Note: Where available, the PDF/Word icon below is provided to view the complete and fully formatted document
Disclaimer: The Parliamentary Library does not warrant or accept liability for the accuracy or usefulness of the transcripts.These are copied directly from the broadcaster's website.
National Press Club -

View in ParlView

Tables are not available online.

Since the early 1970s income inequality among individuals has been growing in most OECD countries. It has arisen from two sources; higher levels of unemployment especially in Europe, and widening wage dispersions, particularly in the US. Australia has also been subject to these trends and the increasing inequality has led to a fast growing research literature which documents the changes (Gregory 1994, Borland 1993, Saunders 1994)(1).

There is one aspect of inequality, however, which has received little attention. This is the association between income inequality and dispersion of economic opportunities across urban neighbourhoods. Individuals in similar economic circumstances tend to live together and increases in economic inequality among individuals may have had a polarising effect within our cities. Neighbourhoods of urban poverty may be increasing at the same time that high income neighbourhoods are becoming more common.

To measure the change in urban inequality we utilise income and employment data collected from each Census between 1976 and 1991. The data show that the economic distance between Australians from different parts of the city has widened to an extraordinary degree. An individuals view of the world is heavily conditioned by his friends, neighbours and the street on which he lives. Consequently, as inequality grows across neighbourhoods, it may not be recognised as quickly by all Australians as it should be. We suggest that it may be increasingly true to say that one half of Australia does not know how the other half lives and that this is not a good thing.

There are many reasons why we are interested in the increase in income inequality across neighbourhoods. One reason is that, ceteris paribus, we prefer an Australia which is more rather than less equal in terms of socio-economic status, income and the distribution of economic opportunities among its citizens. Consequently, we are interested in understanding the processes underlying inequality changes.

Another reason is that if economic disadvantage is being concentrated in particular neighbourhoods then these neighbourhoods may be developing their own pathologies which generate more poverty and produce further increases in inequality through time. In the US, there is a large literature which discusses the way in which urban ghettos foster further inequality but there is not a great deal of systematic Australian work on the economic and social dynamics of neighbourhood poverty. Can it be said, as many would believe, that the personal probability of economic and social success is related to the income level and Socio-Economic Status (SES) of the neighbourhood in which an individual is raised? If so, then greater income inequality among neighbourhoods must distribute life chances less equally.

Finally, the creation of urban ghettos may have important implications for our understanding of the role of macro-economic policy and its impact on the economy. Over the last 20 years, economic cycles in Australia have demonstrated a marked non-symmetry. Unemployment increases during recessions but during the recovery does not decrease to previous levels. Each recession, therefore, has begun from a starting point of higher unemployment. It may be that the changing geographical distribution of employment opportunities is an important part of this phenomenon.

The paper is structured as follows. Part I documents changes in neighbourhood income inequality between the Census dates 1976 and 1991. Part II focuses on increased employment inequality across neighbourhoods. Part m discusses neighbourhood inequality, urban ghettos and the role of policy. Part IV offers concluding remarks.

I The Growth of Ghettos and Urban Poverty

The Data

Australia has always had neighbourhoods that are clearly demarcated by income and socio-economic status. Nevertheless, the undesirability and adverse effects of low income neighbourhoods are not stamped on our national consciousness to the same extent that they are often stamped on the consciousness of citizens of other countries. US citizens, for example, are very aware of neighbourhood ghettos in their inner cities and are well aware of the potential for undesirable ghetto effects on residents (Wilson 1987, Case and Katz 1991).

We believe that income and employment gaps between our best and worst neighbourhoods are not as great as the gaps in many major OECD cities(2). We also believe that Australia is not in danger of creating urban problems to the same degree as the US (3), but, after the data were assembled, we were surprised and alarmed at the extent of the changes for the worse that have occurred since the mid 1970s.

There is not a great deal of data available to enable us to trace out changes in neighbourhood inequality over a significant period of time. The only consistent data base available is the Australian Census. There are four Census collections which include income data that could be used to measure changes in neighbourhood income distributions. Each Census - 1976, 1981-, 1986 and 1991 - coincides with an economic recession. By some measures, the depth of the recessions at each Census are not too dissimilar, but it is noticeable that the rate of unemployment is subject to an upward trend; 4.4, 5.6, 8.0, and 9.5 per cent respectively (4). Since unemployment is higher at each successive date we cannot use Census data directly to analyse income distribution effects of economic cycles and therefore we emphasise the trend from a comparison of 1976 with 1991.

To conduct the neighbourhood analysis the data are presented as group averages from Collectors Districts (CDs) which are the smallest geographical area for which Census data are available. CDs usually contain 200-300 dwellings which are delineated by easily identifiable boundaries. CDs tend to remain unaltered through time and in our sample we exclude those which were subject to boundary changes and not comparable across the four Censuses. The analysis is confined to CDs within major urban areas with populations of more than 100,000 (5). The panel consists of 9483 CDs and about 8 million people in each of the four years. There are no other comparable data sets which allow such a rich analysis of the changing geographical distribution of economic variables.

Although the Census provides by far the best data it is not ideal. Income data are not available by source. Consequently, it is not possible to investigate directly the role of government welfare payments. Another difficulty is that detailed geographic data are released as grouped means for specific variables and it is not possible for us to re-classify the data in many ways that would improve our understanding (6). Finally, the last Census maps economic circumstances at 1991. More recent data would probably show that the trends we are describing have continued but we will not know until the data from the 1996 Census are released sometime in 1997 or 1998.

