Analyzing And Mixing Both Quantitative

Published Date: 02 Nov 2017

In quantitative research, an investigator relies on numerical data (Charles & Mertler, 2002). This research analyzes trends, compares groups, and relates variables using statistical analysis and compares the prior predictions and past research (Creswell, 2012). The researcher uses post positivist claims for developing knowledge, such as cause and effect thinking, reduction to specific variables, hypotheses and questions, use of measurement and observation, and the test of theories (Ivankova, 2002). A researcher isolates variables and causally relates them to see and determine the magnitude and frequency of relationships. To add on, a researcher determines which variables to investigate and chooses appropriate instruments, which will yield highly reliable and valid score (Ivankova, 2002).

Alternatively, qualitative research requires exploring the problem and developing detailed understanding of the central phenomenon (Creswell, 2012). It is "an inquiry process of understanding" where the researcher develops a "complex, holistic picture, analyzes words, reports detailed views of informants, and conducts the study in a natural setting" (Creswell, 1998, p. 15). In this approach, the researcher makes knowledge claims based on the constructivist or advocacy/participatory (Mertens, 2003,) perspectives. What is clear is that in qualitative research, data is collected from those immersed in everyday life of the setting in which the study is framed (Ivankova, 2002). Data analysis is based on the values that these participants perceive for their world. Ultimately, it "produces an understanding of the problem based on multiple contextual factors" (Miller, 2000).

4.2 Mixed Theory in Research

In Chapter 1 (1.3), I introduced the Diversified funding model (DFM) which is guiding this research. This model has been applied in higher education to attain effectiveness and vibrancy financing in the sector. The higher education institutions are increasingly relying on a mixed source of financing. The depiction of the mixed (diversified) model identifies five funding sources: the government, donors, entrepreneurship, parents and students (Johnstone, 1986). According to Johnstone, any reduction in one source must result in an increase somewhere else. The 1996, education policy reforms in Zambia were premised on having a mixed source of funding. The policy is based on cost sharing (government, universities and students), revenue diversification (entrepreneurship, donors, private sources) and student loans. This mix of contributors in most cases depends on many factors and usually varied. In the case of Zambia, since tax funding has been shrinking there must be an increase in other funding possibilities. This research investigates the effectiveness and viability of the current financing policy in Zambia’s public universities. In order to realize its intended goal, the study uses a mixed method approach in its methodological perspective. The findings are discussed in the context of three theories already discussed in chapter two: human capital, neo-liberal and quasi public good.

In the mixing stage of my convergent parallel designs, I analyze data for the quantitative and qualitative data separately, and then results were compared using a convergence model where both approaches analyze the same phenomenon as supported by Creswell & Plano Clark (2007). The procedure involved qualifying quantitative data. Data from questionnaires was analyzed through descriptive statistics and factor analysis. The factors produced became the themes for analyzing and comparing with qualitative data.

4.2.1 Convergent Parallel Design

There are six different types of mixed study designs: the convergent parallel design, the explanatory sequential design, exploratory sequential design, the embedded design, the transformative design and the multiphase design (Cresswell & Clark, 2011). This study of policy of financing public universities in Zambia will use the ‘convergent parallel design’, which is a mixed method of design. This design simultaneously collects both quantitative and qualitative data, merge the data, and use the result to understand the problem (Cresswell, 2012). The rationale for this design is that one data collection form offset the weakness of the other form. In this design the researcher gathers both quantitative and qualitative data, analyzes both data sets separately, compares the results from both data sets separately (see Figure 4.1), compares the results from both data sets and make an interpretation as to whether the result support each other or contradict (Cresswell, 2012). This direct comparison of the two data sets by the researcher provides a "convergence" of data sources (Cresswell, 2012).

In mixed methods approach, the researchers try to build the knowledge on pragmatic grounds (Maxcy, 2003) asserting truth is "what works" (Howe, 1988). Most often, researchers chooses the approaches, variables, unit of analysis which is most appropriate for the chosen research questions (Tashakkori & Teddlie, 1998). A major tenet of pragmatism or simply mix method is that both quantitative and qualitative methods are compatible. By implication, it means both numerical and text data collected sequentially or concurrently, should be in a position to help better understand the research problem. While designing a mixed methods study, the following three issues need consideration: priority, implementation, and integration (Hanson, 2003). Creswell (2012) opines ‘Priority’ refers to which method, either quantitative or qualitative, is given more emphasis in the study and ‘Implementation’ refers to whether the quantitative and qualitative data collection and analysis comes in sequence or in chronological stages, one following another, or in parallel or concurrently. And ‘Integration’ refers to the phase in the research process where the mixing or connecting of quantitative and qualitative data occurs (Creswell, 2012).

