ABSTRACT

Youth unemployment in Australia is considered to be a significant problem, however, the burden of unemployment is not borne equally by neighbourhoods within metropolitan regions. The following study utilises data from the 1991 and 1996 Censuses of Population and Housing to determine the impact of family and neighbourhood on youth employment outcomes in the Melbourne metropolitan region.

The evidence shows that demographic characteristics of neighbourhoods are a significant causal factor in the employment outcomes of youths. It is also apparent that the economic recovery that has taken place since 1991 has not benefited teenagers living in areas with a low socioeconomic status. The implication for policy is that the targeting of individuals or groups at risk of unemployment may not be appropriate unless the relationship between youth unemployment, region and demography are taken into account.

INTRODUCTION

The evidence is mounting in Australia and elsewhere suggesting that the underlying relationships between family, neighbourhood and youth need to be addressed if employment opportunities for disadvantaged youth are to be improved (see, for example, Miller 1998, Le and Miller 1999, Hunter 1995, 1996, Gregory and Hunter 1995, Bradbury et al. 1986, Kelly and Lewis 2000).

There are important linkages between the socioeconomic status of regions and the employment and educational patterns of its constituents, particularly youth. Youth who are identified by certain regions and socioeconomic characteristics are more likely to be over-represented in the pool of young unemployed. They are also more likely to have low educational attainment. The consequences will be continued labour market disadvantage and inequality of opportunity.

Gregory and Hunter (1995) have shown that there is a growing concentration of urban poverty and that these areas are developing their own ‘pathologies’, the consequence being a cycle of increasing disadvantage. Consistent with this is a study by Le and Miller (1999) which found that socioeconomic factors were a significant causal factor in labour market outcomes. As Hunter (1996) points out, it is important to determine whether there are influences other than personal attributes that are contributing to inequality of employment outcomes. If there are, then policy aimed solely at correcting personal attributes will be deficient and will fail to adequately address the problem. For the purpose of the analysis undertaken in this paper youth are defined as those who are aged between 15 and 19 years.

Persistently high youth unemployment has become one of the dominant features of the labour market over the last two decades in Australia and other OECD countries. There has also been a trend away from low skilled employment in advanced economies (BLMR 1987, Committee on Employment Opportunities (CEO) 1993). As this is the main destination for those entering the labour market for the first time (Daly et al., 1998) the prospects for youth are somewhat diminished.

Figure 1 shows the pattern of youth unemployment in Victoria and Australia for the period 1980-2000. Youth unemployment rates observed since 1980 are similar to the rest of Australia, showing a clear pattern of cyclical fluctuation. Since the 1991 recession, however, they generally have been significantly higher than for Australia as a whole. Full-time youth employment in Victoria fell from 136 700 in June of 1980 to 45 200 in May of 2000, a decline of around 67 per cent. Over the same period the youth cohort decreased in size by 18 000 from 344 600 to 326 600, or around 5 per cent.

The major response to the collapse in youth full-time employment has been increased enrolment in further education and training (Lewis and Koshy 1999). A number of factors have contributed to this, such as government incentives (Kenyon and Wooden 1996, Lewis and Mclean 1998), increased government funding of places in tertiary education (Kenyon and Wooden 1996), decreased opportunity costs of study, due to reduced employment opportunities (Lewis and Koshy 1999) and increased relative returns to education (Chia 1991).

Source: ABS PC AUSSTATS, Tables LABCP2C, LABEM2G, LABEM2H
Note: All observations are for August using original data.

Figure 2 shows the percentage of youths in full-time and part-time employment and the percentage who are in some form of post-compulsory education for 1986-1999. The dramatic decline in full-time employment has been accompanied by a rise in the number of youths in part-time employment. There has also been a substantial rise in the number in post-compulsory education over this period. This pattern is consistent with the experiences of the rest of Australia.

Reduced opportunities to gain full-time employment and the increase in part-time employment opportunities have had a dramatic impact on participation in post-compulsory education and training (see, for example, Karmel 1995, Lewis and Koshy 1999).

