Methodology and philsograp of methodology
This thesis is a quantitative research, which based on philosophy view of ‘empiricism’
(Leach,1990) and ‘Positivism’(Duffy,1985). Empiricism, in philosophy, is defined as
the view that all theories originate in experience, that all concepts are about or
applicable to issues that can be experienced, or that all rationally acceptable beliefs or
propositions are justifiable or knowable via experience (Leach, 1990). Positivism,
emphasises empirical data and scientific methods. This philosophy perspective holds
the perspective of regularities establish the world. These regularities are detectable
and conceptualised, and, thus, that the researcher can infer knowledge about the real
world by interpreting and investigating it (Duffy,1985).
Based on ‘empiricism’ and ‘Positivism’, a research employing the quantitative
research can presents an objective, formal, systematic process with employing
numerical data to quantify or measure phenomena and produce findings it describes,
tests and examines cause and effect relationships (Burns and Grove,1987) Moreover,
by employing legitimate quantitative data, which is collected rigorously by applying
the scientific methods and analysing critically, can enhance its objectivity, validity
and reliability (ACAPS, 2012: 6).
28
1.9.2 The comparative method
In this thesis, the main method applied throughout the three empirical studies is the
comparative analysis between the self-employed and wage paid employees. The
comparison is a common research method with outstanding merits and widespread
application, which plays a vital part in the most diverse branches of the humanities
and social sciences alike. Firstly, the comparative approach is a mode of scientific
analysis that sets out to investigate systematically two or more entities concerning
their similarities and differences, to arrive at understanding, explanation and further
conclusions (Azarian,2011). Secondly, the comparative analysis is worthwhile. By
considering subjects, social actions and events under other contexts, the comparative
analysis helps us to better understand the often taken-for-granted basis of our practices
and phenomena. Moreover, the results generated by comparative study approve the
significance of various methods of organising a society’s issues to develop their
efficiency, it also enables us to ‘reflect upon our social systems and cultural ways of
behaving’(May, 2011:249).Thirdly, Comparison detects the potential of revealing and
challenging our less evident hypotheses and conceptions about the world. In light of
this view, the comparison of the phenomenon will allow us to detect the divergent
formations of the phenomenon and investigate why some have processed in similar
ways while others are different ways (Azarian,2011). Lastly, a comparative approach
can not only describe differences and similarities and development of typologies but
also can be used to extract insights about the causal relationships responsible for the
29
observed similarities and differences(May 2011). In other words, the comparative
analysis not only helps identify the different actual or possible paths that social
processes may take but also help develop a causal theory that can explain the
phenomenon. Consequently, by considering the noted advantages of the comparative
method above, in this thesis, it is appropriate to apply this method to highlight the
particularity of the self-employed regard with the workplace wellbeing by conducting
a comparative analysis with employees. It may also reveal causal generalisations
between occupation selection and workplace wellbeing to gather a deeper
understanding of the issues of the workplace wellbeing and the self-employment.
1.9.3 The matching approach
‘What would happen if I had not chosen to be entrepreneurs?’ To answer this kind of
question, one must consider counterfactually. The main problem is that if individuals
chooses to be entrepreneurs, then there is no data on exactly what would have happened
had they not decided to be entrepreneurs. Recently, Schjoedt and Shaver (2007) cast doubt
on previous results by difference-of-means tests relating group averages for the self-
employed and employee(without controlling for other influence). Schjoedt and Shaver
(2007) argued that the methodology of difference-of-means tests may be flawed, because
self-employed individuals differ from other individuals in many ways, and these
differences between the different occupational groups must be controlled. Otherwise, the
selection bias will be produced and will mislead the results. The reason can be statistically
explained by following the common framework set out by Rubin
30
(1974), which uses binary variable Ti {0,1} to represent the occupation groups. If Ti =
1 then subject i are the treatment group individuals, who are the target subject to study
in the research (who are the self-employed in this thesis), that is, the subject is
‘treated’. If Ti = 0 then the subject i are control group individuals, who are employed
to compare with treatment group (who are employees in this thesis). The estimated
outcome of differences between the treatment group and control group on outcome
variable Y in the group of treated subjects (ATT (Average Treatment Effect on
Treated)) can be estimated as
Where refers to the possible outcome of treated subjects without
intervention. However, in practice, such output cannot be obtained because we know only
one outcome after intervention (the actual outcome). As such, both options are not
possible at the same time. Intuitive substituting of by non-participants
is likely to produce selection bias when condition
does not hold. Treatment individuals and control
individuals would have different outcomes even without intervention as a result of
observable and unobservable factors (Caliendo and Kopeinig, 2008). However, this
selection bias can be overcome by employing Propensity Score Matching (PSM)
techniques.
