RDP 1999-07: Job-Search Methods, Neighbourhood Effects and the Youth Labour Market 2. Job-search Behaviour: What do we Already Know?

2.1 Which Search Methods are Most Effective?

Table 1 summarises the job-search methods reported by Australian teenagers who were respondents to the AYS, which covers the period 1989 to 1994. Information about which methods of search proved successful is derived from the responses of teenagers who were employed at the time of the interview and had obtained their job in the year prior to the interview. These responses are summarised in column 1. The number of respondents who have been continuously employed over the year since their previous interview is also included in the second last row of the table.

Roughly one-third of teenagers found their job through friends and relatives, one-third through directly contacting employers, and the final third through indirect search methods such as newspapers and the CES. This information is consistent with the successful methods reported by teenagers for the 12 months to July 1998 from an alternative source, although the proportion using direct employer contact is slightly larger and the reported use of newspapers and the CES is correspondingly lower.^[2]

Respondents who reported that their main activity was looking for work were asked to list all the search methods they were using and, if more than one method was reported, to identify the main job-search method.^[3] For comparison, the main job-search methods of unemployed teenagers are summarised in column 2 of Table 1.

The most striking difference between the two groups is that the unemployed are much less likely to report that friends and relatives are their main search method than the employed are to have used this method successfully.^[4] The proportion of unemployed teenagers reporting that their main search method is either newspapers or the CES is significantly higher.

This is consistent with evidence about the success of different job-search methods used by 15 to 26 year olds who were respondents to the first wave of the Australian Longitudinal Survey (ALS) in 1985. Miller and Volker (1987) show that respondents using the CES or newspapers are less likely to leave unemployment than those using friends and relatives or direct employer contact. Jobs obtained through the CES are also less successful if success is measured by the duration of employment.

The Australian data do not contain any direct measure of how much effort the unemployed apply to different job-search methods, or the success of such effort in terms of the number of job offers made. This information is available for the US and the UK, and it provides further evidence that friends and relatives and direct employer contact are the two most successful job-search methods.

Holzer (1988) finds that 16 to 23 year old males in the United States, who were unemployed the month before their interview, spend more hours searching through friends and relatives and direct employer contact than through state employment agencies or newspapers, and receive more job offers from these direct search methods.^[5] He also shows that job offers generated through friends and relatives have an 81 per cent probability of being accepted, which is much higher than the acceptance rate for offers generated by other methods.

Jones (1989) presents similar evidence for the UK using a sample of unemployed people collected by the Economist Intelligence Unit in September 1982. He finds that there is some decline in the total hours spent on search as the duration of unemployment lengthens and that this is particularly noticeable for the number of hours spent using friends and relatives.

Jones also estimates the effects of hours spent using different search methods on measures of success such as the number of job offers received and the number of interviews obtained. He finds that hours spent using friends and relatives have no effect on the number of interviews received, but increases the number of offers received. While this supports Holzer's finding that the probability of obtaining an acceptable offer using friends and relatives is very high, it makes it difficult to interpret the effect of spending more hours searching on the final outcome.

Another of Jones' results which is difficult to interpret is that as the number of hours spent using a government employment agency increases, the number of interviews and the number of offers decreases. This suggests that the hours spent using different types of search may be proxies for omitted personal or background characteristics which are important. More time spent using newspapers is found to increase the number of interviews, but not the number of offers.

Overall, the evidence suggests that using an employment agency or newspapers are relatively unsuccessful job-search methods, compared with using friends and relatives or direct employer contact. The question then arises: why do unemployed teenagers use job-search methods which are apparently less successful?

Assuming that these job seekers are acting in their own best interests, there are two explanations for this behaviour. One is that job-search methods which proved more successful, on average, for employed teenagers are tried first and when they do not result in an acceptable job offer, alternative job-search methods are pursued. If this is the explanation, we would expect to see that teenagers with longer durations of unemployment are less likely to use direct search methods. The second possibility is that some individuals are living in environments or have characteristics which make it optimal to choose indirect job-search methods, even though they are less successful on average. Before addressing this question more directly, it is useful to consider why direct methods would be more successful.

Motivated by the stylised fact that direct methods appear to be more successful than indirect methods, Montgomery (1991) develops a model of imperfect information to explain why this might be the case. His model assumes that employers cannot observe the quality of potential employees and that, without further information, employers will offer all potential employees a real wage equal to the average productivity of the unemployment pool.

Montgomery also assumes that there is a social structure within which high-productivity workers are more likely to associate with each other than with low-productivity workers. In this environment, one way for firms to increase the probability of hiring a high-productivity worker is to offer jobs to potential employees who are recommended by current high-productivity workers.

The net gain arises because both the employer and the potential employee have better information about each other, increasing the probability that there will be a successful match. This process has also been argued less formally by Rees (1966) who also suggests that employed individuals will only refer capable workers to ensure that their own reputation is not affected. These incentives will reinforce the mechanism described in the Montgomery model.

