The Number Cruncher

Education and income inequality in Australia

Posted on May 14, 2013 by nicholasbiddle

Human capitalism and trends in US income inequality

I would be the first to admit that I have an absolutely fantastic job. Sure, it is cathartic and sometimes strategic to complain about university bureaucracy and the endless search for funding. However, the reality is that to be paid to read, write and speak about ideas, data and policy is a privilege not available to many.

I had occasion to reflect on this recently when I was reading a new book just published by Princeton University Press titled Human Capitalism: How Economic Growth Has Made Us Smarter – and More Unequal. The author, Brink Lindsey, is a senior fellow at the Cato Institute, a think-tank in the US that markets itself as being ‘dedicated to the principles of individual liberty, limited government, free markets and peace.‘

Before getting to the main substance of the book, it is worth outlining a few trends in the US. The first is that most evidence would suggest that inequality in the US has been rising over the last few decades, especially at the upper end of the distribution. For example, the Congressional Budget Office (CBO) found that ‘between 1979 and 2007, income grew by 275 percent for the top 1 percent of households, 65 percent for the next 19 percent, just under 40 percent for the next 60 percent, and 18 percent for the bottom 20 percent.’

In another summary finding, the authors of the CBO paper demonstrated that most of that growth was due to a widening level of inequality in market income (that is, before taxes and benefits are taken into account). Using the Gini coefficient, a summary measure of income inequality that ranges from 0 (perfect inequality) to 1 (perfect inequality), inequality in market income rose from 0.479 in 1979 to 0.590 in 2007. This ‘great divergence’ is in contrast to the ‘great convergence’ in incomes that took place in the immediate post-war period.

So, what is Lindsey’s explanation for this rising inequality? In essence, it is an interesting version of ‘skills-biased technological change’ which is a ‘shift in the production technology that favours skilled over unskilled labour by increasing its relative productivity and, therefore, its relative demand‘ and ‘assortative mating’ where individuals with similar characteristics (in this case education) are becoming more likely to partner with each other. Where Lindsey differs somewhat in his explanation is a focus on the role of complexity in explaining these trends. The following few quotes should hopefully summarise his argument:

‘today the primary determinant of socioeconomic status is the ability to handle the mental demands of a complex social environment’ – p.3
‘economic growth breeds complexity, complexity imposes increasingly heavy demands on our mental capabilities, and people respond by making progressively greater investments in human capital’ – p. 4
‘Among families with college-educated parents, marriages have been growing more stable, parents have dramatically increased the time they spend with their kids, and they are spending that time in a dedicated effort to foster the intellectual, social, and personal skills needed to thrive in an ever more complicated world.’ – p.66

Although I am not completely convinced by all the policy prescriptions at the end of the book, it is an interesting argument. Which got me thinking about what is happening in Australia with regards to education and income inequality.

Changes in income inequality in Australia

The first thing to note about the changes in income inequality in Australia over recent decades is that they are less dramatic and, unfortunately, less consistent across data sources than they are in the US. It would appear from an analysis by Roger Wilkins of the Melbourne Institute that measured inequality using the Australian Bureau of Statistics’ income surveys that the Gini coefficient in Australia rose from a little over 0.30 in 1993/94 to a little under 0.34 in 2008/09. However, there have been a number of revisions to the methodology used by the ABS to collect income which may be driving some of these trends.

When you use a more consistent set of income measures (albeit over a shorter time period) in the Household Income and Labour Dynamics in Australia (HILDA) survey, income inequality seems much more stable. This is demonstrated in the following figure which gives the Gini coefficient for personal disposable income (that is, after taking into account taxes and benefits) as well as household equivalised disposable income (after also taking into account household pooling of income).¹

HILDA inequality through time

Household surveys like the HILDA are a little restrictive as to what they can say about the extremes of the distribution. It is important to note, therefore, that Wilkins also shows using Tax and National Accounts data that the share of income going to the top 1% of the income distribution rose from at little over 6% in 1989/90 to around 8.5% in 2009/10. Nonetheless, it would appear that Australia has not experienced as large a rise in income inequality over the last few decades as the US.

