For Saunders et al. (2003, p. 155), ‘''‘sample size is almost always a matter o f judgement as well as a calculation". Saunders et al. (2003) do not see minimum sample size recommendations as inviolable. Nonetheless, it is important to note that this study did not meet the minimum sample size recommendation for regressions with six independent variables. Tabachnick and Fidell (2001) recommend a minimum sample size of 98 for a regression with six independent variables. Stevens’ (1996) recommendation establishes a lower bound o f 90. For this study, 75 responses met the operationalization criteria for the sample. However, there were only 36 valid responses for the ordinal regression analysis with six regressors.
4 Relaxation of the operationalization criteria of the sample
The original operationalization criteria for the sample outlined in the methodology chapter were relaxed due to concerns about the sample size. The original intention was to limit the study to technology firms located in the XHTZ in the Gaoxin district.
These criteria were relaxed to include firms in all industrial sectors located in any o f X i’an’s districts.
The author constructed the sample frame with a focus on technology companies in the Gaoxin district. The result o f this focus was an over-representation of technology firms located in the Gaoxin district in the sample. Therefore, the sample is not representative of companies located in X i’an.
Only 35 respondents provided information about their annual sales. The author removed three o f these firms from the sample for exceeding the SME revenue threshold (500 million RMB). It was necessary to assume that firms that did not provide information on their annual sales but met the other operationalization criteria were within the SME revenue threshold.
5 Low response rate
There was some concern that a survey o f Chinese SME owners and/or managers would lead to a low response rate. Tse et al. (1995) achieved a response rate o f just 6% for an e-mail-based survey conducted in Hong Kong. This concern was justified: the author’s survey achieved an active response rate of 4.6%. Thietart et al. (2001) warn that high non-response rates can lead to non-response bias, thereby threatening the validity o f the research. Therefore, it was necessary to check whether the sample was representative.
Apart from the over representation o f Gaoxin-based technology firms, the author identified the following issues:
Respondents that scored in the highest quartile o f corporate governance had the highest percentage o f missing responses for the annual sales question. There seems to be a bias, whereby respondents from firms that scored highly on the corporate governance scale were reluctant to disclose information about their annual sales. The cross-tabulation of annual sales against bank credit availability highlights that
10% of the respondents that indicated sales of up to 20 million RMB did not answer the question on bank credit availability.
A cross-tabulation o f education against financial literacy indicates that respondents that scored in the two lower quartiles o f financial literacy were less likely to provide information on their educational background. Furthermore, 10% o f respondents that indicated that they have a Masters degree did not answer the financial literacy questions.
Respondents that scored in the top quartile o f financial literacy were more likely not to answer the loan characteristics questions. Generally, the percentage o f missing responses rises with each quartile of financial literacy.
Only respondents that indicated that they have no difficulty obtaining a bank loan skipped the financial literacy scale: 22.2% o f this group did not answer the financial literacy questions.
6 Non-coverage bias
The construction o f the sampling frame introduced non-coverage bias into the sample. The author used a directory of firms located in the XHTZ published by the X i’an High-Tech Development Authority in 2010. Therefore, it is not a current list o f firms in the XHTZ. Firms that moved into this zone after 2010 would, not have been included in the sample. Furthermore, the contact details for some of the firms listed in the directory were not available.
7 Reliability of the measurement instrument
The internal consistency o f both the corporate governance and the financial literacy scales are below .7, the minimum threshold for establishing scale reliability according to Pallant (2005). The corporate governance scale has a Cronbach’s alpha value of 0.558. George and Mallery (2003) consider such a Cronbach’s alpha value to be poor but not sufficiently so to disqualify the use of the scale for analysis.
The financial literacy scale has a Cronbach’s alpha value of .349. This value is unacceptable for analysis according to the classification o f George and Mallery (2003). This raises questions about the construct validity o f the financial literacy scale, which nonetheless did achieve internal consistency in the pilot study in which participants were from the west.
8 Weak sample monitoring
Weak sample monitoring, a weakness of online administered instruments, means that the author cannot be certain that the survey instrument reached the intended target population. In a limited follow-up survey, using the original sample frame, 73.7 per cent o f the respondents identified as owners and/or managers (see Appendix C).