This study used one of the most popular mixed methods designs in educational research: convergent parallel design in which data was simultaneously collected (Creswell, 2002, 2003; Creswell et al., 2003, Cresswell, 2012). Both the quantitative, numeric, data was collected in the same period with qualitative. The goal of the quantitative phase was to identify potential predictive power of selected variables on policy of financing public universities in Zambia. A qualitative approach was used to collect text data through individual semi-structured interviews, documents, and elicitation materials to help explain key issues regarding the current policy of financing public universities in Zambia. Quantitative data and results provided a general picture of the research problem, i. e., Does the current policy of financing public universities contributed to make universities effective and sustainable? While the qualitative data and its analysis refined and explained these statistical results by exploring participants’ views in more depth. The detailed visual model for convergent parallel mixed method for this study is shown in Figure 4.2.

As shown in Figure 4.2, the priority in this design is given to the quantitative method, because the quantitative research represents the major aspect of data collection and analysis in the study, focusing on detailed variables on policy of financing public universities in Zambia. Qualitative method helped in giving in depth explanation of quantitative analysis by exploring four main independent variables of the current policy. The quantitative and qualitative methods are integrated in the second phase of ‘procedure’ and especially in the last of ‘product’. The results of the two phases were integrated during the discussion of the outcomes of the whole study as followed in chapter six, seven and eight.

4.2.2 Advantages and Limitations of the Design

The strengths and weaknesses of mixed methods designs have been widely discussed in the literature by different academics, which I also experienced when doing this research (Creswell 2012, Creswell, 2002; Haines, 2011; Creswell, Goodchild, & Turner, 1996; Green & Caracelli, 1997; Moghaddam, Walker, & Harre, 2003).

Advantages of this design include:

The design collects both qualitative and quantitative data simultaneously

Both quantitative and qualitative builds on the strength of the other since none of the two might be comprehensive on their own.

Easy to implement for a single researcher if variables and procedures are well defined

One approach may be given more prominence and the other can be complementary

The limitations of this design include:

As any mixed methods design, it requires lengthy time to complete.

It requires ingenuity and feasibility of resources to analyze both type of data

Mixed methods sometimes are confusing to the researcher

4.2.3 Quantitative Phase

The main research questions in the quantitative phase were: How do all categories of respondents (male & female) in general perceive the current policy of financing public universities in terms of cost sharing, revenue diversification and loan scheme? How do student respondents of different ‘universities’ and ‘sponsor’ view the current policy of cost sharing, revenue diversification and loan policy? Can the status of a viable, effective and sustainable financial policy be correctly predicted from different variables of cost-sharing, revenue diversification and student loan policy? Does the current policy of financing public university need to be re-engineered and improved in terms of cost sharing, revenue diversification and student loan scheme? What are the strengths and weaknesses of the current policy of financing public universities? What underlying factors are behind the current policy of financing universities?

Predetermined set of predictor variables from the current policy included: opinion on the current cost sharing policy, perception on the revenue diversification advocated in policy, examining the current policy of student loans and appropriateness of current models (dual track and unit cost). Policy of financing public universities was considered a dependent variable, the outcome or result of the influence of the independent variables (Isaac & Michael, 1981). Selected factors which contribute/ or impede policy of financing public universities have been treated as independent or predictor variables as they generally cause, influence, or affect outcomes. These factors were mostly identified through the analysis of the related literature, theories of financing higher education and also based on the conceptual frame work (Kember, 1994). The survey questions (items) for this study were developed based on the components of the different models of higher education finance discussed in the theoretical perspectives in chapter two of this dissertation (Johnstone, 2008; Barr, 2005)

These factors correspond to the research questions for both Phases (Quantitative and Qualitative) and are the following were considered.

Cost sharing related items: These included among other predictors: good policy option, improve university finance, effectiveness, good implementation, financing own budgets, adjusting current policy, sustainability, re-engineering, encouragement, government involvement and bursaries committee among others (22 items altogether).

Revenue diversification related items: These among other predictors include: self-sustainability, more private sponsored students, re-engineering, alternative sources of revenue, research and consultancy, good policy, change budgeting, operate like corporation and limited government dependency among others (14 items altogether).