The contribution of education to income inequality in Australia

What then can we say about the contribution of variation in education in Australia to income inequality? Well, the first thing to note is that according to the HILDA survey (and all other datasets), education levels in Australia increased substantially between 2001 and 2010. This is demonstrated in the following figure which gives the weighted per cent of the 2001 HILDA sample by their education level (in black) alongside the same figure for the 2010 HILDA sample.

Highest education by year of survey

The per cent of the (weighted) sample who had completed Year 9 or less fell over the 9 years between Wave 1 and Wave 10 of the survey. This fall was greatest amongst those who also did not have any qualifications. There was also a fall amongst those who had completed Year 10 or 11 but did not have any qualifications. At the other end of the distribution, there were large increases in those who had completed Year 12, as well as those with a bachelor or post-graduate degree.²

Between 2001 and 2010, all education types experienced an increase in average income. This is demonstrated in the following figure which gives the percentage change in average disposable personal income over the period (in black) as well as the percentage change in average disposable household equivalised income (in grey).

Change in average income by highest education

Between 2001 and 2010, average personal disposable income for those in the HILDA who had completed Year 9 or less increased by only a small amount (7.7 per cent). Increases were greater for all other education groups, particularly for those with Year 12 and a non-degree (other) qualification (22.8 per cent), those with Year 10 or 11 and a non-degree qualification (21.3 per cent) and those with a post-graduate degree (20.3 per cent).

One way to look at the net effect of such changes is through a decomposable measure of income inequality (the Theil index) which apportions income inequality into the amount of inequality within education levels and income inequality between education levels. I used such a technique in a recent post on inequality within and between urban areas.

In the case of education, the within component is each education group’s inequality level, weighted by their contribution to total income and summed across the population. When I decompose inequality within and between education groups in HILDA, I find a very stable within component. In 2001, it was 0.297 for disposable personal income and 0.155 for disposable equivalised household income. In 2010, the figures were 0.301 and 0.152 respectively. Despite rapid income growth between 2001 and 2010, variation in income within a particular level of education does not seem to have changed too much.

There was, however, a slight increase in inequality between groups. Here, the between component can be thought of as what the overall inequality level would be if there was no variation in income within each education grouping. In 2001, it was 0.053 for disposable personal income and 0.020 for disposable equivalised household income. In 2010, the figures were 0.057 and 0.021 respectively. The increase in between-group income inequality over the decade suggests a small but important widening of education-based income inequality. It may not be as large as what has occurred in the US and documented in Human Capitalism, but it is noticeable in the HILDA survey.

Income inequality in a counterfactual Australia

In a paper that I had published recently in Education Economics (paywall), I showed that youth appear to respond to localised returns to education and are more likely to participate in education in areas where predicted returns are highest. For all its flaws, one of the great insights from economics is that individuals respond to incentives. Responses may not always be completely rational, but incentives matter. It would appear from the data presented here that this is also occurring at the national level, thereby minimising the effect of skills-biased technological change on inequality in Australia.

To highlight this, it is worth exploring what would have happened to inequality in the absence of such a change by setting up a counterfactual scenario. Specifically, what would income inequality in Australia be like in 2010 if the observed changes in income had occurred, but they did so alongside a static education distribution? The answer appears to be that inequality would have increased at a much faster rate than it actually did.

Focusing on personal disposable income, total inequality in 2001 according to the Theil index was 0.350. In 2010, it was 0.358, a small increase. However, if the education distribution had stayed the same (but income increased as it did), then inequality in 2010 would have been 0.369, a much greater increase than was actually observed over the period. Furthermore, this increase in inequality would have been driven by a 13.2 per cent increase in the between inequality component.