Rosenbaum and Rubin (1985) proposed Propensity Score Matching (PSM) to resolve
the selection bias problem as it can reduce multi-dimension matching to only
31
one-dimension matching. PSM is based on the assumption that sample selection bias
can be eliminated by conditioning on observable variables, and does so by matching
each treatment subject(the self-employed) with one or more control
subject(employees) with similar observable characteristics. In essence, matching
models simulate the conditions of an experiment in which treatment individual(the
self-employed) and control individuals(employees) are randomly assigned, allowing
for the identification of a causal link between the career choice and outcome variables.
Statically, PSM consists of four steps: Firstly, logistic regression is conducted to
obtain propensity score by employing predicted probability (p) or log[p/(1 − p)]. The
dependent variable Y = 1, if it is treatment individual; Y = 0, control individuals.
Secondly, check the propensity score is balanced or not between treatment and
comparison groups, and check that covariates are balanced or unbalanced between the
treatment and comparison groups by applying standardised differences or graphs to
examine distributions. Thirdly, matching each participant to one or more
nonparticipants on propensity score by the various statistical method. In this thesis,
the nearest neighbour matching method is employed to produce the balanced data at
the ratio 1:1 of the size of the self-employed to employees. Finally, Verifying that all
the covariates are balanced across treatment and comparison groups in the matched or
weighted sample and continue the other statistic analysis with the new sample.
Matching estimators are preferable because more care is taken to establish an appropriate
control group when the updated sample needs to be used by other statistical
32
methods, like regressions. This is because the researcher is presumably interested in
comparing individuals that have the same values for all covariates, multivariate regression
modelling obscures information on the distribution of covariates in the treatment versus
control groups. Unless there is substantial overlap in the two covariate distributions,
multivariate regression estimates rely heavily on extrapolation, and can, therefore, be
misleading (Ichino et al., 2008). Another advantage of matching method is that it requires no
assumptions on functional forms (Hussinger, 2008). Consequently, the core concept of the
matching theory is that, when examining treatment effect, the treatment sample(e.g., the self-
employed group) should have similar characteristics as those of the controlled sample (e.g.,
employees). Within social science comparison study, other features of observance in two
groups need to be roughly the same to make sure the sample is randomly determined or is
exogenously given (Rubin,1973).
By taking advantage of PSM, the number of researchers utilising the Matching approach
increase continually within the management and economic area. For example, Persson
(2001) used this method to test the effect of joining currency unions on trade growth of
countries. Hutchison (2004) applied the matching approach to investigate the effect of
IMF program participation on output growth. Hofler et al. (2004) using PSM to control
for selection bias problem to study the relationship between institutional ownership and
dividend payout behaviour of the firm. Thus, this approach will be employed into our
updated sample establishment.
Reference
Baron, R. A. (2012). Entrepreneurship: An evidence-based guide. Cheltenham,UK:
Edward Elgar Publishing.
Caliendo, M., and Kopeinig, S. (2008). Some practical guidance for the implementation of
propensity score matching. Journal of Economic Surveys, 22(1), 31-72.
Felstead, A., Gallie, D., and Green, F. (Eds.). (2015). Unequal Britain at Work.
Oxford,UK: Oxford University Press.
Guo, S., and Fraser, M. W. (2010). Propensity score analysis: Statistical methods and
analysis. London,UK: SAGE Publications
Hamilton, B. H. (2000). Does entrepreneurship pay? An empirical analysis of the returns
to self-employment. Journal of Political Economy, 108(3), 604-631.
Joffe, M. M., and Rosenbaum, P. R. (1999). Invited commentary: propensity scores.
American Journal of Epidemiology, 150(4), 327-333.
Lehrer Paul, M., Woolfolk Robert, L., and Sime Wesley, E. (2007). Principles and
Practice of Stress Management. New York, USA: The Guilford Press
Luthans, F., Avolio, B. J., Avey, J. B., and Norman, S. M. (2007). Positive psychological
capital: Measurement and relationship with performance and satisfaction. Personnel
Psychology, 60(3), 541-572.
Rindova, V., Barry, D., and Ketchen, D. J. (2009). Entrepreneuring as emancipation.
Academy of Management Review, 34(3), 477-491.
Rosenbaum, P. R., and Rubin, D. B. (1985). Constructing a control group using
multivariate matched sampling methods that incorporate the propensity score. The
American Statistician, 39(1), 33-38.
Taris, T. W. (2006). Bricks without clay: On urban myths in occupational health
psychology. Work and Stress, 20, 99-104
Uy, M. A., Foo, M. D., and Song, Z. (2013). Joint effects of prior start-up experience and
coping strategies on entrepreneurs’ psychological wellbeing. Journal of Business
Venturing, 28(5), 583-597.
Wiklund, J., Graham, C., Foo, M. D., Bradley, S. W., Shir, N., and Nikolaev, B. (2017).
Entrepreneurship and Wellbeing. Journal of Business Venturing. Call for papers:
https://www.journals.elsevier.com/journal-of-business-venturing/call-for-papers/entre
preneurship-wellbeing . Accessed: 20 SEP 2017
