Equality and Equity Report

Introduction

The job search behaviour of the unemployed is important because it is likely to affect the chances of the unemployed finding a job. In particular, we would expect that the more intensely someone looks for a job, the more likely he or she will be to find a job. The Labour Force Survey has information on the number and type of job search methods used by the unemployed. This information can be used to derive a proxy for search effort or search intensity^[1]. The main purpose of this chapter is to investigate differences in job search behaviour amongst Catholics and non-Catholics.

The Data

In the LFS, the unemployed were asked to indicate which of the following job search methods they used. If they used more than one of these methods then they were also asked to indicate which was their 'main' method.

1 2 3 4 5 6 7 8 9

visit job centre, employment office, job club etc; have name on books of employment agency; advertise for jobs; answer newspaper etc job adverts; study newspaper etc job adverts; apply directly to employers; ask friends and relatives etc about jobs; wait for results of job application; other methods.

Religious Differences in Job Search Behaviour

Table 7.1 shows the percentage of the unemployed using each of the job search methods and Table 7.2 gives the percentage of the unemployed using each of the methods as the 'main' search method. The most common methods used are looking at advertisements in newspapers and visiting the Job Market/Club. The figures show that there are statistically significant differences in the type of methods used by Catholics and non-Catholics. For example Table 7.1 suggests that Catholics are less likely than non-Catholics to visit the Job Market/Club, answer advertisements in newspapers and approach employers directly. However, Table 7.2 suggests that Catholics are more likely than non-Catholics to approach employers directly as their main method of job search. Also, non-Catholics are more likely than Catholics to answer advertisements as their main method of job search.

Note: Employer contact search methods are denoted by (ec). An astrix denotes that the differences between Catholics and Others are significant from a statistical point of view.

Note: Employer contact search methods are denoted by (ec). An astrix denotes that the differences between Catholics and Others are significant from a statistical point of view.

Note: The average number of search methods used by Catholics and others are significantly different from a statistical point of view.

Note: The average number of employer contact search methods used by Catholics and others are significantly different from a statistical point of view.

Tables 7.5A and 7.5B show the percentage of unemployed Catholics and non-Catholics using an employer contact search method and using an employer contact search method as the main search method. The figures show that non-Catholics are more likely than Catholics to use employer contact search methods; 61% of Catholics use employer contact search methods versus 67% of non-Catholics. However the figures also show that there is very little difference between Catholics and non-Catholics in terms of using employer contact search methods as the main search method.

Note: The percentages of Catholics and Others using employer contact search methods in part (A) are significantly different.

Discussion of Religious Differences in Job Search Behaviour

It has been shown above that there are some statistically significant differences between Catholics and non-Catholics in terms of various aspects of job search behaviour. In particular it has been shown that Catholics tend to use fewer job search methods than non-Catholics. This is important because the number of job search methods has been used in the economic literature on job search as a proxy for search effort or search intensity. This means that the LFS data suggest that unemployed Catholics do not look as intensely for work as unemployed non-Catholics.

However, this result cannot be accepted at face value for at least two reasons. Firstly, it may be the case that the total number of job search methods is not a good proxy for search effort. In order to illustrate this consider someone who used one job search method a large number of times, eg someone might have answered ten advertisements in newspapers. Compare this individual to someone who uses two methods once only eg someone might have answered one advertisement in a newspaper and paid one visit to the jobmarket. If the number of job search methods is used as a proxy for search effort, then the first individual is judged to have searched less intensely than the second individual, which is clearly not the case. Therefore, it may be the case that for some individuals, the total number of job search methods is not a good proxy for search effort.

The second reason for being cautious about the idea that Catholics search less intensely than non-Catholics is that it is possible that it might be explained in terms of some of the differences in the characteristics of Catholics and non-Catholics. For example, we have seen that Catholics are more likely than non-Catholics to be long-term unemployed and, for reasons relating to motivation, discouragement etc, the long-term unemployed tend to use fewer search methods than the short-term unemployed. Therefore, it might be the case that part of the reason why Catholics tend to use fewer search methods than non-Catholics is that Catholics are more likely to be long-term unemployed than non-Catholics. A priori, similar arguments could be made in terms of other factors such as, for example, the geographical location of Catholics and non-Catholics.

In order to assess the strength of these arguments a range of econometric models of job search behaviour have been estimated. The results of these models are given in Appendix 7.3. The main reason for estimating the models is to see whether or not there were significant differences in the job search behaviour of Catholics and non-Catholics after a range of factors such as unemployment duration, geographical location and qualifications, etc have been taken into account.