Concluding comments

Taken together, two major trends in Australia seem to be mostly cancelling out, thereby leading to relatively stable income inequality. Although income was growing at a faster rate at the upper end of the education distribution, more people in the most recent data have those higher levels of education.

Despite trends being quite different in Australia, there are still a number of important insights from Human Capitalism. The first, which I haven’t touched on, is the impact of inter-generational income inequality. We don’t have good data on this across the lifecourse, but analysis clearly shows that those children in Australia from relatively disadvantaged backgrounds tend to have worse educational outcomes. The argument made by Lindsey is that this is in part due to complexity. It is worth exploring this in Australia because, even if overall inequality has stayed reasonably constant, if your parent’s income or education level is the major determinant of your own income or education, then this is a clear policy issue.

The second key point from Human Capitalism is that while important, income isn’t everything. It is true that the tax and transfer system in Australia reduces income inequality quite considerably. For example the Gini coefficient for gross personal income in Australia (0.493) was much higher than for gross disposable income (0.444) when the progressive tax system is taken into account. But public policy has fewer levers when it comes to autonomy and satisfaction in a person’s work. That is why I raised how much I love my job at the start of this post.

My education has given me the opportunity to do interesting work which makes me feel like I contribute to a small extent to society and our intellectual knowledge of the world. The unequal distribution of work fulfillment is an interesting and important avenue of future inquiry.

_____________________

Notes:

In the analysis of HILDA, population weights are used and negative and zero incomes are set to $1 (to enable their inclusion in log-based inequality calculations). Varying this has no effect on overall conclusions.
A small proportion of the sample had a degree without having completed Year 12. They were grouped with other degree holders for convenience.

Posted in Book reviews, Economics of Education, Inequality | Leave a comment

Budget night and a profile of Australian Public Service officers

Posted on May 12, 2013 by nicholasbiddle

The right questions on budget night

It is budget night on Tuesday (May 14th). As I write this post, the main stories related to the budget on most of the news websites are how and when the budget will be back in surplus, as well as how big a political cost there will be to the government for the current net financial position.

I agree that the predicted size of any surplus or deficit is newsworthy. It tells us a bit about the economic circumstances of Australia and the wider world. It also tells us about the priorities of the current government and, through its budget reply, how the priorities of the opposition might vary. However, as has been argued by a number of columnists, the singular focus on the size of the deficit or surplus distracts us from many more important discussions.

From the revenue side, in my view what we really should be interested in is (a) whether the amount received is sufficient to meet our collectively identified needs (b) whether the way in which the revenue is collected distorts incentives and if so whether the distortions are in the right direction and (c) whether the capacity to contribute is reflected in the tax system in the way we hope it to be.

In terms of expenditure, we should be thinking about (a) whether overall expenditure reflects our views on the size of the government relative to the economy (b) whether the individual items of expenditure adequately reflect our collective priorities and (c) whether the expenditure is being done efficiently and primarily in areas where markets are likely to fail. As a researcher, I am also interested in whether the expenditure is being adequately evaluated so we know enough about the behavioural and wellbeing effects of particular programs to be able to spend most effectively in the future.

These are the structural questions regarding the budget. We should also be thinking about whether the balance between revenue and expenditure is right given the current and future state of the economy.

As a resident of Canberra, however, I am also particularly interested in the effect of budget decisions on the lives of those in the public service. For people in other cities, the focus tends to be on whether there will be extra money in the bank each week, extra services or a higher approval rating for their favoured political party. For some of the people I know, decisions announced as part of the budget effect whether or not they will have a job next week or, at the very least, what they will be doing in that job. I felt it timely, therefore, to have a bit of a look at the characteristics of those working in the Australian Public Service (APS).

Where do Australian Public Servants live?

There were about 9,950,000 Australian’s working at the time of the 2011 Census. Of these, 4.2 per cent (or about 413,000) were working in the APS, 10.1 per cent were working for State/Territory governments, and 1.5 per cent for local governments. This leaves 84.2 per cent working in the private sector. Focusing on those in the APS, the following figure looks at the per cent of the total APS who were working in each State/Territory (in black) as well as the per cent of the relevant State/Territory workforce that are in the APS.

