APPENDIX B
A Brief Overview of Organisational Research
Introduction
In Chapter 1 of the book it is pointed out that OB and OA both draw on concepts and theories from the social sciences and both are research orientated subjects. While this does not mean that readers of the book need to be experienced researchers, in order to grasp the strengths and weaknesses of theories it can sometimes be useful to have an elementary knowledge of the research process and in particular, the limitations of research in the social sciences. With this in mind, this appendix aims to provide the reader with some insights into these matters. It must be stressed however, that what is given makes no claim to being an extensive treatment of a very complex subject, with an extensive body of knowledge. Rather, it is what the title indicates – a brief overview – and for those wishing to delve deeper, further readings are given at the end of the appendix.
Philosophy and Method in Organisational Research
Organisational research is largely the province of the social sciences, and to many people social science is somewhat woolly and imprecise because it seldom produces definitive laws in the same way that physical science does. Nevertheless, it is still capable of explaining a great deal of behaviour in organisations.
However, there are limits to its explanatory powers, and to understand this it is important to explain the nature of social research. The easiest way to do this is to start by exploring the implications of a rudimentary example from physical science. The one used here is Ohm’s law, which expresses the relationship between voltage, current and resistance in an electrical circuit thus:
V = IR where:
V represents Voltage
I represents Current in Amperes
R represents Resistance in Ohnms
This expresses a precise relationship, and if two of the factors are known, the third can easily be determined by transforming the equation thus:
I = VandR = V
R I
In any science, whether it is social or physical, research aims to deliver knowledge at one of three levels:
1Descriptive,where the primary objective is to learn more about some phenomenon, for example, to answer the question ‘What is happening?’
2Explanation,where the objective goes one stage further and seeks to answer the question ‘How and why does this happen?’
3Prediction,where the objective is to produce an irrefutable law, which specifies that if X happens, Y will follow.
Before going further, it should be noted that the relationship between explanation and prediction has long been a matter of debate in the philosophy of science. Some people assert that if a phenomenon can be explained it can also be predicted, but strictly speaking this is not true and the critical issue is one of control. Prediction implies the ability to control events, and simply because we can explain how and why something occurred does not mean that we can predict and replicate its occurrence again.
Ohm’s law is a theory at the highest (predictive) level, so much so that its use is fundamental in the design of electrical circuitry. However, it is immensely difficult, if not impossible to attain this predictive level in social research. To illustrate the point, consider what appears at first sight to be a rather straightforward organisational issue.
Imagine that an organisation with 300 employees observes that it has an overall level of 5 per cent absenteeism and asks you or I to look into the matter. Absenteeism is usually taken to mean unauthorised absence, and so the first task is to define how the concept will be used. The conventional definition is:
all absence other than for reasons of sickness certified by a medical practitioner, or occasioned by use of annual leave entitlement.
This poses an immediate problem. It assumes that when a person telephones in to report sick, unless the sickness lasts long enough to obtain a doctor’s note there is no sickness, the person is simply absent without authorisation. This is clearly somewhat unfair, and while we could exclude these people from the count, if we do so we run the risk of not including genuine absentees; that is, people who are perfectly healthy and just want a day off. There is no easy way out of this, and so the usual definition would probably have to be used. Note, however, that this means that the measure of absenteeism is questionable from the start.
Now assume that there are three distinct groups of employees in the firm: 200 semi-skilled production workers, 60 skilled craftsmen and 40 office workers. It is noted that absenteeism varies for the three groups as shown in Table B1.
Table B1: Absenteeism by Employee Group
Production Skilled Office
Workerscraftsmenworkers
Number(200)(60)(40)
% absenteeism 6.0 3.3 2.5
This is a descriptive level finding and if this is all that is required, since it explains ‘what is there’, matters need go no further. However, it also contains the hint of an explanation. For example, we could reason that absenteeism might be connected in some way with two factors: first, the type of work done by the three different groups of employees; or second, their personal characteristics. Thus, there is a possibility of delving deeper to produce a plausible explanation of the variation in absence rates.
Using work characteristics as a starting point, we then note that production workers have very boring, routine and fatiguing jobs; skilled craftsmen have more interesting, but nevertheless tiring work; and office workers complete tasks that are varied, interesting and not at all fatiguing. From this it could be reasoned that job interest and fatigue are both factors that contribute to some overall concept which reflects job satisfaction. If it can be measured in a satisfactory way it seems possible that the study can be advanced to give findings at an explanatorylevel.Accordingly, a questionnaire is designed to measure employee job satisfaction. This is no easy matter and there are problems inherent in using this method of collecting information. However, assume that we go ahead with the study and the results shown in Table B2 are obtained.