The geographical analysis is based on CDs ranked by Socio-Economic Status (SES). We use the measure of SES calculated by the Australian Bureau of Statistics for 1986 (1990) (7).

Each CD preserves its SES ranking over the fifteen years. None of the results are affected by the choice of the Census year on which the SES ranking is based.

Neighbourhoods and Household Income

We begin by discussing the marked change in the dispersion of household annual income across neighbourhoods. In 1976, the ratio of the mean household (8) income of CDs from the lowest to the highest five percent of SES areas was 60.4 per cent. This is a fairly equal geographical dispersion of household income. Within the space of 15 years the ratio had fallen to 37.9 per cent. Income distribution has become more unequal and the change is extraordinary. There is a significant increase in the geographic polarisation of household income across Australia The poor are increasingly living together in one set of neighbourhoods and the rich in other set. The economic gap is widening (9).

Figure 1 arranges CDs from low to high socio-economic status (SES) and enables us to identify the pattern of income change across CDs. The CDs are ordered on the basis of their 1986 SES ranking. The first two bars on the left measure the change in mean income over the 1976-1991 period for the one and five percent of CDs with the lowest SES. The last two bars on the right measure the change in mean income from the top five and one per cent of CDs. All other bars refer to the change in annual household income averaged within each CD decile. Each decile includes approximately 500 thousand adults.

As we move across the CDs, from low to high SES areas, the pattern of income changes, measured in terms of 1995 prices, is quite smooth. For the bottom 70 per cent of CDs average household income has fallen in absolute terms and is lower in 1991 than in 1976. In areas of the highest SES household income has increased markedly. In the top five per cent of SES areas household income has increased $12,555 (23 per cent). In the lowest five per cent of areas household income has fallen by $7,589 (23 per cent). The income gap between the top and bottom S per cent of CDs has almost doubled and has widened by $20,144 (92 per cent).

This very significant pattern indicates that the forces making for increased income inequality across households exert a strong and systematic neighbourhood effect. These forces have either impacted upon individuals, according to the neighbourhood in which they live, and/or there is a continual geographic sorting process at work so that households which lose income are-moving to poor neighbourhoods and households which gain income are moving to high income neighbourhoods (10).

The narrow dispersion of neighbourhood household income in 1976, and the increased inequality since then, are so notable that is perhaps worth re-emphasising both facts by comparing household income from the top and bottom one per cent of CDs ranked by SES. In 1976, the weekly income gap between average household income from the bottom one per cent of CDs and the average household in the median CD was not large (Column 1, Table 1).

An additional part-time job for 9 hours per week at $12 per hour would close the gap. Facts such as these explain why most Australians believed that they lived in a fairly equal society in terms of income and employment opportunities. By 1991, however, an additional part-time job could still close the gap but it would need to extend to 19 hours per week, an increase of 10 hours. The bottom and median neighbourhoods are drifting apart and the gap has increased from $116 per week to $230 (1995 prices).

The increased income necessary to move from the average household income in the median CD to the average household income of a neighbourhood in the top one per cent of CDs is larger. The additional income cannot be obtained from the usual part-time job. In 1976, the additional weekly income needed was $442 and by 1991 this had increased to $885 a week. This is not a small step. In 1976, the additional income might be earned from an additional job which paid a little less than average weekly earnings. In 1991, the extra annual income required was $44,489, an income level which far exceeds average weekly earnings.

The increase in income inequality across neighbourhoods continued throughout the fifteen years (Column 3, Table 1) but the principal source of change differed. Between 1976 and 1981 increased inequality was generated by income falls in low SES neighbourhoods.

After 1981, the fall in income continues in low SES neighbourhoods but most of the increase in inequality is generated by income increases in high SES neighbourhoods. The source of the increased inequality appears to have been shifting from large income falls in the low SES neighbourhoods, relative to the median, to large increases in the high SES areas, relative to the median (11). In these days of Super Leagues and large payments to stars in all fields of endeavour the income gap across neighbourhoods might be expected to continue to increase and be driven primarily from increased income in high income neighbourhoods.

To understand further the sources of the increase in inequality of household income the change can be allocated into three parts; (i) the change in average male income of a CD (ii) the change in the average female income of a CD and (iii) a residual which combines the effects of the change in the average number of adults per household and the effects that arise because the average household does not consist of one adult male and one adult female.

Figure 2 documents the first source of change; the change in the male mean annual income of CDs ranked by SES. In the five per cent of CDs with the lowest SES, male annual income fell between 1976 and 1991 by $4102 (1995 dollars). In the top five per cent of CDs average male income increased $916. As a result, the male mean income gap between CDs from the lowest and highest SES widened by $5018.

In 1976 the average male income in CDs from the lowest five per cent of SES areas was 54.9 per cent of the mean income in the highest 5 per cent of SES areas. By 1991 this income ratio had fallen to 42.5 per cent; a change not too dissimilar from the change in the household income ratio. The 1976-1991 period saw little growth of male income. It is noticeable that only 20 per cent of CDs from the highest SES areas experience male income growth over the 15 years. In 80 per cent of neighbourhoods there are real income falls.