State/Territory profile of APS officers

The above figure shows that more than a quarter of the APS were living in New South Wales at the time of the 2011 Census. A further 23.7 per cent were living in Victoria and 15.9 per cent in Queensland. The vast majority (83.7 per cent) of the APS lived outside of the Australian Capital Territory (ACT). It is true that the APS makes up a much greater share of the ACT workforce than in any other jurisdiction (34.6 per cent). However, the APS is far from exclusively Canberra-based and Canberra is no longer exclusively a public-service town.

Nonetheless, there is still somewhat of a Canberra-focus, and because of this, the majority of APS officers live a relatively urban existence. Using the standard remoteness hierarchy used by the Australian Bureau of Statistics, 80.4 per cent of those in the APS live in a major city. This is much higher than the per cent of the total workforce (71.6 per cent) in major cities.

What is the age and gender mix of the Australian Public Service?

The APS doesn’t just have a somewhat different geographic distribution compared to the rest of the workforce, the demography is also different. First, the gender profile varies. Almost exactly half of the APS are male (50.1 per cent). This is much higher than those working for State/Territory governments (36.7 per cent) but somewhat lower than those in the private sector (55.5 per cent).

APS officers are also more likely to be in the middle of their working life. Around 9.2 per cent of the APS are aged 15 to 24 compared to 15.5 per cent of the rest of the workforce. At the other end of the age distribution, 16.4 per cent of the APS are 50 years or over compared to 17.6 per cent of the rest of the workforce.

These age/sex differences are summarised in more detail in the following age pyramid, which gives the per cent of the APS in each five-year age cohort/sex combination (in grey) alongside the rest of the workforce (in white).

Age pyramid of APS and other workforce

What type of work do Australian Public Servants do?

The modern workplace is quite a different one compared to that of even a decade ago. When I started work (in the APS) in 2001, I couldn’t check my work email from home and had to go to a special computer in the office to use the internet. Now, it is 2013 and I am working at home on a Sunday evening, distracting myself every now and then by checking Facebook and how the Knicks are going against the Pacers in their second-round playoff (not so well). Even though I am no longer in the APS, I suspect there are a number of public servants doing a similar thing at the moment (although probably not the Knicks part).

There are, however, still a number of differences in the type of work done by those in the APS and the rest of the workforce. This is demonstrated in the following figure which gives the structure of the APS and the private sector workforce by their occupation. The figure also shows the gender split in the type of work done in the two sectors.

Occupation of APS officers and private sector

Around 14.8 per cent of males in the APS self-identified as being managers. While this was slightly less than the 16.6 per cent of males in the private sector, it is substantially higher than the 10.8 per cent of female APS officers. There is an even more marked split by gender amongst technicians and trades workers.

Leaving gender aside, both males and females in the APS were more likely to be professionals than their private sector counterparts. They were also more likely to be clerical and administrative workers. On the other hand, those in the APS were much less likely to be sales workers; machinery operators and drivers; and labourers. In general, the APS is somewhat more skilled than the private sector.

Those in the APS are less likely to work part-time than the rest of the workforce. However, this is mainly driven by a much lower part-time rate amongst female APS officers. Specifically, 36.7 per cent of female APS officers were working part-time, substantially lower than the 43.7 per cent of females working for State/Territory governments, the 45.4 per cent of females working for local governments and the 51.9 per cent of those in the private sector.

Amongst those who work full-time (35 hours or more per week), those in the APS work slightly fewer hours on average than those in the private sector. However, the difference isn’t as big as some of the public-service stereotypes might suggest. Specifically, APS officers who worked full-time worked on average 43.0 hours per week compared to those full-time workers in the private sector who worked on average 45.4 hours.