Generally speaking, the results of the econometric models suggest that Catholics still tend to search less intensely for jobs even after other factors such as duration etc have been controlled for. For example, it was shown above that the average number of job search methods used by Catholics and non-Catholics were 3.1 and 3.6 respectively. This means that there is a 0.5pp difference in the average number of methods used by Catholics and non-Catholics. The econometric analysis suggests that between 0.2pp and 0.4pp of this can be accounted for by religion and the remainder can be accounted for by the other factors such as duration, geographical location, etc. Similar results are obtained when looking at the number of employer Contact methods used.

Overall, therefore, the econometric results suggest that about half of the observed difference in the number of search methods used by Catholics and non-Catholics can be explained by religion and about half can be explained by other observed factors which are correlated with religion. However, the 'bottom line' in all of this is that our results still suggest that unemployed Catholics search less intensely for work than unemployed non-Catholics. However, extreme care must be taken when interpreting this result. One interpretation of the finding might be that Catholics are less work orientated than non-Catholics. However, the limited available evidence on attitudes to work in Miller (1978) and McWhiter (1989) does not support this explanation. Therefore we reject the hypothesis that Catholics search less intensely for work than non-Catholics because they are less work orientated.

An alternative interpretation is that Catholics believe, either correctly or incorrectly, that they have a lower chance of obtaining a job. Job search theory shows that, all other things being equal, it is rational to search less when the probability of success IS lower.^[2] The findings of the Northern Ireland Social Attitudes Surveys (Osborne, 1989) are consistent with this explanation. The Social Attitudes Surveys show that Catholics believe that they are less likely to be successful in obtaining a job than equally qualified non-Catholics. In our view, therefore, the most likely reason why Catholics use fewer search methods than non-Catholics is that they believe their chances of getting a job are lower.


Notes:
[1] Hughes and McCormick (1990), Schmitt and Wadsworth (1993) and Wadsworth (1991) all suggest that the number and type of search methods used are good proxies for unobservable search effort or intensity.

[2] This is demonstrated formally in Appendix 7.2.

Introduction

This chapter investigates the differences in labour market flows between Catholic and other men using data from the Labour Force Survey. Labour market flows refer to movements between employment, unemployment and economic inactivity. The focus is on three topics - labour turnover, flows from employment for those who left a job and flows from unemployment for those who left unemployment.

We are interested in labour market flows because, at the individual level, the incidence of unemployment equals the probability of entering unemployment times the expected duration of unemployment. We examine flows into and out of unemployment as opposed to the stock of unemployment.

In our LFS sample, most men enter unemployment from employment so it is useful to examine separately labour turnover, ie the rate of leaving jobs, and flows from employment given labour turnover, ie conditional on having left a job. The other large labour market flow in our sample is the flow out of unemployment. However, the sample size is rather small if we restrict our analysis to those who left unemployment. Thus we examine the economic activity of all those unemployed one year before as well as those who left unemployment.

Data and Econometric Models

The LFS contains data on current economic activity and economic activity one year ago for all individuals. We also have data on the length of continuous employment and the duration of unemployment for those currently employed or unemployed. The limited and discrete nature of the available data severely restricts the type of econometric models of labour market flows which can be estimated'. See Appendix 8.1 for details.

In Great Britain, unlike Northern Ireland, the annual LES sample includes some individuals who were interviewed one year before. The recently introduced quarterly LFS in Britain includes many individuals who were interviewed in the previous quarter since only 20% of the households in the sample are replaced each quarter. With similar LFS data, panel data or cohort data for Northern Ireland we could estimate a range of less restrictive models.

Labour turnover is modelled using a probit equation. Flows from employment, conditional on turnover and flows from unemployment are modelled using multinomial logit and bivariate probit models. These models and their limitations are discussed in Appendix 8.1. We regard the models as mainly summarising the relationships in the data.

Labour Market Flows

Tables 8.1 and 8.2 show the flows between employment, unemployment and inactivity, disaggregated by religion, in our LES sample of men aged 20 to 59. Table 8.1 expresses current economic activity as a percentage of economic activity one year ago. For example, of those Catholics who were employed one year ago, 92.3% are currently employed. For completeness in Table 8.2 we reverse the order and express economic activity one year ago as a percentage of current economic activity. For example, of those Catholics who are currently employed, 92.4% were employed one year ago. Table 8.1 is the more important table. There are significant differences in the flows to and from employment and unemployment between Catholics and others. For example, in Table 8.1, 6.3% of Catholics who were employed one year ago are unemployed now and 70.1% of Catholics unemployed one year ago are unemployed now. The corresponding figures for non-Catholics are 3.3% and 64.7% which are significantly lower.