The income changes for women also exhibit a smooth pattern across CDs (Figure 2) but, in this instance, the mean annual income substantially increase in all but the lowest one per cent of CDs ranging from $996 for the five per cent of CDs from the lowest SES areas to $6319 for the 5 per cent of CDs from the highest SES. Women's contribution to the income of a CD has offset the fall in male income, at least in part, in all but the lowest 1 per cent of CDs.

Income distribution has widened for women. The income level of women in the lowest to the highest 5 per cent of CDs, ranked by SES, has changed from 78.8 per cent to 57.8 per cent. Once again a change similar to the change in the household income ratio. The contribution of the increase in women's income has been to further widen the absolute income gaps across SES areas.

Since 1976 women as a group have generally experienced growing income levels. More women are employed, the welfare state has expanded the income sources of low income women through family allowance supplements and sole parent payments and women's wages have been increasing relative to male wages. In CDs from the lowest 5 per cent of SES areas the female to male income ratio has increased from 50 to 70 per cent. In CDs from high SES areas the increase has been from 35 to 51 per cent.

In all periods women's income has increased relative to that of men in all CDs but the pattern differs a little through time. In the early period the largest increases, relative to men occurred in the high SES areas. In the later period the largest increases, relative to men, have occurred in the low SES areas.

It is evident that income inequality has increased across neighbourhoods for both genders and by substantial amounts. Indeed, if the changes in male and female mean annual income are added together, to measure the change in income within a CD, the gap between the bottom and top five per cent of CDs has increased by $12,335 per couple: a very large change indeed but $7809 less than the increase in the gap of household income. The $7809 accounted for by the-residual term, in part a reflection of household size, is an important part of the explanation of the change in household income.

The contribution of the change in male, female and residual income to the change in household income between the bottom and top 5 per cent of neighbourhoods is given in Row I of Table 2. The contribution of the change in income of the average male and female to the widening dispersion of household income is approximately the same, each accounting for about one third of the household income change. The residual effect also accounts for one third of the change.

One important source of the residual effect on household income is the change in the number of adults per household. There is a clear neighbourhood pattern. The average household size has fallen in CDs from low SES areas. This has been generated in part by the growth of single parent households which have increased from 7 to 17 per cent of households in CDs from the lowest 5 per cent of SES areas. In households from the top 5 per cent of SES rankings the proportion of single parent households has increased marginally from 3 to 6 per cent. Household size has increased in high SES areas.

The other rows of Table 2 measure the change in household income at each end of the SES ranking. Row 2 presents calculations for the change in household income for the bottom S per cent of CDs over the 1976-1991 period. Household income in real terms fell by $7589 over fifteen years. This fall is accounted for in equal parts by the fall in male income and a reduction in the contribution of the residual term, primarily the fall in household size. These factors account for $4102 and $4484 respectively. For this group there has been a small increase in women's income. Women are an increasingly important source of income for low SES areas, S because their income has increased substantially, but because male income has fallen so much.

Row 3 lists the increase in household income in the top 5 per cent of CDs. Here the major contributor has been the $6321 increase in the income of women and an increase in the residual contribution of $5318. The increase in male income made a small contribution of $916. The increase in women's income has substantially added to the real income of high income neighbourhoods .

2 Employment changes and the Increase in Income Inequality across Neighbourhoods

The change in male and female employment -population ratios

For most households the principal source of income is employment. The relatively narrow income dispersion across neighbourhoods in 1976 was generated to a significant degree by similar employment-population ratios across neighbourhoods. For men there was no systematic variation in employment-population ratios across CDs ranked by SES (Figure 3). For women, the employment-population ratio in 1976 was marginally less in low SES CDs and the employment-population gap between the lowest and highest 5 per cent of neighbourhoods was small (Figure 4).

In 1976, irrespective of where Australians lived, they shared much the same commitment and access to employment. A social observer could walk across the best and worst parts of Australian urban areas and although the probability of meeting some one who was employed differed by neighbourhood there was no systematic change by socio-economic status. Income inequality across neighbourhoods ranked by SES was generated by different levels of income from all activities and not from differences in the proportions of the population employed.

By 1991, circumstances had changed dramatically. Australian employment growth between 1976 and 1991 had been very poor. The average employment-population ratio fell 15 per cent for men and increased 17 per cent for women. Furthermore, there has been a marked shortage of full-time jobs After adjusting for population growth full-time jobs fell 12 per cent. The lack of employment growth has been reflected in unemployment which increased from 4.7 to 9.5 per cent (12).

This poor employment performance is evident in the pattern of change of neighbourhood employment. In all neighbourhoods the employment-population ratio for men has fallen: 20 per cent in CDs from the top 5 per cent of SES neighbourhoods and 42 per cent in the 5 per cent of CDs from low SES neighbourhoods. It is to be expected that some employment falls would be observed among men - given trends for early retirement and for increased years to be spent in schooling - but the pattern across neighbourhoods is stark.