Lewis and Koshy (1999) argue that increased education enrolments, to a large extent, are a reflection of hidden unemployment. Nonetheless, increased enrolments may be beneficial from both an individual and social perspective, as it has been clearly established in numerous studies that lifetime employment and earnings are enhanced by higher levels of training and education (see, for example, Miller 1998, 1986, 1982, Bradbury et al. 1986, Veum and Weiss 1993; Bell et al. 1992, Le and Miller 1999, Chia 1991).

Evidence suggests that youth unemployment is not only affected by the characteristics of individuals, but also by non-personal factors. Among the suggested non-personal factors contributing to unemployment are neighbourhood demographics and intergenerational rigidities (see for example, Miller 1998, Le and Miller 1999, Bradbury et al. 1986, O’Neill and Sweetman 1997, O’Regan and Quigley 1998, Hunter 1996, Gregory and Hunter 1995, Hunter 1995). Kelly and Lewis (2000) showed that there is significant variation in the geographic and socioeconomic distribution of youth unemployment and employment in the Perth metropolitan youth labour market. They found that neighbourhood characteristics and family effects play a significant role in determining youth labour market outcomes.

Two hypotheses are suggested for differential labour market outcomes between urban regions. The spatial mismatch hypothesis looks at the impact of job decentralisation and the constraints on housing choices of people who have low socioeconomic status. This model suggests proximity to work affects employment outcomes. Vipond (1984), for instance, found that location, measured by distance of a local government area to the Sydney CBD, was a significant determinant of unemployment differentials in the Sydney intra-urban labour market. Vipond argues that there are spatial frictions which prevent complete integration of the metropolitan labour market. The most important of these is accessibility to employment opportunities, which Vipond points out are greatest closest to the CBD.

Another approach suggests that concentration of people who are disadvantaged, that is, socioeconomic concentration, results in social isolation which has a negative impact on labour market outcomes (O’Regan and Quigley, 1998). The effect is felt on high school dropout rates and labour force participation rates.

Several mechanisms are suggested as influencing labour market outcomes under the socioeconomic concentration model alluded to above. Among these are the absence of positive role models, lack of informal job networks, and disruptive influences. The main contrast between the models is that the internal composition of neighbourhoods matters more than external employment opportunities in the socioeconomic concentration models (O’Regan and Quigley 1998).

A study by Gregory and Hunter (1995) found that socioeconomic status and its geographic concentration made no systematic contribution to the likelihood of employment in 1976 among the urban population. However, by 1991 the composition of neighbourhood employment had deteriorated substantially with the likelihood of unemployment increasing for low socio-economic status (SES) areas. In 1991 the employment rate of youths in low SES areas was only 80 per cent of that observed in high SES areas (Gregory and Hunter 1995).

Geographic concentration of unemployment has important consequences for the prospects of future employment. First, it reduces the network of friends and relatives in employment, thereby reducing the opportunities to find employment (Gregory and Hunter 1995, CEO 1993, O’Regan and Quigley 1998). Second, since employment opportunities are more likely to be in other locations (that is, other than low SES areas), then the costs of job search are increased (Vipond 1984). It also lowers the reward for working, since travel or re-location to the place of work will be more expensive, thus increasing the likelihood of spatial mismatch (Gregory and Hunter 1995).

Hunter (1996) estimated a model to test the hypothesis that the worsening employment outcomes in low SES areas was due to a sorting of characteristics. The endowments hypothesis suggests that the reason employment outcomes in low SES areas have deteriorated over time is because people who are unemployed, or have characteristics that make them more prone to unemployment, concentrate in areas which are already disadvantaged. That is, low SES areas experience higher unemployment because the concentration of constituents with attributes not conducive to employment is higher. The main findings from the Hunter (1996) study were that neighbourhood effects play a large part in determining employment outcomes. Hunter (1996) concluded that the returns to a given set of endowments had deteriorated substantially since 1976. Neighbourhood effects were interpreted as being the result of living in a low status collection district or intra-family effects which are correlated with an areas’ socioeconomic status (Hunter 1996).

Hunter (1995) found that industry structure was an important explanator of the decline in employment in low SES areas. The overall structure of industry may change over time in such a way that blue collar or low skilled employment declines. As employment of this type may be expected to be relatively high in low SES areas, then people in these areas are also likely to shoulder more of the unemployment burden. The other possibility is that industry may be in decline in these areas in particular. As a consequence the burden of unemployment will fall directly on these areas. The decline of industry in these areas may reflect general declines in specific industries, for example, manufacturing. Alternatively, the decline may come about due to the decline in household income in the area, due to declining employment ratios, which logically will feed into reduced economic activity and further labour market decline.

