D 319 TMA 04 - 30/6/99 J.P.BIRCHALL PO194869

Part A - Literature review.

Introduction. McNabb poses a crucial question does 'Segmented Labour Market' theory 'simply extend neo-classical models to include the effects of various institutional factors', or embody 'a new theoretical approach'?

Complex reality leads SLM theorists to focus on the exogenous variables in NC models. This does not invalidate NC theory rather confirms all models are necessary simplifications.

Differences.

Neo-classical models of the labour market are based on real wages and homogeneous labour in long run equilibrium.

In Fig.1 at the micro level, if real wages increase, W1 to W2, quantities specific labour supplied and demanded change L1 to Ls, and L1 to Ld. This produces excess supply which, with time, adjusts back to the long run equilibrium W1 L1.

In Fig.2 at the macro level, when nominal wages change, nothing happens but inflation. Employment remains at L1 as supply and demand adjust to S1 and D1 at the new nominal wage at W2.

In Fig.3 if demand shifts following innovative growth which disturbs the equilibrium, a different labour Ln is demanded with a new equilibrium at W2. McNabb stresses 'labour quality and labour productivity must be carefully distinguished'.

NC models are exogenous, only the price of homogeneous labour changes, ceteris paribus.

Cain confirms that omitting variables which are insignificant, or unmeasurable, or which introduce complexity, is a necessary simplification if insights are to be gained from any model.

Segmented labour market theory recognises 'everything depends on everything else' and develops more endogenous models.

SLM theory has roots in NC models, segmentation evolving from marginal productivity as capital and product markets adjust to shifts in technology. McNabb suggests institutional factors can be included in NC 'value judgements' as demand is a perception of marginal productivity and supply is an evaluation of opportunity costs. Segmentation is this scenario, contrasts with discrimination, the type of job is key not the type of person.

However, SLM theory also has roots in Marxist economics, where capitalist manipulation dominates theory.

McNabb suggests the power within firms to structure jobs determines wages not market forces.

In this scenario culture, custom and practice and job evaluation criteria are not market determined. Segmentation results from manipulation by simple supervision, technical control or bureaucratic control in the large corporation. Capitalist divide and rule, and class exploitation, derive from a lack of competition.

Classification of differences.

1 Short or long term. Adjustments whether movement along lines or shifts involve time. NC theory stresses supply and human capital in the long term. SLM theory stresses monopoly demand and insider manipulation in the short term.

2 Micro or macro. NC theory starts with real wages at the micro level. The primary and secondary sectors in SLM are macro markets.

3 Marginal productivity or job type. NC supply and demand theory assumes units of labour are homogeneous. SLM focuses on different types of jobs.

Strengths and weaknesses. The weakness of NC theory lies in its assumptions. When so many variables are exogenous and 'everything depends on everything else' its explanatory power is undermined. However, there is a theoretical counter. When exogenous variables change, the curves shift, successfully accommodating SLM as an evolution of NC theory of monopoly power influencing the price of labour.

NC theory also appears ambivalent about conflict and competition. However, Cain points out that 'harmony' involved in the supply and demand equilibrium is not contradicted by inherent conflict because competition is essential to discover mutual benefit.

The weakness of the more radical SLM model is that the power to manipulate depends on a lack of competition.

Cain suggests confining labour to one segment with no mobility is implausible. Furthermore, discrimination through unfavourable job assignments regardless of marginal productivity maybe a short run expedient but hardly compatible with long run equilibrium.

When 'included in' the endogenous variables make the manipulative task gigantic. How do you manipulate all the variables all the time, particularly in a global economy? Even in a bounded police state the manipulative power of Gosplan proved to be a fiction!

McNabb concedes the point, 'empirical evidence for the SLM suggests there are correlations between characteristics of firms like wages, size and capitalisation, but the causal factors are not clear'.

Conclusion. The fundamental difference is that SLM wages are determined by institutional factors and manipulation and not by the NC idea of marginal productivity.