The pattern of employment change for women is similar but the contrast across neighbourhoods is greater. For the top half of neighbourhoods the proportion of women employed increased approximately 10 per cent. For the bottom half of neighbourhoods employment fell by 40 per cent. We were taken aback by this fall. We are so used to seeing macro data which indicate a rapid growth of part-time work for women, and reading about women's increased-labour force involvement, that it is a shock to see that in 1991, and for half of Australian neighbourhoods, the proportion of women employed in the labour market is substantially less than in 1976. All the growth of the women's employment-population ratio is to be found among women who live in neighbourhoods in the top 50 per cent of high SES areas. By 1991 the probability that a women was employed if she lived in the top 5 per cent of SES neighbourhoods was almost twice the probability that a women was employed in the lowest 5 per cent of SES areas. The next step in the research agenda will be to explain this change. Some factors to consider would be the growth of sole parents - they tend not to be employed and are increasingly concentrated in low SES neighbourhoods -, the very marked tendency for women of unemployed men to be unemployed and any potential for not reporting employment activities among those dependent on welfare.

It is apparent that employment-population ratios are now a major contributor to income variations across areas. For males, Australia has returned to neighbourhood employment patterns of the 1930s where there are substantial ghettos of non-employment. For women however, the pattern is quite different (Gregory et al 1987). In the 1930s there was little variation of female employment-population ratio across neighbourhoods ranked by SES. The pattern was much the same as in 1976. Employment opportunities for women appeared to be spread fairly evenly. The large loss of womens employment in low SES areas by 1991 is quite interesting and worth pursuing.

The Two Australia's

Neighbourhoods in 1991 can be divided into two groups. For neighbourhoods taken from the top 20 to 30 per cent of CDs, ranked by SES, the employment-population ratio of men and women does not change significantly across neighbourhoods and there is no close relationship between employment level changes and income changes (Figure 5, Figure 6). Income dispersion within this group is related more closely to variations in wages and salaries, and earnings from own business rather than variations in employment rates. For our social observer walking through the top 20 to 30 per cent of neighbourhoods the S of employment has changed since 1976 but the pattern of employment across CDs has not. Employment-population ratios continue to vary systematically across neighbourhoods by SES and not to be related to income changes.

For the remaining 50 to 60 per cent of neighbourhoods employment rates now matter. The world has changed and there is now a clear association between employment changes and income changes. Within the group the translation of employment changes into income changes is similar for both men and women. On average an increase in employment by 15 percentage points adds $2300 to male income and $1000 to female income of a neighbourhood (Figure 6).

The widening of the income distribution across neighbourhoods therefore is being driven by different influences at different ends of the income distribution. Employment is strongly associated with income in low income neighbourhoods but not for high income neighbourhoods.

The Census data suggest therefore that those from low income neighbourhoods are increasingly losing contact with the world of employment. The pattern of high rates of joblessness in CDs of low SES is very disturbing.

The joblessness in low SES areas begins with teenagers. In 1991, the employment rate of teenagers in low SES areas is 80 per cent of that of high SES areas, even though most teenagers in high status areas are attending an education institution. Within the age group 20-24 years the employment rate of the bottom 5 per cent of CDs has fallen to 63 per cent of that of the top 5 per cent of CDs and remains there until the age group 45-54 year old where the employment rate falls further. The pattern is the same for men and women (Figure 7). It is remarkable that in 5 per cent of CDs from the low SES areas, which include half a million persons, one half of the men 2544 years are not engaged in full-time employment. Indeed, the employed proportion of the male population of low SES areas in every age group is lower than the employed proportion of females from high SES areas (Figure 8).

3 Policy and Urban Inequality

Although we are very concerned about the rapid growth in income inequality across neighbourhoods it is nevertheless true that there is no "right" degree of urban inequality. Nor is it clear that policy can efficiently and effectively achieve the urban inequality we might choose. In the past Australia has not placed high priority on policies specifically directed towards reducing urban inequality and our experience of policy effectiveness in this area is limited. Policy has been more concerned with income distribution and unemployment among individuals. It has been implicitly based on the premise that if inequality is reduced among individuals it will be reduced across urban areas. This premise seems incorrect. There has been no "trickle down" of macro economic growth to the unemployed and low income earners in low SES areas since 1976. What might be done if we are dissatisfied with a situation where, in 1991, male unemployment is as high as 35 per cent in many neighbourhoods? How might we return to something approaching the distribution of neighbourhood income in 1976? This is obviously an important and challenging topic of research.

It is not possible to answer these questions without some understanding as to the underlying causes of the growth of urban inequality. There is a range of possible causes and we do not have space to discuss them all. We focus on three.

The first and obvious point is that perhaps the influences that are generating inequality outcomes have nothing to do with neighbourhoods. When the macro economy performs badly the poor become unemployed and they just happen to live in poor neighbourhoods. Furthermore, for future employment prospects perhaps the neighbourhood in which you live does not matter. It is more likely, however, that neighbourhoods do matter. Consider an example.-- Suppose that individuals in low SES neighbourhoods were employed disproportionately in manufacturing, as they were, and that factories are located close to home. Then, as a result of the large contraction of the manufacturing sector over the 1976-1983 period, many of these factories close and the neighbourhood loses its most important source of livelihood. Furthermore, suppose the transport system is inadequate to that it is difficult to reach new jobs at low expense in areas further away. In addition, suppose most people find jobs through contacts with employed friends and relatives. Now that unemployment is so high in the area the information on employment opportunities will be less and individuals will find it harder to find employment because of where they live. A spatial mismatch would have arisen between the location of the population and the location of jobs. The solution would seem to lie in better transport and job finding net works that enable individuals to travel and find work further afield.