The relationship between a parent’s unemployment and their child’s may be a reflection of transmission of tastes, constraints, or, true state dependency (O’Neill and Sweetman 1997). Parents who have a low distaste for unemployment may raise children with similar attitudes towards unemployment. Thus, this will increase the likelihood of the child becoming unemployed. Another example of transmission of constraints involves the income of the parents being too low to finance further education of their children (O’Neill and Sweetman 1997) which would increase the likelihood of youth unemployment.

The geographic location of parents may result in the spatial mismatch of youths and employment. This is due to the financial dependency of youths on their parents, something which has been increasing over the last 14 years (Schneider 1999). This will reduce the probability of job matching and sampling where unemployed youth are located in disadvantaged areas with low employment opportunities and who are also financially dependent on their parents.

Using UK data O’Neill and Sweetman (1997) found that the probability of becoming unemployed for a male, if the father was unemployed, was over two times higher than for males whose fathers were employed. Local labour market conditions and education were also found to affect the outcome. However, even after controlling for these factors the tendency for greater unemployment still existed. This also appears to be the case in the Australian context. Bradbury et al. (1986) found that even when personal characteristics of youths are controlled for, high unemployment probabilities are positively correlated with the disadvantage of parents.

In this paper the impact that various demographic characteristics of neighbourhoods have on youth employment-population ratios in the Melbourne metropolitan region are examined. To do this a cross-section model of youth employment is estimated relating the youth employment to population ratio to various neighbourhood demographic variables. The model to be estimated is:

y_i = a + b _j C _ij + u_i

where:

y_i is the youth employment-population ratio

a is the intercept term

b _j is a vector of parameters to be estimated

C _ij is a matrix of observations on the demographic characteristics of neighbourhoods

u_i is the error term, assumed to be identically, independently normally distributed with mean zero and constant variance

i is a subscript denoting the neighbourhood of the observation

j is a subscript denoting the demographic characteristic

Employment outcomes of youths should, in the absence of the transmission of constraints and tastes from parents to youths, be largely independent of the demographic endowments of a neighbourhood. An important caveat to this statement is that the neighbourhood endowments should largely be a measure of the characteristics of adults and exclude that of youths. The choice of variables for the models outlined above have, where possible, been selected on the basis of their independence of youths’ characteristics and correspond directly with those used in Kelly and Lewis (2000).

Data for the estimation of this model are from the Australian Bureau of Statistics (ABS) 1996 Census of Population. The unit of analysis chosen was the neighbourhood, defined for the purpose at hand as being a collection district (CD). This is the smallest available area for analysis from the Census. Typically CDs contain around 225 households which is roughly equivalent to a suburban block. The CD captures the composition of neighbourhoods, whereas individual, or unit record, data does not.

The sample used is for the Melbourne metropolitan region based on the Census Major Statistical Region (see ABS 1996). In total there are 5241 CDs in the metropolitan region. The final sample available for the estimation of the model was 5 207. Those CDs that had no youth aged between 15 and 19 in the labour force were removed from the sample. The following sets out the variables used in the regression model. A statistical summary of the variables is given in table 2.

The distribution of state housing and the extent of home ownership is one of the mechanisms that may contribute to the degree of disadvantage in an area. Where the density of state housing is highest it might be expected that youth employment will be lower (see, for example, Bradbury et al. 1986, and Miller 1998). The own house variable is calculated as the proportion of dwellings within a CD that were owned by the occupants, being purchased by the occupants, or where the occupants were involved in a rent-purchase scheme at the time of the Census.

The state housing variable has been included as a dichotomous variable taking the value 0 or 1. CDs were counted as having state housing present if their state housing density was, statistically, significantly higher than the metropolitan average at the 2.5 per cent level of significance. The corresponding value was found to be 18.96 per cent. Thus, the characteristic was considered present if state housing as a proportion of total dwellings in a CD is greater than 18.96 per cent.