Student Loan system related items: Among other predictors includes: loan policy as a good way, never implemented, opposed by students, no clear guidelines, means testing complicated, change is slow, need for a special bank and cost effective among others (12 items altogether).

These variable items were measured on a continuous 5-point Likert-type scale in the questionnaire. For the test to have a statistical power, each variable was represented by at least two items on the scale in the survey instrument. Demographic characteristics, such as gender, age, marital status, institution, specific discipline, program type, mode of payment (only students), terms of employment (only lecturers) were given consideration in this study.

What was the target population and sample for the quantitative part of the study? A population is a group of elements or cases, whether individuals, objects or events, that conform to specific criteria and to which we intend to generalize the results of research (McMillan & Schumacher, 2001; Chiyongo, 2010). Bless and Achola (1988) also agree that a population is the entire set of objects and events or group of people which is the object of research and about which the researcher wants to determine some characteristics. Babbie and Mouton (2004:173), a population is defined as "the theoretically specified aggregation of study elements". White (2005) defines a sample as a group of subjects or situations selected from a larger population. Bless and Achola (1988) define a sample as the sub-set of the whole population which is actually investigated by a researcher and whose characteristics will be generalized to the entire population.

The target population in this study was the students, lecturers, university administrators and senior ministry of education officials. The students came across different public universities, disciplines and both the government and self supported were considered. The student status varied from Bachelor to Master degree pursuing, though majority were bachelor. For the purpose of the quantitative phase of the study the convenience sample (Dillman, 2000) were selected involving students and lecturers, which encompassed four categories of students and three categories of staff, identified as key respondents in policy of public universities finance. Initially 1000 questionnaires were given for students and only 729 (72.9%) were received and completed and for staff 400 were given and only 200 (50%) were returned. As can be observed Table 4.1, the status of the distribution and returns of the questionnaire from surveyed institutions show a combined response rate of 71.5% (N=929). The criteria with students was: (1) those who were admitted and are active in the school program of education, "higher education management" and "political economics of education" (n=200); (2) those who are admitted but were active in social sciences discipline other than education (n=200); (3) any other category including natural science who had been active (n=200), and (4) all the respondents came from any one of the three big public universities in Zambia (University of Zambia, Mulungushi university and Copperbelt university).

With the lecturers, the convenience sample encompassed three categories (1) those who were in the discipline of education or education related (n=100) (2) those in social science related discipline (n=50) and (3) those in natural science and other discipline (n=50).

How was data collected in the quantitative phase? The quantitative phase of the study focused on identifying factors contributing to financing of public universities in Zambia. The comparative survey design, which implies the data was collected at the same time (McMillan, 2000), was used. The primary technique for collecting the quantitative data was a self-developed questionnaire, containing self-assessment items, measured on the 5-point Likert type, and open-ended questions. A panel of professors teaching higher education and economics of education were used to secure the content validity of the survey instrument. The questionnaire consists of sixty-six (66) questions, which are organized into five sections. The first section "section A" constituted the demographic characteristics of respondents which included gender, age, marital status, discipline, level, mode of payment, terms of employment among others. The second section "section B" constituted survey questions on the perception to the current cost sharing policy of financing public universities. The responses were measured 5-point Likert type scale from "Strongly disagree" to "Strongly agree". The third section measured participants’ significant level with the current revenue diversification policy and financing of public universities. A 5-point rating scale from "Not very important" to "Very important" was used. The fourth section focused on student loan system and public university finance and focused on participants’ support levels of this policy. A 5-point rating scale from "Do not support" to "Strongly support" was used. The fifth section surveyed the support levels of the current policy and models of financing public universities among the respondents. The scale from 1 to 5, from "Strongly disagree" to "Strongly agree", was used.

The survey instrument was pilot tested on the 5.0% randomly selected participants representing both the lecturers and students. A random proportionate by group sample of 30 students and 10 lecturers were selected. These participants were excluded from the subsequent major study. The results of the pilot survey helped to ensure instrument elicited appropriate, meaningful and useful data. It also helped to establish internal consistency reliability, face and content validity of the questionnaire. Based on the pilot test results the survey items were revised and in some cases refocused.

The procedure of Data collection was conducted between the months, January 2012 to April 2012. The participants were drawn from four major classifications which included students, faculty, university administrators and ministry of education officials.