Where there is a big difference though is the per cent who worked an extremely long week. Only 7.3 per cent of full-time workers in the APS identified in the census as working 60 hours or more in the previous week compared to 12.5 per cent of those in the private sector.

Where were those in the Australian Public Service born and what is their ethnicity?

Although there are some limits within the public service in terms of needing to be an Australian citizen for certain roles, it would appear from census data that the APS is not that different from the rest of the workforce in terms of whether they were born in Australia or not. Specifically, around 27.7 per cent of the APS were born overseas compared to 28.0 per cent of the total workforce.

There were, however, some differences by Aboriginal and Torres Strait Islander (Indigenous) status. Interestingly, however, these differences once again showed a clear gender split.

According to their website ‘The Australian Public Service Commission supports the recruitment, development and career progression of Aboriginal and Torres Strait Islander peoples in the APS by implementing the APS Indigenous Employment Strategy.’ Around 1.3 per cent of male APS officers identified as being Indigenous. This is slightly less than the per cent of the total workforce (1.4 per cent). However, there was a much higher per cent of female APS officers who identified as being Indigenous – 1.9 per cent compared to 1.5 per cent for the total workforce.

The Australian Public Service and the budget

My suspicion is that those in the public service will be paying particular attention to budget announcements and analysis over the next few weeks. At the very least, they will be watching the budget coverage with a different set of interests. They will not only have to implement the majority of policies and initiatives in the budget, their jobs may also be directly at risk.

It is easy to dismiss such concerns as being only an issue for Canberrans. And it is true that changes to the APS affect the Canberra economy, labour market and housing market more directly. However, less than one-in-five APS officers live in Canberra.

The previous discussion has also shown that those in the APS are more likely to be female; in the middle of their career; employed as a professional or clerical/admin. worker; working full-time (though not working crazy hours); and an Indigenous female.

The national media in my view could benefit from a more structural and values approach to the discussion of decisions announced on budget night. However, it could also benefit from a consideration of the individuals charged with implementing the decisions.

Posted in Census | Leave a comment

Attacking educational disadvantage through school funding (co-authored with Timothy Cameron)

Posted on May 1, 2013 by nicholasbiddle

“…all students must have access to an acceptable international standard of education, regardless of where they live or the school they attend. …[equity means] differences in educational outcomes are not the result of differences in wealth, income, power or possessions” (pg 105, review of funding for schooling)

The Economics of Education

There are four reasonably well established findings regarding the economics and financing of education:

Education provision is expensive with large infrastructure and salary costs;
The social and economic benefits of education for both the individual and society usually outweigh these costs;
The individuals who benefit from this education are usually not in a position to finance it themselves and are reliant on their families or the state to do so; and
Individuals from disadvantaged backgrounds have relatively low levels of education achievement on average.

These four factors combined mean that education is at the same time a consequence, cause and cure for disadvantage within and across societies. The extent to which education reduces rather than exacerbates inequality, however, is largely determined by the quality of education received by those from disadvantaged backgrounds and whether that quality is at least as high or ideally even higher than that received by the rest of the population.

Education is a positional good where completing some form of education is of secondary importance to how a person is relatively positioned in that course, that class, or across the labour market. For example, the benefits an individual might obtain from seeing a GP won’t be adversely affected by the relatively well off in the city in which they live seeing a specialist or a more expensive GP. However, the benefits an individual gets from completing Year 12 will quite possibly be affected by how they are ranked in terms of university entrance indices and more indirectly, whether or not those around them are all obtaining a Bachelor’s degree.

All levels of government and all major parties recognise the role of the public sector in funding the delivery of education. With regards to school funding, where there is debate is around the following three questions:

What is the total level of funding available to school education?
To what extent should governments subsidise the choices made by families to send their children to a fee-based private school? and
How should the characteristics of students and schools impact on the amount of funding received?