Bradbury et al. (1986) and Miller (1998) found that the level of family income was negatively correlated with youth unemployment. Extrapolating from this it could be inferred that the higher the proportion of people below a given level of income, for example, the median income, the greater the likelihood that the youth employment-population ratio in that area will be low. The median household income bracket for Victoria was $500-$699 per week in 1996. The variable used in the model is the number of households with total household income less than $500 per week as a percentage of all households within a CD. The variable has been normalised by dividing the total working age population (that is, those aged between 15 and 65) of a collection district.

The link between the educational attainment of parents and employment outcomes of children was not statistically significant in a study by O’Neill and Sweetman (1997). However, in other studies (see Miller et al. 1995, Miller 1998, Bradbury et al. 1986) it was shown to be a causal factor. The measures of human capital to be used in the model are primarily based on the extent of post-compulsory schooling that people possess within a CD. The measures used to capture the educational and occupational status of a neighbourhood are:

Five levels of qualification were aggregated to construct this variable. The qualifications used are higher degree, postgraduate diploma, bachelor degree, undergraduate diploma and associate diploma. These qualifications specifically relate to those who have already obtained them. Thus, it is unlikely, if not impossible, that youths will have had time, due to their age, to obtain the qualifications mentioned. On this basis it is reasonable to assume that this variable can be considered a measure of adults’ qualifications. The variable used is the number of people in a CD with these qualifications as a percentage of the CD population over 15 years of age.

This variable is the number of people who have a skilled or basic vocation qualification as a percentage of the CD population over 15 years of age.

This variable includes those categorised as tradespersons and related workers, advanced clerical and service workers, intermediate clerical, sales and service workers, and intermediate production and transport workers in the Australian Standard Classification of Occupations (ASCO) 2^nd edition. The sum of these categories is expressed as a percentage of the CD labour force.

This variable is the number of elementary clerical workers and labourers and related workers as a percentage of the CD labour force. The occupations have been coded using the Australian Standard Classification of Occupations 2^nd edition (ABS, 1996).

This variable is the number of persons in a CD aged between 20 and 65 who are unemployed as a percentage of the labour force aged between 20 and 65 years. The inclusion of this variable creates the link between youth and adult outcomes and a strong indication of whether there is a transmission of constraints within families and neighbourhoods. A significant constraint facing individuals and communities is the absence of viable networks capable of providing employment opportunities and pathways for the unemployed. From the perspective of those who consider themselves to be living in poverty it is ranked as a substantial disadvantage (Johnson and Taylor 2000).

A number of studies have examined the influence of English proficiency on employment (see, for example, Inglis and Stromback 1986, Miller and Neo 1997, Miller 1998). The consensus is that poor literacy is an important determinant of employment outcomes. Miller (1998) shows that human capital, especially in the form of higher qualifications, is less transferable internationally for those with a non-English speaking background than it is for those who come from English speaking countries.

To capture this influence the percentage of the CD population who speak a language other than English at home is used. It includes anyone in a household over the age of five years who normally speaks a language other than English at home. Due to the limitations of the data source it has not been possible to disaggregate this variable by age. The a priori expectation is that NESB will be negatively correlated with employment for two reasons. First, if being from a non-English speaking background lowers the employment prospects for parents, then the network of employment contacts is reduced for youths from these families, regardless of their English proficiency. Second, to the extent that the measure captures deficiencies in youths’ English proficiency, there will be a negative correlation with employment.

Previous studies have shown that Aboriginal and Torres-Strait Islanders are seriously disadvantaged in the labour market (see, for example, Le and Miller 1999, Ross 1990, 1993). However, Harris (1996) shows racial background, including being of Aboriginal descent, to be insignificant when educational attainment is controlled for (see also Bradbury et al. 1986). For the study of CDs in the Melbourne metropolitan area it has been necessary to include this characteristic as a dummy variable, due to the relatively small number of CDs that have Aboriginal and Torres-Strait Islanders resident. There are 96 CDs out of 5 207 with Aboriginal and Torres-Strait Islander residents. The dummy variable takes the value one if there are ATSI residents in the CD and zero otherwise.