2.2 What Factors Affect the Choice of Search Method?

Although the Montgomery model explains why friends and relatives will be a successful method of job search, it does not suggest that this method will increase the chances of finding an acceptable job for all the unemployed. In fact, the model relies on the fact that the unemployed are not homogenous in two respects. Firstly, for imperfect information to be an issue, potential workers must have different productivity levels when they are matched to a given job. Secondly, some are connected to more useful social networks than others.

McGregor (1983) considers the possibility that the job-information network provided by friends and relatives is local in nature. This would imply that the probability that friends and relatives constitute an effective job-search method would be highly dependent on local neighbourhood characteristics. In particular, he argues that information about jobs is more likely to come from employed people and consequently there will be less useful job information in high unemployment neighbourhoods. The expected probability of obtaining a job offer using information from friends and relatives in high unemployment areas is also likely to be lower as there will be more competition for any available jobs.

McGregor (1983) proceeds to test his hypothesis, that local labour market conditions affect job-search behaviour, using a sample of males who were unemployed in Glasgow in 1976. He finds that when personal characteristics are controlled for, neighbourhood unemployment rates do not influence the probability that friends and relatives are used for search. Although this is a disappointing result, there are several technical reasons why these estimates should be treated with caution: the unemployment rates are presented in high, medium and low bands and the estimation technique does not take proper account of the nature of the data being used.^[6]

Schmitt and Wadsworth (1993) find that the most important determinant of job-search method choice is the unemployed person's previous occupational status using a sample of unemployed male benefit recipients in Great Britain.^[7] The other consistently important variables are the duration of unemployment and a dummy for the 50 to 65 year old age group. Unfortunately, however, the use of personal contacts is not available as a separate category, but is subsumed in ‘other’ job-search methods.

Jones (1989) estimates the relationship between the number of hours used on different search methods and individual characteristics. Being male increases the number of hours spent searching through all methods except newspapers. Older individuals were less likely to spend time searching through government employment agencies, and individuals with technical qualifications spent more time directly approaching employers. A more puzzling result is that individuals from high unemployment areas spent less time searching through government employment agencies or newspapers. Holzer (1988) examines the factors which affect the probability that an individual will uses different search methods, although the results are only marginally significant at best.

2.3 A New Model of Job-search Method Choice

In Appendix A, we develop an economic model to formalise the way in which characteristics of the local environment can affect the job-search behaviour of unemployed individuals. This is done by extending the job-search model developed by Pissarides (1990) to allow individuals to have access to two different search methods with characteristics reflecting the differences between the indirect and direct search methods discussed above.

The first job-search method, labelled the general search method, captures search methods such as newspapers and employment agencies which provide general job information. The probability that the general job-search method will create a match between an unemployed worker and an unfilled vacancy is assumed to be a function of aggregate labour market conditions and the search effort of the individual.

The second job-search method, labelled the local search method, is designed to capture job-search methods, such as information from friends and relatives or direct employer contact. The effectiveness of these methods is assumed to be influenced by conditions in the local labour market, but not by the amount of effort applied by the job seeker. This implicitly assumes that the job-information network provided by friends and relatives is confined to the local area, and that the employed neighbourhood residents receive job information at a fixed rate, which the unemployed teenager cannot influence.

Unemployed individuals decide how much effort to devote to the general search method, taking into account the direct costs and the expected benefits involved. This choice will be affected by factors such as the local unemployment rate, which affects the probability of finding a job through the local search method, and the aggregate unemployment rate, which affects the probability of finding work through the general search method. It is shown in Appendix A that the level of search effort chosen, and hence, the probability of reporting that the main method of search is a general search method, increases as the local unemployment rate rises and falls as the flow of benefits to being unemployed or the efficiency of the local search method increases.

This formalises the arguments put forward by McGregor (1983) that higher local unemployment rates should increase the amount of search effort put into newspapers and employment agencies, and that job seekers in low unemployment neighbourhoods are more likely to be using friends and relatives for information. Although this is a plausible theoretical argument, the importance of job-information networks and local labour market conditions is essentially an empirical question which is addressed further in Section 4.

Footnotes

Successful and Unsuccessful Job Search Experience, ABS Cat. No. 6245.0. It should also be noted that the CES was replaced by Centrelink and a competitive employment services market, the Job Network, during the period covered by the survey. All services which replace the CES have been classified as the CES here for the purposes of comparison with the older survey data. [2]

Thus, teenagers looking for work, who are enrolled in full-time education are not included in this sample. [3]

Around one-third of the employed respondents moved directly into their current job from an earlier job. The differences observed in Table 1 could perhaps be explained if the job-search behaviour of on-the-job searchers was significantly different from that of individuals who are not already employed. An examination of the data suggests that this is not so. [4]

The sample of 16 to 23 year old males is taken from the 1981 Youth Cohort of the National Longitudinal Survey. [5]

The dependent variable used by McGregor (1983) takes the value one if the individual is using friends and relatives as a search method and zero otherwise. The analysis uses a linear probability model to estimate the proposed relationship rather than a method designed to deal with dependent variables of this type, such as logit or probit. [6]

The sample includes unemployed male benefit recipients who were interviewed in the General Household Survey between 1979 and 1982. [7]