Table B2: Employee Job Satisfaction
ProductionSkilled Office
Workerscraftsmenworkers
Number(200)(60)(40)
% absenteeism6.0 3.3 2.5
Mean job satisfaction
score for group 7.0 10.0 12.0
(max obtainable = 20)
Range of scores3-15 3-15 3-15
Clearly there seems to be some sort of relationship between average levels of job satisfaction and absenteeism for the three groups. However, this explanation is very crude. It uses the average level of job satisfaction in each department and because in each case the range is very wide, there must be some individuals who are satisfied and some who are not.
It is individuals who take unauthorised absence, though clearly not all do and to make the explanation more plausible, we would need to match job satisfaction scores with individual absence records, and this is where the problems start to magnify. Individuals would need to indicate their identities on questionnaires, which removes all anonymity. They might ask themselves why the information is required and protect themselves by giving untruthful answers. Another problem is our assumption that fatigue and job interest are the factors that influence job satisfaction. There could be many other things that people find satisfying or dissatisfying about work. Therefore, while the results approximately satisfy the ‘how or why does this happen’ criterion for explanatory level work, the explanation is nowhere near as precise as in the example of Ohm’s law given earlier.
If the aim was to take the study of absenteeism to a predictivelevel, the problems would be even more formidable, and probably not capable of being overcome. If all the problems of measuring absenteeism and job satisfaction could be resolved, theoretically we could conduct a well-designed experiment. For example, working on the assumption that low satisfaction results in absenteeism, we could redesign the work to make it more satisfying and then observe the result. However, employee conceptions of what makes a job satisfying are likely to be highly individual, which means that, if satisfaction is to be increased for all of them, redesign is a difficult matter. There is also the problem that any change we make could have a novelty value. It might temporarily make the job more interesting, but the effect could wear off after a while. Perhaps the most difficult problem of all is that when a change is made people inevitably put their own interpretations on why it has occurred. Unlike inanimate objects, people are not passive recipients of change. They generally impute some reason for it, and respond accordingly. Since their thoughts cannot be observed directly, we once again encounter the problem of not really knowing why behaviour has changed. Thus, even if absenteeism falls, employees might have some other reason for attending regularly, one that has nothing to do with the redesigned jobs. All of this means that it is virtually impossible to produce social science theories that have the same predictive validity as those in the physical sciences.
Here the reader should be aware that in social science there are two strongly competing views about the whole matter of studying and explaining human behaviour. What can broadly be classified as the positivist perspective holds that there is an objective reality about the social world, which is there to be uncovered. Therefore, organisations can be investigated in a detached way, by objective methods that largely follow the research philosophy of the natural sciences, as in the example above. The opposing view is that of phenomenology, which asserts that in the social world there is no such thing as an objective reality. People interpret their surroundings and construct their own meanings from what they perceive to be there, and respond accordingly (Berger and Luckmann 1966). Therefore there can be no independent, objective reality, only the reality that people construct for themselves, which means that in trying to explain their behaviour account must be taken of the way they experience matters. Clearly an approach such as this can be frustrating to those looking for infallible, off-the-shelf solutions to a problem, but it also has its bright side. It is the huge amount of variation between people that gives them their individuality and it is this individuality that makes people fascinating.
Methods of Social Research
Social science uses a wide range of different research designs to conduct investigations and an equally large number of methods of collecting information. Each has its own strengths and weaknesses and there is no such thing as a perfect design or perfect data collection method. In general terms the design selected should flow from the questions addressed by the research and the methods used to collect data should be those which have the greatest practical utility in obtaining the information required. Space precludes giving more than a brief overview of the wide variety of methods, and in any event the aim here is to give the reader an outline, if somewhat cursory awareness of all the methods, rather than a detailed knowledge of any single one. For those interested in exploring matters more deeply further readings are suggested at the end of the chapter.
Research Designs
In social science there are three main types of research design: survey designs, experimental designs and ethnographic methods (case studies), each with a number of variants. Since research design is a topic in its own right, and the aim here is simply to give an outline of the alternatives, discussion will be confined to these three primary types.