Mobility may be a factor influencing employment outcomes for youth and adults alike, as they both share the same local labour market (Bradbury et al. 1986). In standard neoclassical analysis mobility of labour is important in improving employment outcomes. However, it has also been suggested that new entrants to an area lack detailed knowledge of local labour markets, consequently mobility may be an important contributor to poor employment outcomes (Le and Miller 1999, Bradbury et al. 1986). Other reasons suggested by Bradbury et al. (1986) are that families at greater risk of unemployment, for example, renting families, are also more likely to move. The measure used is the number of people who were enumerated at a different address in the 1996 Census than in the 1991 Census, as a percentage of the CD population.

Vipond (1984) tested the impact of distance on labour market outcomes in the Sydney labour market. The main finding was that there is a positive intra-urban unemployment gradient, that is, unemployment increases with distance from the CBD. Vipond gives evidence showing that the greater proportion of employment opportunities are within the CBD area. One of the reasons suggested for a positive intra-urban unemployment gradient is that there is informational friction associated with space (distance in this context) and also the transport structure is such that it is radial in nature, with the CBD at its centre. As a consequence, suburban industry is not well serviced by public transport networks. The counter argument to this is that the unemployment gradient should be negative, due to the decentralisation of industry away from the city centre and into the suburbs and outer regions. Those who can not afford to shift, due to economic disadvantage, are the ones who end up trapped in the inner city, thus a negative intra-urban unemployment gradient. In the context of this study where the dependent variable is the employment-population ratio, this is equivalent to a positive employment gradient.

The measure used here for the distance variable is the distance of the collection district from the CBD. This was calculated by taking the longitudinal and latitudinal coordinates of the collection district centroid and taking the straight-line distance to the CBD.

This variable captures the influence of sole parents on the employment outcomes of youth. Bradbury et al. (1986) found that labour market outcomes were significantly worse for youths when they were from sole parent families. This variable is the percentage of one parent families in a collection district.

The youth employment-population ratio was regressed on the variables listed in table 1 using Ordinary Least Squares (OLS). As is common with models that use cross-section data the presence of heteroscedasticity was detected. Although it can be shown that the derived estimates are still unbiased they will, nonetheless, be inefficient. As the sample is particularly large, the issue of inefficiency is less important than would otherwise be the case with a small sample size. However, the problem of biased variances remains and, as a result, can lead to incorrect inferences due to the standard errors of the estimated parameters being biased. To overcome this the model was re-estimated using OLS based on White’s heteroscedasticity adjusted standard errors (White, 1980). The results are presented below in table 2.

Consistent with prior expectations, the estimated model showed that the higher the proportion of housing in an area that is owned, the higher the youth employment rate. The estimated coefficient for the state housing dummy variable was not statistically significant at the ten per cent level but had the expected negative sign.

The estimated parameter of the income variable although significant at the one per cent level, did not have the expected sign. This suggests that the higher proportion of households below the median household income the higher the youth employment-population ratio. This runs counter to the argument that household financial constraints impinge on the employment opportunities of youth. Alternatively, and more intuitively, it is consistent with the proposition that the higher household income the lower the demand for paid employment and the greater the demand for education.

The NESB variable was highly significant and the negative sign consistent with findings from other studies (see, for example, Miller and Neo 1997, Inglis and Stromback 1986 and Miller 1986). Nonetheless, it is not clear from the model estimated here whether the effect is operating through the family and surroundings or through the youth themselves. It was not possible to separate NESB within households on the basis of age.

The estimated coefficient of the mobility variable was significant at the ten per cent level and had a negative sign. This result brings into question the neoclassical view that increased mobility reduces unemployment. It is consistent with the view that those who are disadvantaged, in terms of employment outcomes, are more likely to move or that new entrants lack detailed knowledge of local labour markets. It is also consistent with the results from a recent study of the Perth metropolitan region by Kelly and Lewis (2000).

The estimated parameters of both skilled and unskilled labour were strongly significant and also have a substantial impact on the employment outcomes of youths compared to professional labour (the omitted category). This is consistent with the estimated parameter for the income variable and adds further support to the notion that financial constraints have a bearing on the demand for education and the imperative to be in paid employment. Other mechanisms at work are the increased exposure to employment networks and positive role models.

The proportion of people holding higher qualifications in a CD exerts a strong influence on the employment-population ratio relative to no qualifications. Two possible explanations for the significant result in the model are demonstration effects leading to higher human capital development in youths and increased employment networks.