Table 8.3 shows the figures from Table 8.1 broken down by age group. The table is in two parts. The first part refers to Catholics and the second part refers to non-Catholics. Although the sample sizes in some cells are small, the statistically significant differences between Catholics and others still remain. For example, looking at 20 to 24 year olds, we find that 16.2% of Catholics employed one year ago are unemployed now whilst 23.9% of those unemployed one year ago are employed now. The corresponding figures for non-Catholics are 9.0% and 39.8% respectively which are significantly better.

Labour Turnover

In Table 8.4 labour turnover is disaggregated by age group. The turnover rate is the percentage of those employed one year ago who are either employed in a different job or else not employed now. The aggregate Catholic turnover rate is significantly higher than the non-Catholic rate. Even when we take account of age, Catholic turnover rates are a little higher. However, the differences are small and are not statistically significant.

Flows from Unemployment

In Table 8.5 we examine what happened to those who left a job ie we condition on labour turnover. Our sample consists of those who were employed one year ago but who are either currently unemployed or inactive or else employed in a different job. There are significant differences in outcomes between Catholic and other men. In our sample 46.3% of Catholic men who left employment in the past year were unemployed as opposed to 30.3% of non-Catholics. Catholics are significantly more likely to be unemployed even when we control for age. For example, 56.4% of Catholic men aged 20-24 were unemployed as opposed to 36.2% of non-Catholic men. Catholic men aged 55-59 are significantly more likely to leave the labour force.

Unemployed One Year Ago

In Table 8.6 the current economic activity of all those who were unemployed one year ago is examined. Unemployed Catholics are significantly more likely to be unemployed and less likely to be employed one year later. When we disaggregate by age group we still find significant differences between Catholics and others. Of course some of the significant differences between Catholics and others may be due to differences in other observable characteristics apart from age and religion.

The sample used in Table 8.6 consists of all those unemployed one year ago. Thus the significant religion effects we find are due to some combination of religion effects on the hazard or exit rate from unemployment and, conditional on having left unemployment, the state entered on leaving unemployment. Ideally we want to disentangle these duration and transition state effects.

Flows from Unemployment

In Table 8.7 the sample is restricted to those unemployed one year ago who left unemployment, ie those who are currently employed, on a scheme, inactive or unemployed less than one year. With this sample we are mainly picking up the second effect noted above, ie differences between Catholics and others in the labour market states entered on leaving unemployment.

Although the sample size in Table 8.7 is small, we find that Catholics are significantly less likely to obtain a job, and are more likely to be on a scheme or be inactive. Overall 33.5% of Catholics who left unemployment are employed (other than on a government scheme), 15.4% are on a scheme, 8.8% are unemployed and 42.3% are inactive. The corresponding figures for non-Catholics are 46.3% employed (other than on a scheme), 13.7% on a scheme, 9.3% unemployed and 33.1% inactive. Most of the inactive are discouraged. The rates of discouragement are 71.7% for Catholics and 52.6% for non-Catholic men.

Econometric Analysis of Labour Market Flows

In Appendix 8.1 we consider how to model the labour market flow data in the LFS. Econometric model results are presented in Appendix 8.2. Labour turnover is modelled using probit models. A large number of explanatory variables are included, for example socio-economic group and industry dummies. In the various models which we estimate, being Catholic does not have a significant influences on labour turnover. From this we conclude that, when a range of other relevant variables are taken into account, there are no significant differences in labour turnover between Catholic and other men.

A number of different models were used to estimated outflows from employment and from unemployment. The details of these models are given in Appendix 8.2. Flows from employment are modelled conditional on labour turnover by considering those who left their job in the past. For these individuals, being Catholic reduces the probability of being employed by about 11.2% (0.112), increases the probability of being unemployed by 12.6% (0.126) and reduces the probability of being inactive by about 1.5% (-0.014). All three effects are statistically significant. In terms of flows from unemployed, the results suggest that Catholic men who are unemployed one year ago are about 4% (-0.038) less likely to be employed now than other men.


Conclusions

In this Chapter LFS data are used to examine differences in labour market flows, ie movements to and from employment, schemes, unemployment and inactivity, between Catholic and other men. We focus on labour turnover, outflows from employment and outflows from unemployment.

According to raw figures, Catholic men have a slighter higher labour turnover rate than non-Catholic men, ie Catholics are more likely to leave their job for whatever reason. However, in our econometric models of labour turnover, which take account of socioeconomic group, industry and other relevant variables, we find no significant difference in labour turnover between Catholic and other men.

When we examine outflows from employment and unemployment significant Catholic effects are found. Looking at all those employed one year ago who left their job for whatever reason, we find that ceteris paribus, Catholic men are significantly less likely to be employed and more likely to be unemployed. These effects are found for all age groups.