Boyd has spent considerable time looking at the importance of the decline in manufacturing and it appears to be true that it has disproportionately reduced employment in low SES neighbourhoods. With regard to the breakdown of job networks we have not yet had time to investigate this possibility

Another important factor could be that the demand for labour is moving away from unskilled workers towards workers with skills and education. The unskilled live in low SES areas and the skilled in areas of high SES but there is no specific neighbourhood effect. It takes time for individuals to upgrade skills and so we observe falling income and employment in low SES areas. In this instance, it might be better if policy were directed towards enhancing worker skills wherever workers live.

Finally, increasing inequality may be the result of major structural problems in the macro economy - such as emerging inflation or balance of payment difficulties - that lead to insufficient jobs being created. If there is insufficient employment opportunities then workers compete for scare jobs and it is to be expected that the least skilled will tend to miss out. It is further expected on average that those who are last in the queue will tend to live in disadvantaged areas. Disadvantaged areas are low rent areas and their inhabitants have always been among the more marginal of workers and among the low paid. In these circumstances we should look to improve the macro-environment.

The analysis of the role of the manufacturing decline is somewhat complicated and will not be discussed here but the basic research can be found in Hunter (forthcoming). Instead we focus on the final two points, changing demands for skilled workers and general influences of macro economic policy.

Education Policy

-

Many countries, including Australia, have attempted to use an expansion of education and skill training to offset growing income inequality and unemployment among the low paid. Students have been offered means tested living allowances for high schools and tertiary education and interest free loans to pay university fees. Tertiary and high school places have increased substantially. Indeed, over the last decade and a half, Australia has embarked upon one of the most ambitious education programs in the OECD.

This education expansion has had a large impact on the average neighbourhood from areas of median SES. Between 1976 and 1991 the proportion of the population with degrees increased from 3.7 to 14.7 per cent and the proportion of the population without qualifications fell from 66 to 45 per cent. And yet, despite this large increase in education of the potential workforce, male unemployment in median neighbourhoods has risen from 4.4 to 13.0 per cent. In addition, average income per adult has risen by less than one half a per cent per year.

Although, income and employment outcomes may have been worse without education increases, it appears, nevertheless, that increased education levels have not been sufficient to offset significant employment losses or to generate significant income increases for the median neighbourhood. Education and skill training may primarily determine who gets jobs and may have very little influence on the number of jobs available or average rates of pay.

A similar sober assessment also appears inescapable from a comparison of the changing interrelationship between education levels and income inequality among neighbourhoods. The various measures of the education of a neighbourhood's residents are highly correlated and for our analysis we use the proportion of residents 15 years and over with a degree (13).

In 1976, there was a strong positive association between the average education level of a neighbourhood and the income of its residents. On average, a one percentage point increase in degree holding among men was associated with additional income of $1000 (Figure 9). For women, the relationship was $500 for each additional percentage point increase in the proportion of the population with degrees (Figure 10). Among neighbourhoods, as among individuals, higher education brings higher income.

It is noticeable, however, that there is no systematic relationship between employment--population ratios and education for either men or women. More education is associated with more income but not because employment is increased. This is a restatement of the fact that in 1976 employment opportunities were distributed equally across neighbourhoods ranked by SES.

By l991, the relationships have changed a great deal. For men, more education is still positively associated with more income but the relationship has shifted so that for any given proportion of the population with degrees the income level has fallen by about $8000. If the employment-education relationship can be thought of as a causal one then to achieve the same level of male income as in 1976 a neighbourhood needs to achieve a higher education level. Consider a neighbourhood from a low SES area. To maintain male income this neighbourhood needed to increase the proportion of its male population with degrees by 6 percentage points between 1976 and 1991. The actual increase was 2.5 percentage points, hence the fall in male income. In high SES areas the increase needed in the proportion with degrees was around 8 percentage points. The actual increase was 9 percentage points, hence the increase in male income.

This shift in the education-income relationship is very important. On the basis of the 1976 relationship between the incidence of degrees, and the income of a neighbourhood, the increased education attainment of the average neighbourhood within the bottom five per cent of CDs should have brought about an income increase of $3500. In fact, there has been a fall of $6000. The $9500 gap clearly illustrates the importance of the change.

The principal source of the shift in the male education-income relationship is a shift in the employment-education relationship. For neighbourhoods from the bottom 70 per cent of SES areas the education-employment relationship has moved down but, in addition. there is now a strong neighbourhood relationship between less neighbourhood education and less neighbourhood employment. A relationship that did not exist in 1976. The lower the male education level of a neighbourhood the lower the male employment-population ratio. Education not only affects income, as it always has, but now it also affects the employment-population ratio. Poor neighbourhoods are now twice disadvantaged by low education levels.

For neighbourhoods from the top 30 per cent of SES areas further education does not bring further employment. For these neighbourhoods nothing has changed with respect to changes in education and changes in employment. But the education-employment relationship has also shifted downward so at each neighbourhood education level there is 15 percentage points less employment.

Women's labour market changes are similar to the male labour market changes in all but one respect. The difference is that the education-income relationship has changed little since 1976 except in areas of low SES where additional degrees among residents have not brought neighbourhood income increases, But, unlike the male relationship, the large increase in womens income across all but the low SES areas is associated with the large increase in education. There has been no systematic shift down in the employment-income curve as in the male market.

There is a clear dichotomy among neighbourhoods and the two Australias are evident. For the top 30 per cent of SES areas income has fallen for each education level for men but increased for women, The relationship between changes in income and changes in education however has not shifted for this group.