Survey Designs
The surveydesign is probably the one that most closely matches the public perception of social research. Because surveys are essentially cross-sectional they are useful for identifying characteristics of individuals, groups or organisations at a point in time, or different points in time if longitudinal methods are used. However, unless all of a population is to be surveyed, it is necessary to take steps to try to ensure that the sample chosen is representative of the whole population. For example, to draw conclusions about an organisation it would be desirable to have appropriate percentages of males and females, managers and non-managers and people drawn from different departments, in order to reflect the composition of the whole workforce. A wide variety of data collection methods can be used in surveys, such as interviews, questionnaires and observations.
Experimental Designs
The word experimental conjures up a picture of laboratory work and in psychology many investigations use laboratoryexperiments. The great advantage of this approach is the degree of control that can be exercised over the variables at work, which means that a more precise evaluation of the effect of one variable on another can be obtained. For example, suppose that we wished to investigate the effects of background noise on the capability of people to perform a complex task accurately. We could start by stating a simple hypothesis that ‘accuracy falls as the level of distracting background noise rises’. Therefore, the independentvariable, the one that is assumed to influence the level of accuracy, is noise level, and we could measure this in decibels, using a sound meter. The dependentvariable, accuracy, would probably be measured as the number of mistakes made. In order to have confidence in the results, it would be necessary to run a large number of tests, and the accuracy of people would probably improve as they became more familiar with the task. For this reason, it is usual to have two groups of people from whom readings are obtained: a controlgroup, where background noise is kept constant, and an experimentalgroup, for whom noise is varied. The basic design is as shown in B3.
Figure B3: Basic Experimental Design
MeasurementTreatmentMeasurement
1 2
Control groupAccuracy at NoneAccuracy at
background noisebackground noise
level ‘A’level ‘A’ after 30
minutes at task
Experimental groupThe sameApply higherAccuracy at noise
backgroundlevel ‘B’ after 30
noise level ‘B’minutes at task
Measurement 1 gives a baseline measure that tells us whether the two groups differ in accuracy at the start, and from measurement 2 we can evaluate the extent to which accuracy is reduced by background noise but, at the same time, increases with learning. From the measurement 2 for the control group we can obtain an indication of how learning effects accuracy. Therefore, if we assume that the accuracy of the experimental group is effected by learning by the same amount, we could obtain a picture of how much performance has decreased due to the effects of the distracting background noise.
This is a very simple example and is merely used to illustrate the features of the design. In a well-designed experiment conducted in a laboratory setting many other precautions would be taken to control the conditions, so that the possible effects of other extraneous variables that could affect accuracy are minimised. For example, temperature and lighting would probably be kept constant. Thus there can usually be a high degree of confidence in the results obtained from a well-conducted experiment. However, there are severe criticisms of the approach. In the real world people rarely operate in sterile surroundings such as these and, compared to the day-to-day conditions under which people have to perform complex tasks, the laboratory is a highly contrived and unrealistic situation. Nevertheless work of this nature should not be condemned out of hand. The experimental method is the only one that gets near to being able to identify cause and effect relationships, and work of this type can be vital in gaining insights into the ways that variables interact.
Sometimes, experiments are carried out in a real-world context, and these are known as fieldexperiments. For example, if we wanted to know the effect of changes to work methods on attitudes and job performance, this could be investigated in an actual work setting using control and experimental groups. However, in this type of experiment it is clearly not possible to protect against the effects of extraneous influences to the same extent that this can be done in a laboratory, and this can affect the results. For instance, if something else, such as management refusing to meet a pay claim, occurred at the same time, this might create a great deal of resentment and influence attitudes and job performance.
Ethnographic Designs
These usually consist of case studies, in which detailed information is collected about individuals, groups, or even a whole organisation. Work of this type is often far more focused on process than products (Garfinkel and Sacks 1970; Zimmerman and Pollner 1970); that is, it not only identifies what people do, but how they make sense of their surroundings. Data is often gathered by a wide variety of methods, including direct observation, so that a very rich and detailed picture of events is obtained as they happen, which sometimes permits very strong inferences to be made about causality. Usually no attempt is made to introduce controlled conditions, although for comparative purposes several different groups or organisations can be studied and the same data collection methods used in each one.
The main drawback of this approach is that it often focuses on one setting alone and, since this only gives a picture of what happened there, it can be difficult to generalise the findings to other contexts. For this reason it is often used in an exploratory way, for example to study new phenomena or generate ideas for more carefully controlled work later on. However, this is not an infallible rule. Some sociologists and anthropologists argue that all social situations are so inherently different that it is meaningless to even try to make generalisations. They also argue that if anything other than a case study approach is used something vital can be overlooked.
Data Collection Methods