Amongst those who were unemployed one year ago and who left unemployment at some point in the past year, Catholic men are found to be significantly less likely to be employed and more likely to be on a scheme or unemployed now.


Notes:
[1] For example, it is not possible to estimate interesting models along the lines of Tuma and Robins (1980), Burdett et al (1984), Olsen et al (1986) and most of the other models surveyed in Chapter 6 of Devine and Kiefer (1991).

Overview

In this paper two large household survey datasets are used to provide a comprehensive picture of the relationship between male unemployment and religion. A range of econometric models are used to examine various aspects of unemployment in addition to the incidence of unemployment. These include economic inactivity and discouragement, the duration of unemployment, job search behaviour, labour turnover and labour market flows.


IFS and CHS Data

Labour Force Survey (LFS) and Continuous Household Survey (CHS) data for men aged between 20 and 59 are used. This choice of age group avoids most of the problems associated with the decisions to participate in education and training schemes and to retire.

These two datasets are large. The effective LFS and OHS samples are 11,300 and 6,600 respectively and so sampling variability is not a major problem. The LFS data are for the four years 1985, 1986, 1990 and 1991 and the OHS data are for the four years 1986 to 1989. The data have been recoded to ensure consistency over time.

The LES and OHS data are not ideal for answering all the questions we are interested

in. Panel or cohort data would be better. However, we use econometric models to extract as much information as possible from the existing data. With LFS and OHS data we cannot directly address the issues of the black economy and security related employment. Removing the security related jobs effect is likely to reduce the unemployment differential by between 10% and 15%.


Econometric Modelling

Econometric models are used to disentangle the effects of many factors which simultaneously contribute to the various aspects of unemployment. A large number of explanatory variables are included in the models. Our list of explanatory variables is more comprehensive than those used by other researchers and is based on studies of the incidence of unemployment for Northern Ireland, the UK and elsewhere.


The Unemployment Differential

In public debate a great deal of attention is attached to the male unemployment differential, the ratio of Catholic to Protestant (or non-Catholic) unemployment rates. The unemployment differential in our OHS data is 2.5.

In our LFS data this ratio fell from 2.6 in 1985/86 to 2.3 in 1990/91. However, a clustered sample design was used to collect the LFS data in 1985/86. As a result we cannot say that the fall from 2.6 to 2.3 is statistically significant.


Review of Previous Studies

We critically review two important studies which model the incidence of unemployment in Northern Ireland, namely Smith and Chambers (1991) and Compton (1990). These studies reach opposite conclusions and represent the two opposing views of the factors which account for the unemployment differential. We review some of the criticisms which have been made of these studies. In addition we set out some important new criticisms of Compton's methodology.


Incidence of Unemployment

A series of econometric models of the incidence of unemployment by religion were constructed and a range of statistical tests were carried out to check the robustness of our results. The models control for a large number of relevant factors including age, number of children, housing tenure, educational and other qualifications and area of residence. We find that religion accounts for about half of the unemployment differential in our two samples. Differences in the personal and other characteristics of the Catholic and non-Catholic populations account for the rest of the unemployment differential. Ceteris paribus, therefore, we find that Catholic men are significantly more likely to be unemployed than non-Catholic men. These findings are robust and are consistent with the results of Smith and Chambers (1990).


Interpretation of Findings

Our findings must be interpreted carefully. For example the large and significant Catholic effect on the incidence of unemployment does not necessarily equate with discrimination.

A large and significant Catholic effect may be explained by a range of factors which are not in our models because they are not measured in our data. These factors must be both correlated with religion and have a large effect on the incidence of unemployment. Possible factors commonly mentioned include differences in labour force growth, subject mix at school or college, motivation, as well as direct discrimination, indirect discrimination, or the 'chill factor'.^[1]

The effect of differences in labour force growth were examined using a stylised model. Under a range of plausible assumptions, we do not obtain a large effect on the unemployment differential. Some combination of very low labour turnover rates, zero or negative employment growth rates, high Catholic labour force growth rates and segregation in employment are required to obtain large effects on the unemployment differential. In addition we include age and the number of children as explanatory variables in our microeconometric models. These two variables are significant and we argue that they capture some of the labour force growth effects.

There is little evidence that subject mix in Northern Ireland or elsewhere has a large effect on the incidence of unemployment. Also the limited evidence available does not suggest that Catholics are less motivated, less flexible or have a poorer attitude to work than non-Catholics.