SLM theory is not an alternative theory. The differences don't invalidate NC theory they merely provide new insights into shifting and changing curves as 'institutional factors' are no longer exogenous but readily influence marginal productivity perceptions and opportunity cost evaluations.

Wages depend on the specific job demand and on the human capital supply.

Part B Fieldwork.

a) Piore's initial objective was to establish the effect of automation on skill levels in manufacturing jobs. The proposed methodology was to compare skill profiles in a series of cases where old factories with old techniques were being replaced with new ones for the same product through analysis of the engineering designs.

The fieldwork uncovered inadequacies in the original brief associated with data gathering difficulties and interpretation problems.

The issue could not be pinned down to the specifics of engineering design but appeared to depend on complex patterns of behaviour.

Data gathering from engineering designs although specific and apparently straight forward was only part of the story. Furthermore the data gathering proved to be expensive and time consuming to obtain, involving issues of selection, accuracy, definition and comparability and required specialist interpretation.

Clarification of these issues forced Piore to change his methodology and adopt open ended interviews and participant observation because no other information was available.

Behavioural data is subjective and involves the motivation of responders, it is tacit and incomplete. Instructions don't produce co-operation. Questions merely initiate stories and prejudices or misinformation. Extensive sensitive probing was required.

The new information generated a new model of how skill profiles emerge in complex change situations.

Piore found that his preconception was not reality. Skill levels were not associated with automation but with a plethora of variables which evolved through experimentation and from the complexity of human behaviour.

He could not calculate the answer he had to discover it. The start point was to know the right questions to ask.

In one case the model is assumed and fieldwork validates.

In the other case a new model is developed by the fieldwork.

b) introduction. Understanding economic behaviour involves the application of scientific methodology; observation and data gathering, theory development, generation of testable hypotheses, validation and peer review.

Fieldwork and data gathering is necessary during two parts of this process, prior to developing an explanatory new theory and when validating an existing theory.

These two elements of scientific methodology must be carefully distinguished.

Piore's lesson. Piore commenced fieldwork with the intention of validating a theory by testing the hypothesis that automation effects skill profiles.

During his field work he learned that a new theory of skill accumulation was needed. Crucially this necessitated a different type of data collection.

Open ended interviews and participant observation are the first steps in the scientific process before a model is developed. This type of fieldwork is essential because no testable preconceptions should exist. Theory then follows to explain the observations. This is the empirical tradition of British science.

Piore is being disingenuous when he says 'this has been difficult for economists to recognise because the process provides answers to problems which economists thought had been already resolved'. Good economists always observe in order to develop robust theory in the first place.

Piore is confusing theory development with hypothesis testing.

The confusion has been well understood ever since the 1923 Hawthorn experiments, when behavioural scientists investigated the effects of a 'pleasant' physical environment on worker output. Output did indeed increase in the improved environment but the scientists discovered that the control group output also responded!

The old theory was not validated, workers were not responding to a pleasant environment, instead a new theory was developed suggesting both groups responded to the 'social' environment created by the experiment itself [1]!

Hypothesis testing. The more specific fieldwork in Piore's initial plan is the type used to test hypotheses where the precise nature of the data needed is identified by the hypothesis itself. Multiple regression of this data type disentangles multiple variables and gives powerful support for theories when hypotheses are validated.

However validation is always a problem.

Exogenous models are difficult to validate because of the ceteris paribus assumption when variables can't be held constant. The linearity and aggregation assumptions in regression analysis can't cope with interconnected non-additive variables.

Endogenous models are also difficult to validate because of the immense complexity of interconnections. However scientists are making progress and recently simple rule based behavioural models have been developed which when iterated exhibit dynamic patterns which do mimic real world observations [2].

Theory development. Invariably economic theory starts with behavioural assumptions. Theory development must therefore involve investigations into human behaviour where open ended interviews and participant observation are essential to identify traits which are often tacit.

Examples of where this type of fieldwork is needed abound in economic science; exploring the shapes of indifference curves, the size of income and substitution effects, the realities of conspicuous consumption or distinctiveness, the utility of Z-goods, the position of threat points and household consumption possibility frontiers.