For data analysis, prior to the statistical scrutiny of the quantitative survey results, the screening of the data was conducted on the univariate and multivariate levels (Kline, 1998; Tabachnick & Fidell, 2000). Data screening helped in identifying potential multicollinearity in the data, because multivariate tests are sensitive to extremely high correlations among predictor variables (Ivankova, 2002). Normality of Data was a must. Normality tests are used to determine whether a data set has a normal or even distribution or not, or to compute how likely an underlying random variable is to be normally distributed. Practically, as most researches we found some biased questions because of a poor research design, in which respondents tend to rate their experiences on different items of policy of financing public universities in Zambia disproportionately high, low, or moderate. Usually the length of the survey or the situation contributes to either skewed or peak distributions. We also discovered that some respondents just routinely mark boxes or rate a particular experience. According to Hadyuk (1987), explains that a multivariate normal distribution implies that each variable in a sample has a univariate normal distribution and each pair of variables has a bivariate normal distribution. Univariate normality simply explains the distribution of only one variable in the sample, while multivariate normality tends to explain the joint distribution of all variables in the sample. Univariate normal distribution of each variable is very essential, but not a sufficient condition for having a multivariate normal distribution as explained by West et al. (1995)

Outlying cases were excluded from the analysis, as a case that actually is in one category of outcome may show a high probability for being in another category. (Ivankova, 2002). Data screening included the descriptive statistics for all the variables, information about the missing data, linearity and homoscedasticity, normality, multivariate outliers, multicollinearity and singularity. Different questions were analyzed differently as shown below. Since this research uses a convergent parallel design, qualitative data was put in themes which were compared. The results from qualitative data were directly compared with result from quantitative data collection. Statistical trends were supported by qualitative themes and vice versa. As a way of consolidating data, the two approaches were combined to form new variables. The research questions in the study were analyzed following the procedures highlighted below.

What are the social-demographic characteristics of selected participants with respect to gender, age, institution, marital status, sponsor, and affiliation?

Part of preparing and presenting statistical data is to offer a broad picture of a sample representing something specific. A biased sample typically means that inferences can be made only to the particular group represented by the sample. To help readers interpret data, this question begins with offering information about the demographic characteristics of the respondents or simply put respondents’ profile. Data for this question was analyzed through descriptive statistics. The survey items were summarized in the text and reported in tabular form. Frequency analysis was conducted to identify valid percent for responses to all the questions in the survey.

To what extent are perceptions of students & lecturer’s concerning the current financing policy linked to the ‘gender’, ‘institution’ and ‘sponsor’ variables in terms of cost sharing, revenue diversification, student loan and operational models?

To answer this question, a holistic measure of concern of the policy of financing public universities in Zambia was examined on 59 items in the survey. This question helped in giving general description of quantitative data. I wanted to find out the gender variable had any influence on the responses. I also wanted to know if the institution variable had any influence on the respondents since one university had no students enjoying the bursary scheme while the other two did. Did self sponsored students differ in their views to those supported by government? Descriptive statistics for the survey items was summarized in the text and reported in tabular form. Frequency analysis was conducted to identify valid percent for responses to all the questions in the survey. The mean differences of each scenario for the male and female respondents were carried out by one way ANOVA (Analysis of Variance). This was done on both student and lecturer data sets. The level of concern and perception was done between different sex for each of the items under cost sharing, revenue diversification, student loan policy and current operational models. The Means and Standard Deviation of each scenario was given. A sample t-test was performed to compare the mean differences of two sets of respondents to check the level of significance of each.

Two socio-demographic variables of ‘different universities’ (UNZA, CBU & MU) and ‘mode of payment’ (Self, Govt. & other) were explored to ascertain as to whether the institution and mode of payment were significant factors to respondents. The responses were analyzed through comparison of means (M) and Standard Deviation (D). The mean differences of each scenario for the universities and payment mode for student respondents were carried out by one way ANOVA (Analysis of Variance)

How do respondents perceive the current policy on cost sharing, revenue diversification and student loan policy in relation to its viability, effectiveness and sustainability?

Descriptive statistics was carried out on item related to the sustainability of the current policy. Frequency analyses were conducted to identify valid percent for responses to the specific questions related to viability, effectiveness and sustainability of the financing policy. All items related to both students and lecturers were carefully selected and summarized in the comparative bar charts and each scenario was examined closely. Three comparative charts were generated based on cost sharing, revenue diversification and student loan policy. This was compared and supplemented by qualitative data from the interviews. Interviewee’s questions asked how participant perceived the current financing policy in terms of its viability and effectiveness.