Suppose, for the sake of argument, we believe that some mix of direct or indirect discrimination or the 'chill factor' is important in explaining our results. With the data available, it is not possible to split this effect into direct discrimination, indirect discrimination or 'chill factor' components. In addition, it is difficult to apportion these effects into current and past components. The reason is that past unemployment increases the risk of current unemployment. This is likely to be due to the loss of human capital or 'stigma' or 'scarring' effects. Thus the current incidence of unemployment depends, in part, on the past incidence of unemployment. As a result, it is rather difficult to disentangle the effect of past factors from ongoing factors.

However, we may obtain some indication of whether the relevant factors operated only in the past or continue to operate more recently by examining the incidence of unemployment for young males, say those aged 20 to 24, and by examining flows to and from employment and unemployment. In both cases we find significant negative Catholic effects. This suggests that some current disadvantage is present for Catholics, and that the unemployment differential cannot be explained solely in terms of past disadvantage.


Economic Inactivity and Discouragement

Catholic men are significantly more likely to be economically inactive than non-Catholic men. In our models about half of the difference in inactivity rates between Catholics and others is explained by religion; the remainder is explained by differences in other characteristics.

In the raw data significantly more inactive Catholics are discouraged ie they are not actively looking for work because they believe that there are no, presumably suitable, jobs available. However, in our models when we control for a range of relevant personal characteristics and other factors, we do not find a significant religion effect on the incidence of discouragement. Ceteris paribus, inactive Catholic men do not appear to be more discouraged than non-Catholic men.

Surprisingly we find that there is no need to model the incidence of unemployment and economic inactivity jointly. When we control for a range of factors no significant correlation between unemployment and non-participation is found.

More inactive Catholics claim benefits. However the higher Catholic rate of claiming is largely accounted for by their higher rate of discouragement.


Duration of Unemployment

The duration of unemployment was examined using a variety of simple and more elaborate econometric models. The latter models attempt to take account of the way the data are generated.

Using LFS duration data we find significant religion effects. Catholic men are significantly more likely to be long-term unemployed than other men, ceteris paribus. Also, the exit rate of Catholic men from unemployment is significantly lower. These effects are found both in the raw data and in the models which control for a range of other factors.

With OHS duration data, no significant religion effects were found either in the raw data or in our models. This is highly surprising and implausible. If true, it implies that, ceteris paribus, differences in unemployment rates between Catholic and other men are solely due to differences in entry rates into unemployment. There is no evidence for this in the larger LFS samples. Examination of the data suggests that sampling variability is the reason for these strange results. In support of this is should be noted that there appears to be significant religion effects in the OHS duration data for the earlier years 1983 to 1985.


Job Search Behaviour

In our LFS sample, there are significant differences between Catholic and other men in the number and type of job search methods used. These are commonly used as proxies for search intensity. If we follow this approach, our findings imply that Catholic men search less than non-Catholic men. Of course, it may be the case that the number and type of job search methods used may be poor proxies for search intensity.

Catholic men used fewer job search methods and slightly fewer 'employer contact' search methods. 'Employer contact' methods are more direct and likely to be more successful.

The interpretation of these findings is not clear cut. Job search theory suggests that, ceteris paribus, it is rational to search less when the probability of success is lower. Thus Catholic men may search less because either they are less work orientated or they correctly or incorrectly believe that they have a lower chance of obtaining a job.

The limited available evidence does not suggest that Catholics are less work orientated and so the former explanation should probably be discounted. The findings of Social Attitudes Surveys are consistent with the latter explanation. High discouragement rates and a preference for a job amongst inactive Catholics are also consistent with this. In our view, therefore, the most likely reason why Catholics use fewer search methods is because they believe their chances of getting a job are less.


Labour Market Flows

The limited data on labour market flows in the LFS severely restrict the type of econometric models which can be estimated. We regard the models in this paper as mainly summarizing the relationships in the data. Labour turnover, flows from employment and flows from unemployment were all examined.

In the raw data Catholic rates of labour turnover are a little higher. However, when we control for socio-economic group, industry and other relevant variables we find no significant difference in labour turnover between Catholic and other men. Therefore, higher Catholic unemployment rates are not attributable to higher Catholic turnover rates.

When we consider all those employed one year ago who left their job for whatever reason, Catholic men are significantly less likely to be employed and more likely to be unemployed, ceteris paribus. This effect is found for all age groups.

Amongst all of those who were unemployed one year ago, Catholic men are significantly more likely to be unemployed and less likely to be employed one year on, even when we control for a range of relevant factors. Again, this effect is present for all age groups. We obtain much the same results when we examine what happened to those who left unemployment in the past year so this is unlikely to be just a high Catholic unemployment duration effect.