It is also required to understand the competitive strategies of firms, maximising or satisficing strategies, moral hazard and adverse selection, the production function itself and isoquant shapes.

Most basic of all, open ended interviews and participant observation are required to get an inkling into the perceptions of marginal productivity and the evaluation of opportunity costs which determine the shape and position of the demand and supply curves.

Conclusion. Understanding economic behaviour starts with the development of theory based on assumptions of behavioural traits. Investigation requires open ended interviews to carefully probe and tease out what are often tacit motivations, a completely different data type than the more specific fieldwork required for hypothesis testing.

Part C Quantitative Analysis.

Economic issues. Discrimination occurs when individuals are subject to unequal treatment resulting from actual or perceived differences in the average rates for groups, or populations, to which they belong.

Neo-classical theory indicates labour market discrimination occurs when individual wage rates are not explained by differences in marginal productivity.

This empirical analysis does not attempt to measure the effect of marginal productivity directly but analyses how much the difference in wage rates can be explained by education and experience, variables which may impact on productivity.

The hypothesis to be tested follows from decomposing the male and female wage equations, Wm = a0 + a1Em and Wf = b0 + b1Ef

into an intercept term - a0 - b0, indicating pre-entry differences

a productivity related term - b1(Em - Ef)

and term reflecting discrimination - (a1 - b1) Em

In this case the wage equation involves not one variable E, but two, experience and years of schooling. The squared terms push greater weight on higher figures and can be used to reflect the diminishing productivity of older workers.

Segmented Labour Market theory, on the other hand, suggests wages are not determined by marginal productivity but by external institutional factors associated with firms or the internal 'politics' of the organisation and the manipulation of job types. This job specific discrimination can be analysed by comparing coefficients of dummy variables shown in this study in a similar way.

Thus to answer the question are white females discriminated against compared to white males we need to test the null hypothesis (a1 - b1) Em = 0.

Statistical analysis. The sign of the coefficient indicates a positive or negative influence on wages.

Both experience and schooling have positive effects on wages.

The negative sign on the squared terms indicates a falling off of wages as productivity falls in older workers.

The size of the coefficient indicates it's significance. To test the statistical significance of the coefficients a t test statistic is used. A t-statistic > 1.96 suggests we can be more than 95% confident that the result is not chance.

The intercept terms are highly significant and indicate higher wages for males due to pre-entry conditions.

The experience and schooling coefficients are also significant but not the squared terms.

Thus the null hypothesis is not true, for both variables, the male coefficient is higher than the female, reflecting differences which are not explained by productivity related variables. Discrimination may exist.

For SLM theory there is supporting evidence for the theory in that job type (occupation, industry, location and environment) affects wages but the evidence for discrimination is mixed. There is discrimination against females in most occupations apart from Government, crafts and construction. There is discrimination in favour of females in most industry, perhaps reflecting recruitment of only the most 'able' females in the first place? There is consistent discrimination against females in regions but it is generally less significant than in industry, interestingly, many regions have negative effects on wages, perhaps reflecting the rich north and poorer south? There is no clear discriminatory trend with city size although female wages do fall away in the smaller cities.

The R2 figure is the error term indicating how much of the dependent variable has been explained by the chosen independent variables. In this case only 28% of the male wage rate and 34% of the female rate can be explained by the chosen variables, indicating other variables are involved as well.

Conclusions. Data and information is often not available explicitly it is widely dispersed, tacit, incomplete, selective, inaccurate and subjective. Furthermore data changes with the different individual interpretations, beliefs & emotions, free choices & feedback from learning.

Nevertheless regression analysis can provide useful insights into causal relationships where nothing is assumed to remain constant. If pinned to a theory, relevant supporting questions can be asked. However the emerging relationships are not precise but patterns, noise can be mistaken for a variable. Although the influence of variables is removed and not ignored, when 'everything effects everything else' the linearity assumption is flawed.