The skilled and basic vocation variable makes a substantial contribution to the differences in youth employment outcomes for a neighbourhood. The most compelling interpretation of this result is that direct employment opportunities are associated with the presence of tradespeople, whether they be in the family or in the neighbourhood. This might best be described as a transmission of opportunity, rather than constraint.

Adult unemployment was significant at the one per cent level and had a relatively large coefficient. This is consistent with studies on intergenerational mobility by O’Neill and Sweetman (1997) and the findings of Miller et al. (1995). That is, the labour force status of parents influences the outcomes of their children. It cannot be clearly ascertained as to whether the influence is a direct one from parents, or, whether overall neighbourhood adult unemployment exacerbates the outcome. Nonetheless, the strong significance suggests the presence of adult unemployment exerts a strong downward influence on youth employment. One possible explanation is that high youth unemployment is a result of transmission of constraints.

The ATSI variable was statistically insignificant which contrasts with the findings of Le and Miller (1999) who found that indigenous status increases the likelihood of unemployment. However, Harris (1996) found that this variable is insignificant when other factors, such as educational attainment, are controlled for. The insignificance of the variable in this study may well be due to the relatively low densities of indigenous people living in Melbourne’s neighbourhoods.

The distance variable was significant at the 5 per cent level and is consistent with the negative unemployment gradient (positive employment gradient) explanation of intra-urban labour market adjustment. It is of some interest that this is the opposite result to that obtained by Kelly and Lewis (2000) in their study of the Perth metropolitan region. It also contrasts with an earlier study of the Sydney labour market by Vipond (1984) who found that unemployment was positively correlated with distance from the CBD. The reason suggested by Vipond is that the greatest opportunities are closest to the city and that proximity to employment may be driving the employment outcomes in that labour market. That is, there are spatial frictions that prevent the full integration of intra-urban labour markets.

The one-parent families variable was also significant at the 5 per cent level and had the expected negative sign. Given high ratios of family separation this is of some interest in the context of youth labour markets and is certainly one of the overlooked side effects of divorce. Exactly what the mechanism is that is affecting the employment outcome however is difficult to determine. Whether it is a resource constraint or some kind of scarring effect cannot be ascertained from the results.

The main determinants of youth employment outcomes, on the basis of statistically significant coefficients, are the extent of home ownership, NESB, income, mobility, skilled and unskilled labourers, people holding higher qualifications and skilled and basic vocations, adult unemployment, one parent families and distance from the CBD.

One of the most substantial determinants of youth employment-population ratios is the level of adult unemployment in a neighbourhood.

The Melbourne metropolitan region youth unemployment rate displays substantial variation between collection districts. Of interest is whether the variation that exists within the metropolitan area is evenly distributed on the basis of socioeconomic status or whether it is highly concentrated.

The ABS Socioeconomic Index for Areas (SIEFA) 1996 Socioeconomic Index of Disadvantage was used to group collection districts into deciles according to their index value. High scores of this index indicate relative wellbeing for an area. The components of the index include, for example, household wealth, unemployment, and educational attainment of the residents in a collection district (ABS, 1998). The same sample of collection districts is used here for the 1991 and 1996 observations. Figure 3 illustrates the unequal distribution of the unemployment burden on areas that have a low socioeconomic status in the Melbourne metropolitan region. Youth unemployment rates have been determined for each socioeconomic decile.

In 1991 high levels of youth unemployment were observed across all deciles, although collection districts in the lower socioeconomic deciles had higher youth unemployment rates than collection districts with a higher socioeconomic status. However, by 1996 the situation for the lowest decile had deteriorated further.

It appears that the economic recovery that was well under way by this time did not benefit disadvantaged areas, at least as far as youths are concerned. The average youth unemployment rate for the metropolitan region at the time of the 1996 census was 18.4 per cent, around 1 percentage point lower than the Victorian youth unemployment rate. As can be seen from figure 3 the top six deciles, ranked by the SEIFA Index, recorded youth unemployment rates below this figure. Lower deciles recorded very high youth unemployment rates, the lowest decile being over 33 per cent. Whereas all other deciles recorded some improvement compared to 1991, some of them quite marked, the most economically disadvantaged areas experienced a worsening of youth unemployment. The average unemployment rate in the lowest decile was around 3 times as high as the most well-off decile.