Does the current policy of financing public university need to be re-engineered and improved in terms of cost sharing, revenue diversification and student loan scheme?

To effectively answer this question, for quantitative data I used descriptive statistics (Means and Standard Deviation) and also examined all the predictors which specifically asked the respondents on the need for re-engineering the current higher education policy. There were four predictors used: Q15. The current cost sharing policy should be adjusted (re-engineered); Q43. There is need to cultivate and adjust the current policy of revenue diversification (re-engineer); Q50. There are no clear guidelines to the universities to implement the student loan system (need to re-engineer) and Q66. A mixed approach (unit cost and dual track) could be an appropriate way to operate universities (need for re-engineering). Descriptive statistics was central. Individual predictive items were analyzed and then compared on the line graph. This was supplemented by qualitative data obtained from interviewing the university administrators and senior ministry of education officials. Based on their experience about the policy of financing public universities, it was relatively easy for them say whether the current policy needed re-engineering and in what form.

What are the strengths and weaknesses of the current policy of financing public universities?

The question was analyzed by examining descriptive statistics based on lecturer and student respondents. Key predictive items were identified for the analysis. Inspecting the means of each item and standard deviation, it was then easier to identify the current strengths and weaknesses for the financing policy. This was further complemented by qualitative views. This question was better suited in probing the policy makers what they made of the current policy of financing public universities. Quantitative responses were compared to qualitative responses. The views by the two strands of respondents and interviewees were complementary.

What underlying factors (nature of constructs) were deduced for policy of financing public universities in terms of: cost sharing, revenue diversification and student loan?

To answer this question Exploratory Factor Analysis (EFA) was used to analyze the components of the policy based on views collected using the likert scales. Factor analysis is a statistical procedure especially used to basically identify a small number of factors that can be used to represent relationships among sets of interrelated variables (DeCoster, 1998). In other words it is a method which is used to examine how underlying constructs influence the responses on a number of measured variables. We have to know that there are basically two broad types of factor analysis, the exploratory and confirmatory. The exploratory (EFA) attempts to discover the nature of the constructs influencing the set of responses, while the confirmatory (CFA) tests whether a specified set of constructs is influencing responses in a predicted way (DeCoster, 1998). Both are based on the Common Factor Model. For instance, this model proposes that each observed response (measure 1 to 5) is influenced partially by underlying unique factors (factor 1 & 2) and partially by underlying unique factors (E1 to E5) as shown in Figure 4.3. The strength between each factor and measure varies. Other factors may have more influence on the

It is always important to remember that factor analyses are performed by examining the pattern of correlations (or covariances) between the observed measures from the analysis. Measures that are highly correlated (either positively or negatively) are likely to be influenced by the same factors, while those that are relatively uncorrelated are likely influenced by different factors in the analysis. Four steps were crucial in the final analysis of the exploratory factor analysis of the four components of policy of financing public universities in Zambia.

Computation of a correlation matrix for all variables for different aspects

Determination on the number of factors necessary to represent the data and also the method of calculation them (factor extraction)

All factors were transformed so that they become easily interpretable (rotation)

Finally, the scores were computed for each factor

After performing exploratory factor analysis on all the 22 cost sharing items, 7 factors were deduced: "Cost sharing is a good policy option" (Q.8, 9, & 17), "re-engineering of the current cost sharing policy to make it effective and viable" (Q. 20, 13, 16 &10), "government to continue sponsoring students in public universities" (Q. 12, 11 & 18), "negative perception to self sponsorship" (Q. 24, 22 & 23), "need for clear guidelines on current cost sharing policy" (Q.19, 21 & 25), "what government should finance" (Q. 26, 29 & 27) and "what universities should finance" (Q.15 14 & 19).

On the 14 items on revenue diversifications, 5 factors were deduced: "positive perception to alternative revenue through research and consultancy" (Q. 33, 34 & 32), "universities to operate like corporations, sale patents and collaborate" (Q. 41, 40 & 42), "revenue diversification currently is not effective and viable" (Q. 38, 39 & 37), "private sources of revenue are crucial for universities" (Q. 31, 30 & 36) and "need for re-engineering current policy of revenue diversification" (Q. 35 & 43).