For-the remaining 70 per cent of neighbourhoods the less the education level the greater the income fall. Employment and education are now associated and hence there is less income at each education level.

To conclude, we look at the change in the distribution of education levels across neighbourhoods to assess the general impact of the large increases in education levels of the potential workforce. In 1976, 10 per cent of all residents 15 years and over who resided in CDs from the top 5 per cent SES possessed degrees. Now the proportion is 20 per cent. In the lowest five per cent of CDs in the proportion of the population with degrees has increased from 0.5 per cent to 3 per cent. The absolute gap in the degree distribution between areas has widened and the increased incidence of degree qualifications has been disproportionately concentrated in CDs from high SES. Neighbourhoods have not become more equal. For every ten new degree holders in the top 5 per cent of CDs there has been an additional three in low SES areas. A similar pattern is evident if different measures of education are used.

It is not as though areas of low employment and low income have been untouched by the expansion of education. Education levels have increased. The difficulties are two fold. First, the increase in education in absolute terms has been greater in high SES areas so that inequality has increased. Second, the relationship between employment and education levels has shifted in low SES areas that a given level of education now delivers much less income than before and the move to a more disadvantageous relationship has dominated the improvement in the education level.



Macro Policy, Wages and Employment Bias

We begin by stating the obvious. A necessary condition to reduce urban inequality to levels more closely approaching those of the mid 1970s is that macro policy must be directed towards strong employment growth. Furthermore, it seems obvious that a rate of aggregate job creation that implies unemployment rates above 6 per cent will not be sufficient. The change in urban inequality between 1976 and 1991 occurred in an environment where the unemployment rate averaged about 6.9 per cent.

It is clear from the pattern of neighbourhood job loss presented in Figures 4 and 5 that if urban inequality is to be reduced there must be a bias in new job growth towards those who live in depressed neighbourhoods. The extent of the bias needed is substantial. If we make the unlikely assumption that participation rates are fixed and we wish to return to the 1976 unemployment relationship then for each additional job taken up in the top 5 per cent of SES areas about 12 jobs are needed for those from the lowest 5 per cent of SES areas. It seems unlikely that macro policy alone will be able to achieve this outcome.

First, even if the economy continues to create jobs at a fast rate, and for a sufficiently long time, the economy is unlikely to generate the neighbourhood job bias needed. Long run employment trends seem to be against the unskilled, the lowly educated and those who live in depressed neighbourhoods.

Second, it also seems likely that the new pattern of unemployment across neighbourhoods will make it more difficult to pursue fast economic growth without the development of inflationary wage pressures. It seems likely that as the economy expands that inflationary wage pressures will emerge from people living in good neighbourhoods and the growth cycle will come to an end before new jobs extend to depressed neighbourhoods.

The problem that we have in mind can be discussed with the help of the following simple diagram. Figure 11 presents the unemployment pattern across neighbourhoods in 1976. Unemployment is higher in areas of low SES but the 5.0 percentage point unemployment gap between neighbourhoods from the highest and lowest 5 per cent of neighbourhoods ranked by SES is small. Suppose now that the economy grows sufficiently fast to create enough employment in aggregate that the number of jobs offered to those who live in a neighbourhood from the median SES group just equals the number of unemployed there. Furthermore, assume that there is no bias in initial job offers so that they are spread equally across all neighbourhoods (Line A). Under these circumstances those who live in neighbourhoods with income above the median will receive more job offers than there are unemployed people - the vertical distance between the two lines. This will generate two responses. One response is that as job offers are unfilled they will move down the neighbourhood rankings and offers will spill over to those who live in poorer neighbourhoods. This is the "trickle down" effect of economic growth which seemed to operate throughout the 50's and 60's. The other response is that there will be pressure for wage increases among those who live in good neighbourhoods and find that job offers are plentiful. These two effects are obviously inter-connected. The greater the spillover effect to those in low SES areas the less the wage pressures created by any given level of job offers.

The wage pressure emanating from those who live in good neighbourhoods depends on the position and slope of the unemployment curve and line A, the job offer curve. The slope of the unemployment curve provides a measure of the degree of substitutability of labour across areas. A flat unemployment curve suggests there has been substantial "trickle down" of job offers. The more substitutable labour the more similar the rate of unemployment should be across areas. If the unemployment curve rotates and becomes steeper so that unemployment increases in low SES areas and decreases in high SES areas this will suggest that labour from different SES areas has become less substitutable for each other. Since 1976 the unemployment curve has moved in ways which suggest that job offer spillovers have become weaker and, as a result, it has become more difficult for macro-economic policy alone to achieve full employment without generating inflationary pressures (Figure 12). The unemployment gap between those who live in high and low socio-economic status areas has increased. The difference in male unemployment rates from the bottom to the top 5 per cent of SES areas has increased from 5.0 percentage points in 1976 to 17.5 percentage points in 1991.

The slope of line A measures the bias in the job offers across areas. If line A is horizontal there is no neighbourhood bias. Over the last two decades it is probable that the slope of line A has been positive and jobs have been disproportionately offered to inhabitants of high SES areas. These individuals are the better qualified and more skilled. As a result of this bias the wage pressure has been greater than if the job offer line was horizontal or negatively sloping.