One possibility is that the areas in question are disadvantaged due to their distance from areas of employment. Some studies have found proximity to employment to be a significant determinant of employment outcomes (see, for example, O’Regan and Quigley, 1998; Vipond, 1984). The results obtained in this study brings into question whether the employment opportunities are best closest to the Melbourne city centre. If they are then there is cause for concern, since there is a positive link between distance from the centre of the city and employment outcomes. This would support the idea that there is some kind of neighbourhood and family effect at work among inner city neighbourhoods. Alternative explanations are that entry level employment opportunities are located away from the city centre, suggesting spatial frictions are the culprit. Finally, it may be that the financial resources of households for areas closest to the CBD are such that youths are supported by their families, while they are undertaking post-compulsory study for example, without the need to work.

In a study of the Perth metropolitan region it was shown there is a concentration of youth unemployment in the areas surrounding the Central Business District. Some of the heaviest concentrations of the unemployed are located in areas that are well serviced by major road networks, such as the freeway system that cuts through the centre of the city. These areas also have a modern rail system and are well serviced by bus transport. The geographic distribution of youth unemployment suggests that there were factors other than proximity to employment that contributed to employment outcomes. Nonetheless, proximity was still an important factor.

The principal argument of this paper is that neighbourhood characteristics play a significant role in determining youth employment outcomes. Most prominent in the Australian literature is the argument relating to the growing concentration of urban poverty and the effect that this is having on employment outcomes. Some of the suggested mechanisms at work in these models are the lack of informal job networks, absence of positive role models, and disruptive influences. The internal composition of a neighbourhood appears to be just as important as external employment opportunities. Whether the adverse employment outcomes experienced in poor neighbourhoods are a result of the sorting of endowments or the result of the decreasing returns to a given set of endowments is, arguably, inconclusive. In the context of youth unemployment it still remains to be determined. What is more certain, and is the finding of this paper, is that there is a strong relationship between the demographic characteristics of a neighbourhood and the employment outcomes of youths. This has been shown to be the case in previous research by the authors for the Perth metropolitan area.

Clearly youth unemployment is not a general problem which can be addressed by macro level solutions, such as stimulating aggregate demand. The analysis of youth unemployment needs to address the role that neighbourhood characteristics are playing in determining youth employment outcomes. A more complete understanding of the transmission of constraints and intergenerational rigidities is required before appropriate policy can be put in place to address youth employment opportunities and emergent youth employment equity issues. Suffice to say, more emphasis needs to be placed on the geographic and socioeconomic concentration of unemployment in policy approaches to the issue.

Australian Bureau of Statistics (1992), Labour Statistics Australia: Catalogue No. 6101.0, AGPS, Canberra

- (1996), Census Dictionary: Catalogue No. 2901.0, AGPS, Canberra

- (1998) Census of Population and Housing Socioeconomic Indexes for Areas, Catalogue No. 2039.0, AGPS, Canberra

- CDATA 91 (database), Catalogue No. 2019.0.30.001 (CD-ROM), Canberra

- CDATA 96 (database), Catalogue No. 2019.0.30.001 (CD-ROM), Canberra

- (1999), PC AUSSTATS (database) (online), Available Gopher: gopher://gopher.abs.gov.au:70/11/PUBS/labou (updated daily)

Bradbury, B., Garde, P. and Vipond, J. (1986), ‘Youth Unemployment and Intergenerational Immobility’, Journal of Industrial Relations, Vol. 28, No. 2, pp. 191-210

Bureau of Labour Market Research (BLMR), (1987), Structural Change and the Labour Market, Research Report No. 11, AGPS, Canberra

Chia, T. T. (1991), ‘Has the Value of a Degree Fallen? Cross-Sectional Versus Time-Series Evidence’, in International Economics Postgraduate Research Conference Volume, Supplement to Economic Record, pp. 41-52

Committee on Employment Opportunities (CEO), (1993), Restoring Full Employment: A Discussion Paper, AGPS, Canberra

Daly, A. Duc, N. H., Elridge, D., Gabbitas, O. and McCalman, P. (1998), Youth Wages and Employment, Productivity Commission, Staff Research Paper, October, Ausinfo, Canberra