After running factor analysis on the on the 12 items related to the current loan policy, 4 factors were identified as crucial namely: "no clear guidelines for the current student loan policy" (Q. 50, 49 & 53), "Zambians not opposed to student loan policy" (Q. 48, 47 &51), "student loan policy are crucial for vulnerable students" (Q. 55, 54, 52, 51) and "government Inertia was the cause of lack of implementation of the loan policy" (Q. 46, 45)

The exploratory factor analysis was supplemented by qualitative data obtained from interviewing university administrators and senior ministry of education officials. With only a few perception differences, the interviewees explained the nature of constructs influencing some of the identified underlying factors.

How effective is financing of higher education (universities) in China?

Document analysis on key papers and government documents on the financing arrangements in China’s higher education were used. The data base from the graduate school of education at Huazhong University of Science and Technology was extremely helpful in tracing the current trends. I had also a lot of input from colleagues and professors in the department. Different education reforms and specific policies of higher education were highlighted. All forms of student aid and support system available in China’s higher education were explored.

What then would be the appropriate model and arrangements for financing public university education in Zambia?

This question was answered in the concluding chapter as an attempt of giving solution to the current nightmare of financing public universities. Using the data obtained in this study, international experience, and Jonstone’s conceptual frameworks, I propose a model; formula and creation of a special fund through an act which if implemented can make universities viable, effective and sustainable.

How was reliability and validity attained? Clearly, in quantitative research, reliability and validity of the instrument are very important for decreasing errors that might arise from measurement problems in the research study (Ivankova, 2002). Reliability refers to the accuracy and precision of a measurement procedure (Cresswell, 2012; Thorndike, 1997). The stability or test-retest reliability of the survey instrument was obtained through the pilot testing of the instrument. Test-retest reliability will show if the same results are obtained with repeated administering of the same survey to the similar study participants (Ivankova, 2002). Reliability of the results was mainly through factor analysis in which the Cronbach’s alpha was used to check reliability of both internal consistency and final result. Equally, internal consistency reliability analysis of the items measured on the Likert-type scale also was conducted on the results of the pilot study. This helped in assessing how well the various items in a measure appear to reflect the attribute, the respondents’ views on re-engineering the policy of financing public universities in Zambia. Inter-item correlation was conducted on the basis of the correlation matrix of items on the scale, corrected item-total correlation, and alpha if an item is deleted. This particular analysis provided information on which items needed rewording or even complete removal from the scale.

Mertler & Vannatta (2010) opine that validity refers to the degree to which a study accurately reflects or assesses the specific concept or construct that the researcher is attempting to measure. Thorndike (1997) content, criterion-related, and construct validity of the survey instrument were established. Content validity showed the extent to which the survey items and the scores from these questions are representative of all the possible questions about policy of financing public universities in Zambia. The wording of the survey items was examined by both professors in HUST and US who are familiar with the field of higher education finance. This helped a lot in assessing whether the survey questions seem relevant to the subject matter it is aimed to measure, and also a reasonable way to gain the needed information, and if generally was well-designed.

Ivankova (2002) points, criterion-related validity, also referred to as instrumental or predictive validity, is used to demonstrate the accuracy of a measure or procedure by comparing it with another measure or procedure, which has been demonstrated to be valid (Mertler & Vannatta, 2010). For this purpose, the self-designed survey questionnaire was compared on the consistency of the results with existing instruments, measuring the same construct, different respondents’ views of financing public universities. In this case, construct validity sought agreement between a theoretical concept and a specific measuring device even a procedure. In this thesis, this was achieved (construct validity) by using factor of the Likert type survey items which was performed, both after the pilot and the major study. The factor loading for survey item, showed the correlation between the item and overall factor (Mertler & Vannatta , 2010)

Generally or ideally, the analysis should produce some kind of simple structure, which is characterized by the following: firstly, each factor should have several variables with strong loadings, secondly, each variable should have a strong loading for only one factor, and thirdly each variable should have a large communality, such as, degree of shared variance (Kim & Mueller, 1978; Ivankova, 2002). Construct validity also helped in addressing the concern of having the results produced by one’s measuring instrument being able to correlate with other related constructs in the expected and planned manner (Carmines & Zeller, 1991).