There are reasons to believe that as time passes the unemployment rate curve will become steeper and job offer spillovers may become weaker. If there is increasing disconnectedness with employment opportunities as time passes the unemployment curve will become steeper and if this were to occur then macro policy would become less effective in its attempt to increase more jobs without inflation. Macro policy needs to be accompanied by successful labour market interventionist policies to bring residents of depressed areas back into the labour market. If there is a widening of income inequality and job opportunities across urban areas then the internal dynamics of depressed areas may be increasingly creating islands which are largely outside the main traffic routes of economic growth. Unemployed people in depressed neighbourhoods may be associating primarily with other unemployed people and as a result may not hear of available job opportunities. Most individuals find new jobs by being told of opportunities by friends or relatives (14). It may also be that individuals living in depressed neighbourhoods develop behaviour patterns that make it difficult for them to be successful in job search.

Wages Policy

There are two obvious wage policy reactions that might be made to the above analysis. The first could be thought of as a macro response. If we could control wage increases for those who live in high SES neighbourhoods, so that their response to excess demand does not lead to an outbreak of wage inflation, the economic cycle may continue longer and there will be more job offer spillovers to those who live in low SES areas. This is essentially the policy that was followed by the Accord process throughout most of the 1980s and early 1990s.

The second reaction, which might be thought of as a micro response, is to attempt to flatten the unemployment rate curve and change the job offer curve by lowering wages of those who live in low employment neighbourhoods. This might be achieved by deregulating the labour market so that the wages and income of those who live in low SES areas will fall further but be offset to some degree by increased employment opportunities. The lower wages will either create more low paid jobs and/or divert some wage offers away from higher priced labour.

It is not known how much wages might need to fall. To increase employment of the bottom 5 per cent of SES areas back to 1976 levels, relative to high SES areas, would require at least a 44 per cent increase in male employment and a 70 per cent increase in female in employment. It appears likely therefore that a substantial wage fall would be required. This raises a number of problems. First, it takes time to create jobs so that the short run wage fall might be substantial. So substantial in fact that individuals may prefer not to work and be supported by unemployment benefits and other welfare payments and perhaps a range of black economy activities. If wage reductions were to occur this were to occur on a substantial scale, it might be expected that governments would eventually react and reduce benefit levels on the grounds that the increased benefit levels, relative to low wages, were discouraging individuals from accepting jobs.

Benefits are the main source of income for most individuals in low SES areas and any reduction must inevitably increase poverty and widen income distribution further. It is obvious why governments and communities are reluctant to go down the path of substantial reductions in wages and benefits and why it is often suggested that it might be better to try and increase the employability of individuals in low SES areas rather than reduce their potential wage. The main policy instruments to increase employability have been-developed in the context of the Working Nation statement and include wage subsidies for the long term unemployed and increased education and training for the low skilled. The neighbourhood analyses of this paper strengthens the support for these programs and we hope that subsequent evaluations of the Working Nation initiatives indicate that these programs are effective.

4 Concluding Remarks

Since the early 1970s the Australian economy has had a major job creation problem. According to the Census the proportion of men 15-64 years employed in a median neighbourhood is 23 per cent less than in 1976. The proportion of women employed is one per cent more. The shortage of jobs has not been rationed evenly throughout our society. Job loss and income falls are concentrated in low SES neighbourhoods and job growth and income rises are concentrated in neighbourhoods of high SES.

Between 1976 and 1991, the lowest one percent of neighbourhoods, based on a 1986 SES ranking, have lost one third of their employment, 23 per cent of their household income and male unemployment has increased from 6.4 to 28.1 per cent. The contrast with areas of high SES is marked. In areas of the highest SES, employment has fallen marginally, household income increased 31 per cent and male unemployment increased only to 4.8 per cent. The proportion of women employed in high SES areas now exceeds by 10 per cent the proportion of men employed in low SES areas. The household average income gap between the two areas has increased by $26,580.

To lose employment and to suffer significant income losses are bad outcomes for anyone but does it matter that these undesirable outcomes increasingly possess a geographic component? It is sometimes suggested that it does not and that nothing is gained by knowing that it is people who live in poor neighbourhoods who are increasingly not at work, that part-time jobs are going to young people and women who live in high SES neighbourhoods and that income is rising in the best SES neighbourhoods but falling in poor neighbourhoods? Our intuition suggests that neighbourhoods do matter. It seems likely that the greater the economic polarisation within our cities the less equal are the opportunities for young people and the more likely that bad neighbourhood pathologies will emerge.

But what should be done? It is not easy to know. There has not been a strong Australian tradition of thinking about economic policy and neighbourhoods and it is not always easy to move from thought patterns that revolve around individuals or the macro economy to thought patterns that stress geography. The necessary economic concepts and models that are needed for a geographic analysis do not role of the tongues of economists with quite the same ease as the macro concepts that we have become used to over the last two decades. It is partly as a response to this imbalance that we have spent so much time on laying out the basic numbers and the extent of the change that is occurring in our society.

Steve Monaghetti is alleged to have said about marathon racing that " There is a beginning, a middle and then another beginning". The same remark might apply to economic research. There is always more to be done and we do not really believe that we are even at the middle yet. We are very conscious that we do not know enough about social and geographical mobility, changing income and employment opportunities over the life times of people who live in poor neighbourhoods and the role of job finding networks in poor neighbourhoods. The Census data can only play a very limited part in our pursuit of these issues. Boyd, however, who is finishing his thesis next week after three and bit years is at the stage that he just wants to finish. He does not like the Monaghetti remark. He prefers to say "There is a beginning a middle and an end".