Gregory, R. G. and Hunter, B. (1995), ‘The Macroeconomy and the Growth of Ghettos and Urban Poverty in Australia’, Centre for Economic Policy Research Discussion Paper, No. 325, April 1995, Canberra

Harris, M. N. (1996), ‘Modelling the Probability of Youth Unemployment in Australia’, Economic Record, Vol. 72 No. 217, pp. 118-29

Hunter, B. (1995), ‘The Social Structure of the Australian Urban Labour Market’, Australian Economic Review, 110 (2), April-June, pp. 65-79

Hunter, B. (1996), ‘Explaining Changes in the Social Structure of Employment: the importance of geography’, Social Policy Research Centre, Discussion Paper No. 67, July 1996, University of New South Wales, Sydney

Inglis, P. A. and Stromback, T. (1986), ‘Migrants Unemployment: The Determinants of Employment Success’, Economic Record, Vol. 62, No. 178, pp. 310-24

Johnson, J. and Taylor, J., 2000, ‘The Invisible Australians’: Conceptions of Poverty in Australia, Paper delivered to the 7^th Australian Institute of Family Studies Conference, 24-26 July 2000, Sydney

Kelly, R. and Lewis, P.E.T, (2000), ‘The Impact of Intergenerational Effects and Geography on Youth Employment Outcomes: A Study of the Perth Metropolitan Region’, Australasian Journal of Regional Economics, Vol 6 No 1, 2000.

Kenyon, P. and Wooden, M. (1996), ‘Labour Supply’, in The Changing Australian Labour Market, K. Norris and M. Wooden (eds), Commission Paper No. 11, Economic Planning Advisory Commission, AGPS, Canberra

Le, A. T. and Miller, P. W. (1999), A Risk Index Approach to Unemployment: an Application Using the Survey of Employment and Unemployment Patterns, Australian Bureau of Statistics Occasional Paper, February 1999, Canberra, ABS catalogue No 6293.0.00.001

Lewis, P.E.T. & Koshy, P. (1999), ‘Youth Employment, Unemployment and School Participation’, Australian Journal of Education, Vol 43, No. 1, forthcoming

Lewis, P.E.T. & Mclean, B. (1998), The Youth Labour Market in Australia, paper presented to the Productivity Commission workshop on youth unemployment, Australian National University, November.

Miller, P. W. (1986), ‘Unemployment Patterns in the Youth Labour Market’, Australian Economic Papers, Vol. 25, No. 47, pp. 227-235

Miller, P. W. (1998), ‘Youth Unemployment: Does the Family Matter?’, Journal of Industrial Relations, Vol. 40, No. 2, pp. 247-276

Miller, P. W. and Neo, L. M. (1997), ‘Immigrant Unemployment: The Australian Experience’, International Migration, Vol. 35, No. 2, pp. 155-83

Miller, P., Mulvey, C. and Martin, N. (1995), ‘Earnings And Schooling: An Overview Of Economic Research Based On The Australian Twin Register’, Centre for Labour Market Research, Discussion Paper 95/4, Murdoch University, Perth

O’Neill, D. and Sweetman, O. (1997), Intergenerational Mobility in Britain: Evidence from Unemployment Patterns, September 2 1997, National University of Ireland, Maynooth

O’Regan, K. M. and Quigley, J. M. (1998), Spatial Effects Upon Employment Outcomes: The Case of New Jersey Youths, March 1998, University of California, Berkeley

Ross, R. T. (1990), ‘Employment prospects for Aboriginals in New South Wales’, in Aboriginal Employment Equity by the year 2000, J. C. Altman (ed), Centre for Aboriginal Economic Policy Research, Australian National University, Canberra

Ross, R.T. (1993), ‘A Probit Analysis of Aboriginal Employment Prospects in New South Wales’, Economic Record, Vol. 69, No. 206, pp. 253-58

Schneider, J. (1999), ‘The Increasing Financial Dependency Of Young People On Their Parents’, Social Policy Research Centre, Discussion Paper No. 96, University of New South Wales, Sydney

Vipond, J., (1984), ‘The Intra-Urban Unemployment Gradient: The Influence of Location on Unemployment’, Urban Studies, Vol. 21, pp. 377-88

White, H. (1980), ‘A Heteroscedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroscedasticity’, Econometrica, Vol. 48 No. 4, pp. 817-838