The National Plan for School Improvement

The responses to each of these questions are different in the eight Australian States and Territories. While the Federal Government has a minimal direct role in school education, they do provide significant funds to the States and Territories. The National Plan for School Improvement and the Australian Education Bill 2012 represent the Federal Labor Government’s response to the three questions posed above. With regards to expenditure per student, the base amount of funding is $9,271 per primary school student and $12,193 for secondary students. The base level of funding per student is considered the amount of money required to support a student with minimal educational disadvantage in a school where at least 80 per cent of the students exceed a specified national minimum standard in reading and numeracy. The level of recurrent funding is guaranteed for all government school students and is allocated to independent schools on a ‘capacity to contribute’ basis.

With regards to the last of the questions posed above, the Australian Education Bill 2012, explicitly identifies six types of educational disadvantage which will attract additional funding. These measures of educational disadvantage relate to (1) a school’s size and (2) location, as well as (3) how many students are Aboriginal or Torres Strait Islander, (4) have a disability, (5) are from a low socioeconomic background or (6) have limited English proficiency. These measures of disadvantage featured heavily in the Review of Funding for Schooling, also known as the Gonski Review. Furthermore, there is empirical support for these measures being associated with poorer education outcomes.

Using data from Wave 1 of the 2009 cohort of the Longitudinal Surveys of Australian Youth (LSAY), the following figure shows the difference in terms of standard deviations for maths (in black); reading (in grey) and science (white) for six characteristics of a nationally representative group of 15 year olds, as well as from a one standard deviation increase in a standardised index of the child’s economic, social and cultural status (escs). The results are from a regression analysis with the child’s sex, age and grade level held constant. Differences which weren’t significant at the 5% level of significance have been set to zero in the figure (but still included in the analysis).

LSAY analysis of test scores and education disadvantage

The results from the analysis of the LSAY were reasonably consistent across the three domains. Students attending a Catholic or independent school had higher test scores than those attending a government school. Students from schools in a provincial or remote location had worse test scores than those in a metropolitan location. Finally, students from more advantaged families had higher test scores, whereas Indigenous students and those who spoke a language other than English at home had lower ones. Unfortunately, disability status is not available on the LSAY, but other research clearly shows that those children with a disability tend to have lower test scores than those who do not.

This socioeconomic distribution of school outcomes has long been established. Identifying an efficient, equitable and politically palatable way to recognise this disadvantage through school funding has been more difficult. At the moment, determining the exact amount of funding that is currently being provided to disadvantaged students and schools is almost impossible due to complex and varied funding arrangements at the state and territory level. Recognising this, the Prime Minister and her relevant ministers are attempting to get the states and territories to sign up to the National Education Reform Agreement (NERA) which would lay the foundation for a consistent funding approach in order to tackle this disadvantage.

The new funding model, the Schooling Resource Standard (SRS), is designed to provide the aforementioned base level of funding for all students as well as additional loadings for those subject to one or more of the measures of educational disadvantage. Unlike the base funding, the loadings for educational disadvantage will be constant across government and non-government schools. Most of the loadings for educational disadvantage allocate funding so that that as disadvantage becomes more concentrated within a school, funding rises at an increasing rate.

Under the SRS, schools will be funded depending on the percentage of students in the lowest two quartiles (Q1 and Q2) of the socio-educational advantage (SEA) scale. These loadings range from 15% to 50% of the per student amount for each Q1 student depending on the concentration of Q1 students and 7.5% to 37.5% per Q2 student depending on the concentration of Q2 students. The following chart shows the additional funding a high school receives per low-SEA student. The chart is indicative only as we have assumed linearity between known points and haven’t taken into account that there is limit of 100% in terms of combined Q1 and Q2 students.

SEA and funding

The school size loading provides extra funding to smaller schools in recognition of the fact that there are economies of scale in school provision. This funding will provide up to $240 000 for high schools and $150 000 for primary schools. The chart below shows the funding levels dependent on student enrolment. For high schools with less than 100 students and primary schools with less than 15 students the funding amounts are still unclear.