To conclude I hope that we have given you an idea of some of the work that is done within the economic group at the RSSS and some inkling of the opportunities we offer young scholars to research into matters of vital concern to our society.

References

Borland, J., "Wage Inequality in Australia". National Bureau of Economic Research, Boston, 1992

Case, A and Larry Katz, "The Company you keep: The Effects of Neighbourhood on Disadvantaged Youth", NBER Working Paper 3705, May 1991.

Gregg, Paul and Jonathon Wadsworth How Effective are State Employment Agencies? Job Centre Use and Job Matching in Britain. (Mimeo) Working Paper No. 476

Gregory, R.G., "Aspects of Australian Living Standards The Disappointing Decades 1970-1990" Economic Record, Vol No 69 No 204, 61-76 March 1993

Gregory, R.G., V. Ho, L. McDermott and J. Hagan, "The Australian and US Labour Markets in the 1930s" in B. Eichengreen and T. J. Hatton (eds), Interwar Unemployment in International Perspective, Kluwer Academic Publishers, London, 1987

Jenks, Christopher and Susan E Mayer, The Social Consequences of Growing up in a Poor Neighbourhood; A Review in L Lynn and M. McGeary (eds) Inner City Poverty in the United States National Academy Press, Washington, DC, 1990 111-187

Montgomery, Holmes D.," Social Networks and Labor-Market Outcomes: Towards an Economic Analysis". American Economic Review, Vol 81 No. S Dec 1991 1408-1417

Saunders, Peter, Welfare and Inequality, Cambridge University Press 1994

Wilson, William J ,The Truly Disadvantaged, Chicago, The University of Chicago Press, 1987

Socio-Economic Indexes for Areas, Australian Bureau of Statistics, Cat. No. 1356 1990

Footnotes

(1) There is no consensus, however, as to the source of these large changes. They seem to be related to shifts in labour demand away from men and towards women workers and away from unskilled workers towards those with higher education levels. There are some areas of agreement among researchers as to what is S driving the increased inequality. It does not seem to be the case that inequality among individuals is being driven primarily by the decline in manufacturing, the growth of trade with Asia or immigrant flows of low skilled labour. We are more agnostic.

(2) We have not been able to find any cross-country comparisons of differences in inequality across urban neighbourhoods .

(3) The poverty of the US ghettos is compounded by the concentration of disadvantaged Americans of African descent (see Wilson 1967). Another contributing factor is the US Federal system that places emphasis on local taxes as a revenue source. The Australian Federal system, in contrast, is a force for equalising income and government services across neighbourhoods.

(4) Unemployment at August each Census year taken from the Labour Force Survey.

(5) CDs were omitted from the panel if the total population was less than 50 to avoid the sampling error deliberately introduced by the ABS to protect the confidentiality of persons in the neighbourhood. In each successive Census new CDs are added and in some circumstances the boundaries of CDs are changed. Our sample is a fixed number of CDs with unchanging boundaries that are to be found in each Census plus a small number where the CD may have been divided in-to two. We begin with a list of CDs from the 1986 census and if there was more than one CD that corresponded to the 1986 CD then the first was taken to be representative of the 1986 CD.

(6) The data are released by the Australian Bureau of Statistics as group averages within each neighbourhood to protect the confidentiality of individual census returns. As an example of the difficulties of data that are released as group averages consider the following: data are available for mean income and mean employment within each CD but not mean income per employed person and there is no way that we can accurately calculate this figure.

(7) We use the Urban and Rural Indexes of Relative Advantage. The Indexes are calculated by the application of Principal Components. The relevant variables include data such as family income greater then 50,000, the proportion of CD residents with degrees, the occupational distribution of the employed workforce and the number of bedrooms per person.

(8) A household consists of a person living alone, or two or more related or unrelated persons who live and eat together in private residential accommodation.

(9) The data collection method for the Census is for household members to complete a questionnaire. Many other ABS data sources are collected by household interviews. There are systematic differences in data according to the collection method. The Census tends to understate income and employment of those whose involvement in the workforce is peripheral.

(10) An investigation of the coefficient of variation of income within neighbourhoods suggests that within neighbourhoods income allocation across households and individuals is not becoming more homogeneous. This suggests that we are observing household income changes within neighbourhoods, ranked by SES, rather than a sorting phenomenon which is reallocating households across neighbourhoods.

(11) Between 1976 and 1991 there has been little further change in the inequality of household income of the bottom 79 per cent of neighbourhoods. But there has been no clawing back of the large losses of income among low SES neighbourhoods in the 1976-1986 period.

(12) These data are taken from the Labour Force Survey August 1976 and August 1991.

(13) Other possible variables include per cent with a qualification, per cent who leave school at fifteen years or less and average number of years of schooling.

(14) In a recent UK study Gregg and Wandsworth (1995) show that the most successful method utilized by unemployed males to find a job is through Friends and Contacts. The utilization rate of this method is not the highest but it has the highest success rate. Among males one third of jobs are found this way. Among women one quarter of jobs are found from this method. Montgomery (1991) estimates that 50 per cent of all workers currently employed in the US found their jobs through friends and relatives.