Notes:
[1] Few individuals employed in security related jobs or engaged in the black economy respond, either directly or by proxy, to household surveys such as the LFS or CHS, so we do not need to consider these two factors in detail. Of course, some men employed in security related jobs may disguise their true occupations by classifying themselves as civil servants. However, we do not believe that our findings are affected by this to any great extent.

Amemiya T (1981), "Qualitative Response Models: A Survey", Journal of Economic Literature, vol 19, pp 1193-1205.

Bera A K, Jarque C M and Lee L F (1984), 'Testing the Normality Assumption in Limited Dependent Variable Models, International Economic Review, vol 25, pp 573-578.

Borsch-Supan A and Hajivassiliou V (1993), "Smooth Unbiased Multivariate Simulators for Maximum Likelihood Estimation of Limited Dependent Variable Models", Journal of Econometrics, vol 58, pp 347-368.

Burdett K, Kiefer N, Mortensen D T and Neumann G (1984), "Earnings, Unemployment and the Allocation of Time over Time", Review of Economic Studies, vol LI, pp 559-578.

Cain G (1986), "The Economic Analysis of Labour Market Discrimination: a Survey" in Ashenfelter O and Layard R (eds) Handbook of Labour Economics, vol 1, North Holland, New York.

Compton P A (1981), "Demographic and Geographical Aspects of the Unemployment Differential between Protestants and Roman Catholics in Northern Ireland", in Compton P A (ed), The Contemporary Population of Northern Ireland and Population Related Issues, Institute of Irish Studies, Queens University, Belfast.

Compton P, Cormack, R and Osborne, R (1988), "Discrimination Research Flawed", Fortnight, no 288, pp 11-12.

Compton P A (1991), "Employment Differentials in Northern Ireland and Job Discrimination: A Critique" in Roche P J and Barton B (eds), The Northern Ireland Question, Myth and Reality, Avebury, Aldershot.

Cormack R J and Osborne R D (1989), "Employment and Discrimination in Northern Ireland", Policy Studies, vol 9, pp 49-53.

Crouchley R, Pickles A R and O'Farrell P N (1991), "Duration of Unemployment Among Men in Northern Ireland: Regional Inequality or Denominational Discrimination?", Environment and Planning A, vol 23, pp 903-920.

Davidson R and McKinnon J G (1993), Estimation and Inference in Econometrics, New York, Oxford University Press.

Devine T J and Kiefer N M (1991), Labour Economics: The Search Approach, Oxford University Press, New York.

Dhrymes P J (1986), "Limited Dependent Variables", in Griliches Z and Intriligator M D (eds), Handbook of Econometrics, vol III, Amsterdam, North Holland.

Eversley D (1989), Religion and Unemployment in Northern Ireland, Sage, London.

Gallie D and Vogler C (1990, "Unemployment and Attitudes to Work", Working Paper No 18, Social Change and Economic Life Initiative, Nuffield College, Oxford.

Godfrey L G (1988), Mis-specification Tests in Econometrics: The Lagrange Multiplier Principle and Other Applications, Cambridge University Press, Cambridge.

Greene W H (1993), Econometric Analysis, Macmillan, New York. Hajivassiliou V (1991), "Simulation Estimation Methods for Limited Dependent Variable Models", Cowles Foundation Discussion Paper No 1007, Yale University.

Hajivassiliou V, McFadden D and Ruud P (1991), "Simulation of Multivariate Normal Orthant Probabilities: Methods and Programs", Cowles Foundation Discussion Paper No 1021, Yale University.

Han A and Hausman J A (1990), 'Flexible Parametric Estimation of Duration and Competing Risk Models', Journal of Applied Econometrics, vol 5, pp 1-28.

Hausman J A and McFadden D (1984), "Specification Tests for the Multinomial Logit Model", Econometrica, vol 52, pp 1219-1239.

Heckman J (1979), "Sample Selection Bias as a Specification Error", Econometrica, vol 47, pp 153-161.

Heckman J and Singer B (1984), "A Method for Minimizing the Impact of Distributional Assumptions in Econometric Models for Duration Data", Econometrica, vol 52, pp 271-320.

Hoffman D and Pagan A (1989), "Post Sample Prediction Tests for Generalised Method of Moments Estimators", Oxford Bulletin of Economics and Statistics, vol 51, pp 333-344.

Horowitz J L (1985), 'Testing Probabilistic Discrete Choice Models of Travel Demand by Comparing Predicted and Observed Aggregate Choice Shares", Transportation Research B, vol 19B, pp 17-38.

Howe L (1989), "Doing the Double or Doing Without: The Social and Economic Context of Working on the Side in Belfast" in Jenkins R (ed), Northern Ireland: Studies in Social and Economic Life. Avebury/ESRC, Aldershot.