School size and funding

The average costs of delivering education into regional and remote areas are higher than in metropolitan areas and student outcomes are lower in these areas even after accounting for a number of factors. This has long been reflected in Commonwealth Grants Commission allocation of GST revenue to the states. Each school’s remoteness will be calculated by the Accessibility/Remoteness Index of Australia (ARIA). Inner regional schools will receive up to 10% of the loading amount, outer regional schools will receive 10%-30%, remote schools 30%-70%, and very remote schools will receive a loading of 70%-80%.

Through the Closing the Gap targets all Australian governments have recognised and committed themselves to addressing the fact that Indigenous students, on average, achieve significantly worse education outcomes compared to non-Indigenous students. We have both together and independently done considerable work on this topic and along with other authors have shown that disadvantage starts off early and often widens throughout the school years. Furthermore, there are ongoing questions regarding the relevance of current assessment and curricula to all Indigenous children. Under the SRS, the Indigenous loading starts at 20% of the per student amount for the first Indigenous student and increases to 120% for schools with a solely Indigenous student population. The graph below shows funding per Indigenous student [assuming linearity] based on the concentration of Indigenous students.

Indigenous status and funding

Loadings for students with disabilities (SWD) and students with limited English proficiency are yet to be finalised. In 2014 interim loadings will be applied that provide 186% of the per student amount for SWDs and 10% for students with language backgrounds other than English. The Government recognises there are data collection issues which lead makes identifying the needs of these students difficult and is working on improved loadings for 2015.

Limitations of the National Plan for School Improvement

If implemented as outlined above, the Schooling Resource Standard will represent a significant increase in the amount of resources available to schools and, in particular, those schools that are attempting to educate some of Australia’s most disadvantaged students. There are, however, a number of limitations of the plan outlined by the Gillard government.

The first of these limitations is that the six measures of disadvantage explain only a small proportion of the variation in school outcomes. For example, the models summarised in the first figure in this article explain between 20% and 23% of the variation in the relevant test scores in the LSAY. Other factors like the quality of the early childhood education experience of the students, the stability of the child’s home environment and the local labour market or social conditions are all beyond the control of the school but likely to impact on student outcomes.

The second limitation of the model is that there is considerable variation within schools in terms of the educational disadvantage faced by the students. Not all disadvantaged children attend a relatively disadvantaged school with some children in even the most advantaged school environment likely to struggle. Furthermore, there are likely to be many children at the schools that receive additional funding that would do quite well regardless of their context. In the LSAY data analysed above, the sample is clustered within 353 schools with an average of 43 students from each school surveyed. The school in which the child is currently attending explained only 24-25% of the variation in the maths, reading and science test scores. While a strong case can be made for maintaining principal and teacher autonomy (within bounds), and there is the distinct possibility that attending a school with other disadvantaged children has negative impacts, it is still the case that under the current plan there is no guarantee that the most needy students within the schools will receive the benefits from the additional funding.

This raises the question of whether the additional funding allocated as part of the NPSI is best targeted towards schools or whether at least some of it should go to the families themselves. The overarching aim of the Australian Education Bill 2012 is that ‘by 2025, Australian students will be in the top 5 in the world in reading, science and maths.’ With a finite budget and an uncertain revenue base, this goal might better be reached by spending some of the proposed $14.5 billion on high quality early childhood education; cash incentives for at-risk students to complete education or enter university; greater income support for the families of disadvantaged children; health interventions for Indigenous children in remote areas; remedial literacy outside of school hours for those struggling with reading or writing; or a range of other potentially worthwhile programs.

The reason why we said this goal *might* better be reached through other programs is that, in Australia, we really have very little quantitative evidence on what actually works to encourage disadvantaged children to engage with formal schooling and achieve education outcomes that are commensurate with their potential. The gaps in achievement which we are grappling with in Australia between those from socioeconomically advantaged and disadvantaged backgrounds; Indigenous and non-Indigenous children; those with a disability and those without; those from an English speaking background and those from other backgrounds; and across our vast continent are in many ways quite small compared to those in the US. In a relatively recent summary article, Curto, Fryer and Howard ‘describe recent social experiments and evaluations of investments in schools, communities, and family engagement strategies’ and conclude that ‘the evidence to date suggests it may not take a village to increase the achievement of the poorest minority students, just a high-quality school.’