4.2.4 Qualitative Phase

This phase first considered data collection and respective samples. The qualitative phase of the study involved the purposeful sample, which implies intentionally selecting individuals to learn to understand the central phenomenon (Cresswell, 2012; McMillan & Schumacher, 1994; Miles & Huberman, 1994). The idea is to purposefully select informants, who best answered the research questions and who were "information rich" persons (Patton, 1990). In this sample method, the researcher purposely targets a group of people believed to be reliable for the study. The power of purposive sampling lies in selecting information related to the central issues being studied (Kombo & Tromp 2006). Here the investigator selects the particular units from the population from which relevant samples are drawn.

The qualitative phase in the study mainly was complementary and focused on explaining the results of the statistical tests obtained in the quantitative phase. The primary technique used in conducting in-depth semi-structured interviews with two categories of respondents which included ‘University administrators’ and ‘Ministry of Education HQ officials’ was triangulation. Triangulation of different data sources is important in qualitative analysis as it helps in comparing and contrasting information from different sources (Creswell, 1998; Creswell, 2010). The Interview Protocol included seven open-ended questions. The content of the protocol questions was grounded and similar to those of quantitative. The interviews were premised on getting detailed views on the current policy of financing public universities. More specific question on cost sharing, revenue diversification, student loan and model of financing public universities were asked.

At least 3 participants from each of the three public universities were sampled and another 3 from the ministry of education headquarters. All these participants were in administrative positions either in the university or ministry. A total of 12 participants were sampled. In the survey informed consent form, the participants were informed that three of them were selected from each of the universities. Due to the nature of the convergent parallel design of this study, the selection of the participants was simultaneously done for both qualitative and quantitative. Details of interviews conducted are contained in Table 4.2

The protocol was pilot tested among three officials; two from the universities and one from the Ministry of education from the same targeted population, but three were excluded from the study. Debriefing with the participants was conducted to obtain information on the clarity of the interview questions and the relevance to the study aim. Most participants, received interview questions prior to the scheduled interview and a few opted to be interviewed without prior knowledge of questions. Participants were also informed that the interview was tape recorded and transcribed verbatim. Respondents were further reminded that they will have an opportunity to review, and if necessary, correct the contents of the interviews after it had been transcribed. This was clearly communicated to all selected respondents at University of Zambia, Mulungushi University, Copperbelt University and Ministry of Education.

Why did the study also employ document analysis? Without doubt, document analysis is considered as one of the important approaches of getting information especially in the qualitative investigation. Documents captured for this study includes annual reports, government expenditures based on the yellow book, annual education plans, publication from the ministry of education policy document, universities strategic plan and some memoranda from meetings. In this research of re-engineering financing policy to make public universities sustainable, a myriad of documents, reports and all important information concerning the financing of higher education in Zambia were obtained. Special focus was given to government policy papers on higher education, especially looking at policies pertaining to cost sharing, revenue diversification and student loan. Patterns of financing public universities were also confirmed and coupled with major challenges especially in the last decade (Chapter three reviews most documents on evolution and current status).

In addition, policy papers, reports and publications from directorate of planning in the ministry of education and universities’ management accountant and bursars were also consulted and extensively incorporated in the study as well. In doing this I considered how the policy of financing higher education impacted on the direction of higher education in Zambia. An attempt was made to use reports and policy recommendation from World Bank and International Monetary Fund on higher education in Zambia. Moreover, public universities’ reports and budget papers were reviewed to make the study more comprehensive. Finally, comprehensive examination of publications and reports and other important documents from the Ministry of Education were critically considered and integrated into the study. For easy facilitation, I contacted the appropriate offices such as the Ministry of Education and universities, to access all the necessary documents for the study. The access and understanding of these documents improves the richness of information for qualitative study.

Patton (2002, p. 295) contend "learning to use, study, and understand documents and files is part of repertoire of skills needed for qualitative inquiry" In the same line, Guba & Lincoln (1981) points to certain reasons why documents and records need to be integrated into a naturalistic investigation: (a) Documents and records are stable and rich resources that serve as the basis for investigation into a phenomenon; (b) they serve as checks and balances on the truthfulness or falsity of a statement, thereby checking misrepresentation and libel during the investigation; (c) they provide information about events that exist and arise from the context of the investigating phenomenon, hence they serve as a primary purpose; (d) access to documents and records especially in the public record is easier and even sometime at no charge to the researchers; and (e) they are non-reactive, - available for the sharing of knowledge. These document serves as checks and balances to the truth and falsity of information obtained from interviews (Guba & Lincoln, 1981)