Howe L (1990), Being Unemployed in Northern Ireland: An Ethnographic Study, Cambridge University Press, Cambridge.

Hughes G and McCormick B (1990), "Measuring Unemployment and Cyclical Participation in the British labour Market", Scandinavian Journal of Economics, vol 92, pp 247-292.

Jarque C M and Bera A K (1982), ""Efficient Specification Tests for Limited Dependent Variable Models', Economics Letters, vol 9, pp 153-160.

Jones S (1989), "Job Search Methods, Intensity and Effects", Oxford Bulletin of Economics and Statistics, vol 51, pp 277-296.

Kiefer N M and Skoog G R (1984), "Local Asymptotic Specification Error Analysis", Econometrica, vol 54, pp 873-885.

Kiefer N (1988), "Economic Duration Data and Hazard Functions", Journal of Economic Literature, vol 26, pp 646-679.

Lancaster T (1979), "Econometric Models for the Duration of Unemployment", Econometrica, vol 47, pp 939-956.

Lancaster T (1990), The Economic Analysis of Transition Data, Cambridge University Press, Cambridge.

Lancaster T and Nickell 5 (1980), "The Analysis of Re-Employment Probabilities for the Unemployed", Journal of the Royal Statistical Society, Series A, vol 43, pp 142-152.

Lancaster T and Chester A (1984), "Simultaneous Equations with Endogenous Hazards", in Neumann G R and Westergard-Nielson N C (eds), Studies in Labor Market Dynamics, Berlin, Springer-Verlag.

Layard R, Nickell S and Jackman R (1991), Unemployment: Macroeconomic Performance and the Labour Market, Oxford University Press, Oxford.

Lechner M (1991), 'Testing Logit Models in Practice", Empirical Economics, vol 2, pp 177-198.

Lindeboom M and Theeuwes J (1993), "Search, Benefits and Entitlement", Economica, vol 60, pp 327-346.

Maddala G 5 (1983), Limited Dependent and Qualitative Variables in Econometrics, Cambridge University Press, Cambridge.

McCormick B (1988), "Unstable Job Attachments and the Changing Structure of British Unemployment 1973-83", Economic Journal, vol 98, pp 132-147.

McCormick B (1991), Unemployment Structure and the Unemployment Puzzle, Employment Institute, London.

McCormack V and O'Hara J (1990), Enduring Inequality: Religious Discrimination in Employment in Northern Ireland, National Council for Civil Liberties, London.

McFadden D (1989), "A Method of Simulated Moments for the Estimation of Discrete Response Models Without Numerical Integration", Econometrica, vol 57, pp 995-1026.

McKinnon J G (1992), "Model Specification Tests and Artificial Regressions", Journal of Economic Literature, vol XXX, pp 102-146.

McWhirter L (1989), "Longitudional Evidence on The Teenage Years", in Harbison J (ed), Growing Up in Northern Ireland, Stranmillis College, Belfast.

McWhirter L, Duffy V, Barry R and McGuinness G (1987), "Transition from School to Work: Cohort Evidence", in Osborne R D, Cormack R L and Miller R L (eds), Education and Policy in Northern Ireland, Policy Research Institute, Belfast.

Mackinnon J G and Magee L (1990), "Transforming the Dependent Variable in Regression Models", International Economic Review, vol 30, pp 315-339.

Magnac T and Robin J (1991), "An Econometric Analysis of Labour Market Transitions Using Discrete and Tenure Data", Discussion Paper No 91-18, Department of Economics, University of London, forthcoming Journal of Labour Economics.

Manski C (1975), "The Maximum Score Estimator of the Stochastic Utility Model of Choice", Journal of Econometrics, vol 3, pp 313-333.

Meyer B D (1990), "Unemployment Insurance and Unemployment Spells", Econometrica, vol 58, pp 757-782.

Miller R L (1978), Attitudes to Work in Northern Ireland, Belfast, Fair Employment Agency.

Miller R L and Osborne R D (1983), "Religion and Unemployment: Evidence from a Cohort Study", in Cormack R J and Osborne R D (eds), Religion, Education and Employment, Belfast, Appletree Press.

Miller R L, Osborne R D, Cormack R J and Williamson A P (1990), Higher Education and Labour Market Entry: The Differing Experiences of Northern Ireland Protestants and Catholics, Coleraine, University of Ulster.

Mortensen D (1986), "Job Search and Labour Market Analysis", in Ashenfelter, O and Layard R (eds), Handbook of Labour Economics, vol II, North Holland, Amsterdam.

Murphy A (1994a), "Further Tests of Mis-Specification for Discrete Choice Models", mimeo, paper presented at 1994 Irish Economic Association Annual Conference.

