The hypothetico‐deductive method requires that hypotheses are falsifiable: they must be written in such a way that other researchers can show them to be false. For this reason, hypotheses are sometimes accompanied by null hypotheses. A null hypothesis (H0) is a hypothesis set up to be rejected in order to support an alternate hypoth- esis, labeled HA. When used, the null hypothesis is presumed true until statistical evidence, in the form of a hypothesis test, indicates otherwise. For instance, the null hypothesis may state that advertising does not affect sales, or that women and men buy equal amounts of shoes. In more general terms, the null hypothesis may state that the correlation between two variables is equal to zero or that the difference in the means of two groups in the population is equal to zero (or some other definite number). Typically, the null statement is expressed in terms of there being no (significant) relationship between two variables or no (significant) difference between two groups.
The alternate hypothesis, which is the opposite of the null, is a statement expressing a relationship between two variables or indicating differences between groups.
To explain further, in setting up the null hypothesis, we are stating that there is no difference between what we might find in the population characteristics (i.e., the total group we are interested in knowing something about) and the sample we are studying (i.e., a limited number representative of the total population or group that we have chosen to study). Since we do not know the true state of affairs in the population, all we can do is to draw inferences based on what we find in our sample. What we imply through the null hypothesis is that any differ- ences found between two sample groups or any relationships found between two variables based on our sample are simply due to random sampling fluctuations and not due to any “true” differences between the two popula- tion groups (say, men and women), or relationships between two variables (say, sales and profits). The null hypothesis is thus formulated so that it can be tested for possible rejection. If we reject the null hypothesis, then all permissible alternate hypotheses relating to the particular relationship tested could be supported. It is the theory that allows us to have faith in the alternate hypothesis that is generated in the particular research investi- gation. This is one more reason why the theoretical framework should be grounded on sound, defendable logic to start with. Otherwise, other researchers are likely to refute and postulate other defensible explanations through different alternate hypotheses.
The null hypothesis in respect of group differences stated in the example “Women are more motivated than men” would be:
H0: M W or
H0: M W 0
where H0 represents the null hypothesis, μM is the mean motivational level of the men, and μW is the mean motivational level of the women.
The alternate for the above example would statistically be set as follows:
HA: M W which is the same as
HA: W M
where HA represents the alternate hypothesis and μM and μW are the mean motivation levels of men and women, respectively.
For the nondirectional hypothesis of mean group differences in work ethic values in the example “There is a difference between the work ethic values of American and Asian employees,” the null hypothesis would be:
H0: AM AS or
H0: AM AS 0
where H0 represents the null hypothesis, μAM is the mean work ethic value of Americans and μAS is the mean work ethic value of Asians.
The alternate hypothesis for the above example would statistically be set as:
HA: AM AS
where HA represents the alternate hypothesis and μAM and μAS are the mean work ethic values of Americans and Asians, respectively.
The null hypothesis for the relationship between the two variables in the example “The greater the stress experienced in the job, the lower the job satisfaction of employees,” would be H0: There is no relationship between stress experienced on the job and the job satisfaction of employees. This would be statistically expressed by:
H0: 0
where ρ represents the correlation between stress and job satisfaction, which in this case is equal to 0 (i.e., no correlation).
The alternate hypothesis for the above null, which has been expressed directionally, can be statistically expressed as:
HA: 0 The correlation is negative.
For the example “There is a relationship between age and job satisfaction,” which has been stated nondirec- tionally, the null hypothesis would be statistically expressed as:
H0: 0
whereas the alternate hypothesis would be expressed as:
HA: 0
Having formulated the null and alternate hypotheses, the appropriate statistical tests (t‐tests, F‐tests) can then be applied, which indicate whether or not support has been found for the alternate hypothesis – that is, that there is a significant difference between groups or that there is a significant relationship between variables, as hypothesized.
The steps to be followed in hypothesis testing are:
1. State the null and the alternate hypotheses.
2. Choose the appropriate statistical test depending on whether the data collected are parametric or nonparametric.
3. Determine the level of significance desired (p = 0.05, or more, or less).
4. See if the output results from computer analysis indicate that the significance level is met. If, as in the case of Pearson correlation analysis in Excel software, the significance level is not indicated in the print- out, look up the critical values that define the regions of acceptance on the appropriate table (i.e., (t, F, χ2) – see the statistical tables at the end of this book). This critical value demarcates the region of rejec- tion from that of acceptance of the null hypothesis. When the resultant value is larger than the critical value, the null hypothesis is rejected, and the alternate accepted. If the calculated value is less than the critical value, the null is accepted and the alternate rejected.
Note that null hypotheses are rarely presented in research reports or journal articles.
Now do Exercise 5.11, Exercise 5.12, and Exercise 5.13.
EXERCISE 5.11
Create a diagram to illustrate the relationships between the relevant variables in Exercise 5.9 and develop five different hypotheses.
EXERCISE 5.12
A production manager is concerned about the low output levels of his employees. The articles that he has read on job performance frequently mention four variables as being important to job performance:
(1) skills required for the job, (2) rewards, (3) motivation, and (4) satisfaction. In several of the articles it was also indicated that only if the rewards were (attractive) to the recipients did motivation, satisfaction, and job performance increase, not otherwise. Given this situation:
1. Define the problem.
2. Create a diagram.
3. Develop at least six hypotheses.
EXERCISE 5.13
A recent study has investigated the effect of corporate social responsibility (CSR) on the market value of the firm. This study developed and tested a conceptual framework, which posits that (1) customer satis- faction mediates the relationship between CSR and the market value of the firm, and (2) two firm factors (“innovativeness capability” and “product quality”) moderate the relationship between CSR and cus- tomer satisfaction. For this situation, define the problem, draw a schematic diagram, and formulate the hypotheses.
Hypothesis testing is strongly associated with designing experiments and the collection of quantitative data.
However, as exemplified by Box 5.1, hypotheses can also be tested with qualitative data.
BOX 5.1
HYPOTHESIS TESTING WITH QUALITATIVE RESEARCH: NEGATIVE CASE ANALYSIS
Hypotheses can also be tested with qualitative data. For example, let us say that, aft er extensive interviews, a researcher has developed the theoretical framework that unethical practices by employees are a function of their inability to discriminate between right and wrong, or due to a dire need for more money, or the organization ’ s indiff erence to such practices. To test the hypothesis that these three factors are the primary ones that infl uence unethical practices, the researcher should look for data to refute the hypothesis. When even a single case does not support the hypothesis, the theory needs revision. Let us say that the researcher fi nds one case where an individual is deliberately engaged in the unethical practice of accepting kickbacks (despite the fact that he is knowledgeable enough to discriminate right from wrong, is not in need of money, and knows that the organization will not be indiff erent to his behavior), simply because he wants to
“get back” at the system, which “will not listen to his advice.” Th is new discovery, through disconfi rmation of the original hypothesis, known as the negative case method , enables the researcher to revise the theory and the hypothesis until such time as the theory becomes robust.
BOX 5 1
We have thus far seen how a critical literature review is done, theoretical frameworks are formulated, and hypotheses developed. Let us now illustrate this logical sequence through a small example where a researcher wants to examine the organizational factors influencing women ’ s progress to top management positions. The literature review and the number of variables are deliberately kept small, since the purpose is merely to illustrate how a theoretical framework is developed from the literature review, and how hypotheses are developed based on the theoretical framework.