The thesis now concretizes data analysis procedure under the qualitative phase. As the common practice, in the qualitative analysis, data collection and analysis proceed simultaneously (Merriam, 1998; Miles & Huberman, 1994). In the qualitative phase of the study, the text and image data obtained through the interviews, documents and elicitation materials were coded and analyzed following the particular question asked to interviewees. The following steps were strictly followed: Firstly, the preliminary exploration of the data by reading through the transcripts and writing memos; secondly, coding the data by segmenting and labeling the text; thirdly, using codes to develop themes by aggregating similar codes together (See Figure 4.4); fourthly, connecting and interrelating themes; and lastly constructing a narrative (Creswell, 2012). To augment the further discussion, the visual data displays were created so as to show the evolving conceptual framework of the factors and relationships in the data (Miles & Huberman, 1994).

One of the biggest challenges of qualitative research is synthesis of data. The volumes amounts of interviews and the quality of information acquired could make qualitative data analysis threatening, difficult and sometimes overwhelming. There is therefore need for a complete reduction of information, to makes the data more manageable and meaningful for easy interpretation. Data reduction according to Miles and Huberman framework (1994) "… refers to the process of selecting, focusing, simplifying, abstracting, and transforming the data that appear in written up field notes or transcriptions" (p.10). The main purpose of this is to make the data manageable, well condensed and therefore requires a careful process of deduction and inductions there by taking into account the research questions framed and emerging ideas and themes in the study. The information which was obtained from university administrators and from the ministry of education officials was subjected to data reduction.

Data analysis involved developing a detailed description of each of the parts of policy of financing public universities, beginning with cost sharing, revenue diversification, student loan and current models. Descriptions and themes related to specific activities earlier framed. Based on this analysis, a researcher provided a detailed narration of each of the situation and variable. In other words, since this research is using a mixed method, qualitative data was coded, codes were assigned and the number of times the code appeared was recorded as numeric data. This makes it easier to compare to the quantitative datasets (Green, 2007; Cresswell, 2012).

In convergent study design, the analysis is performed at two levels: within each approach and across the both approaches (Stake, 1995). Then, analysis of this data is done holistically focusing on specific identified variables. The researcher then had an opportunity to report the meaning of embedded variables and generally also lessons learned. The objectives of the research were therefore answered. It is at this point that an accurate picture which appraises the current policy of financing public universities was comprehensively realized.

These following questions were considered when soliciting data from public university administrators and ministry of education officials from the Headquarters.

Is the current cost sharing policy in public universities viable, effective and sustainable?

The current policy of financing public universities emphasizes revenue diversification, how effective has this policy been? Which ventures are you involved in?

In policy, support for students in public universities should be through student loan system. Despite the existence of this policy for 15 years and more, it has never been implemented. Why is it so? Is this policy appropriate for public universities?

What are the strengths and weaknesses of the current policy of financing public universities?

Does the current policy of financing public university need to be re-engineered and improved in terms of cost sharing, revenue diversification and student loan scheme?

Which model/approach is appropriate for funding public universities, Unit cost (run it like a business), Dual track (Some pay and some supported) or Mixed approach? Do you have other suggestions?

Establishing credibility is very essential in qualitative surveys. Here, the criteria for judging a qualitative study differ from quantitative research. In qualitative design, the researcher seeks believability, based on coherence, insight, and instrumental utility (Eisner, 1991) and trustworthiness (Lincoln & Guba, 1985) through a process of verification rather than through traditional validity and reliability measures (Chiyongo, 2010). The uniqueness of the qualitative investigation within a specific context precludes its being exactly replicated in another context (Atuahene, 2006). Even though, statements about the researcher’s positions – the central assumptions, the selection of informants, the biases and values of the researcher – enhance the study’s chances of being replicated in another setting (Creswell, 2003; Creswell, 2012).

Different scholars validate the findings, such as, determine the credibility of the information and whether it matches reality (Merriam, 1988). The four primary forms are used in the qualitative phase of the study: in the fore is triangulation – converging different sources of information (interviews, documents, artifacts); then, member checking – getting the feedback from the participants on the accuracy of the identified categories and themes; followed by, providing rich, thick description to convey the findings; and lastly external audit –asking a person outside the project to conduct a thorough review of the study and report back (Creswell, 2003; Creswell & Miller, 2002; Creswell & Clark, 2011).