Murphy A (1994b), "Artificial Regression Based Mis-Specification Tests for Discrete Choice Models", mimeo.

Murphy A and I Shuttleworth 1(1994), "Educational Attainment, Subject Choice and the First Destinations of Northern Ireland School Leavers in 1990/91", mimeo.

Murphy A and Armstrong D (1993), "The Incidence and Duration of Male Unemployment in Northern Ireland", in Gudgin G and O'Shea G (eds), "Unemployment Forever: The Northern Ireland Economy in Recession and Beyond", Northern Ireland Economic Research Centre, Belfast.

Narendranathan W and Elias P (1993), "Influences of Past History on the Incidence of Youth Unemployment: Empirical Findings for the UK", Oxford Bulletin of Economics and Statistics, vol 55, pp 161 -185.

Nickell S (1979a), "Estimating the Probability of Leaving Unemployment", Econometrica, vol 49, pp 1417-1426.

Nickell S (1979b), "The Effect of Unemployment and Related Benefits on the Duration of Unemployment", Economic Journal, vol 89, pp 34-49.

Nickell 5 (1980), "A Picture of Male Unemployment in Britain", Economic Journal, Vol 19, pp 776-794.

Olsen R J, Smith D A and Farkas G (1986), "Structural and Reduced-Form Models of Choice Among Alternatives in Continous Time: Youth Employment Under a Guaranteed Jobs Program", Econometrica, vol 54, pp 375-394.

Osborne R D (1991), "Discrimination and Fair Employment", in Stringer P and Robinson G (eds), Social Attitudes in Northern Ireland: 1990-9 1 Edition, Blackstaff Press, Belfast.

Pagan A and Vella F (1989), "Diagnostic Tests for Models Based on Individual Data: A Survey", Journal of Applied Econometrics, vol 4, pp S29-S59.

Pakes A and Pollard D (1989), "Simulation and the Asymptotics of Optimisation Estimators", Econometrica, vol 57, pp 1027-1057.

Pissardes C A (1990), Equilibrium Unemployment Theory, Basil Blackwell, Oxford.

Pissarides C and Wadsworth J (1990), "Who are the Unemployed?", Discussion Paper No 12, Centre for Economic Performance, London School of Economics.

Pudney 5 (1989), Modelling Individual Choice: The Econometrics of Corners, Kinks and Holes, Basil Blackwell, Oxford.

Small K A and Hsiao C (1985), "Multinomial Logit Specification Tests", International Economic Review, vol 26, pp 619-627.

Smith D J (1987), Equality and Inequality in Northern Ireland: Part 1: Employment and Unemployment, Occasional Paper No 39, Policy Studies Institute, London.

Smith D J (1988), "No Substance to Criticisms", Fortnight, no 228, p 12.

Smith D J (1988), Policy and Research: Employment Discrimination in Northern Ireland", Policy Studies, vol 9, pp 41-60.

Smith D J and Chambers G (1991), Inequality in Northern Ireland, Clarendon Press, Oxford.

Schmitt J and Wadsworth J (1993), "Unemployment Benefits Levels and Search Activity", Oxford Bulletin of Economics and Statistics, vol 55, pp 1-24.

Stratton L (1993) "Racial Differences in Men's Unemployment", Industrial and Labour Relations Review, vol 46, pp 451-463.

Tuma N and Robbins P K (1980), "A Dynamic Model of Employment Behaviour: An Application to the Seattle and Denver Income Maintenance Experiments", Econometrica, vol 48, pp 1031 -1052.

Vuong 0 M (1989), "Likelihood Ratio Tests for Model Selection and Non-Nested Hypotheses", Econometrica, vol 57, pp 307-333.

Wadsworth J (1991), "Unemployment Benefits and Search Effort in the UK Labour Market", Econometrica, vol 58, pp 17-34.

White H (1982), "Maximum Likelihood Estimation of Mis-Specified Models", Econometrica, vol 50, pp 1-25.

Whyte J (1990), Interpreting Northern Ireland, Basil Blackwell, Oxford. Wilson T (1989), Ulster: Conflict and Consent, Blackwell, Oxford.

Wright R and Ermisch J (1981), "Gender Discrimination in the British Labour Market: A Reassessment", Economic Journal, vol 101, pp 508-524.

Yaron G (1991), "Sex Discrimination and Low Pay: An Empirical Analysis Using UK Data", Applied Economics Discussion Paper No. 108, Institute of Economics and Statistics, Oxford.

Yatchew A and Griliches Z (1985), "Specification Error in Probit Models", Review of Economics and Statistics, vol 66, pp 134-139.