We have no idea whether such conclusions are of relevance in Australia because there are no trials of the type summarised by Curto and co-authors. This is partly because unit record data on education outcomes is rarely made available for independent evaluation. It is also because education interventions are not currently set up in such a way that a clear comparison can be made between a treatment school or group of students and an otherwise identical control group. So, while support for additional funding in education is reasonably easy to justify, strong evidence to support how that additional funding should be allocated is more difficult to come by.

A recent article by Dr. Ben Goldacre, the author of the Bad Science blog, outlines the potential for the insights from randomised trials in medicine to be incorporated into education research. He argues that ‘by collecting better evidence about what works best, and establishing a culture where this evidence is used as a matter of routine, we can improve outcomes for children, and increase professional independence.’ In the article, Dr. Goldacre canvasses the limits to such trials (including applying insights from one trial to another country or context) and the role of qualitative research in understanding the details of why and how a program might be achieving its aims. However, he makes a convincing case for much greater use of studies with treatment and control schools or students to help teachers and school administrators effectively test what works well and what doesn’t work so well in education provision. He also, it should be noted, makes a strong case for academics and policy makers to work more closely with education providers to answer the questions which they are actually interested in and would support their work.

In our opinion, the NPSI as it is currently structured represents a missed opportunity in one key respect. The Federal government could have followed the approach of in essence saying to the states ‘you can spend the additional funds however you like, as long as you evaluate it properly.’ We could then in a few years say whether providing additional pay for teachers, introducing certain technologies into the classroom, providing direct financial assistance/incentives or any of the many other avenues of expenditure are effective.

There are no certainties in politics and it is possible that the Gillard government will be able to turn around its poor polling figures and win another term of government. It is also possible that an incoming Abbott government will retain much of the substance of the NPSI. However, it is also a distinct possibility that the new funding approach will be substantially modified. Although it might seem a second order issue, there is a real opportunity for the Gillard government to change the way such major reforms are implemented in the future by building in a careful and robust evaluation component to the additional funding.

Posted in Economics of Education | 1 Comment

Indigenous population – Initial projections and implications

Posted on April 25, 2013 by nicholasbiddle

Not long after the 2006 Census, the ABS estimated that there were around 517,000 Aboriginal and Torres Strait Islanders (Indigenous Australians) living in Australia. This represented about 2.5 per cent of the overall Australian population. Although final estimates aren’t available, the ABS’s preliminary estimates from the 2011 Census are around 670,000 Indigenous Australians or around 3.0 per cent of the total Australian population.

This is a very rapid population increase over just a five-year period. In my view, there are six main reasons for this increase:

Indigenous Australians are concentrated in the main childbearing years (at least relative to the non-Indigenous population);
Indigenous females continue to have a greater number of children than non-Indigenous females, especially when they are relatively young;
In urban areas in particular, there is a high partnering rate between Indigenous males and non-Indigenous females with the children of these partnerships tending to be identified as being Indigenous;
The ABS may be getting better at counting Indigenous Australians in the census;
The ABS may have historically underestimated the number of Indigenous people who were missed by the census in previous years; and
There may have been a non-negligible number of people who previously did not identify as being Indigenous in the census but now feel more comfortable in doing so.

Although we won’t know till towards the end of the year when new data is released, it is also quite possible that mortality rates amongst the Indigenous population have fallen. Whatever the reasons for this rapid population increase, it is clear that it has implications for policies that seek to address the socio-economic and health disadvantage faced by the population.

As important as the population increase at the national level is, the change in composition of the Indigenous population through time is equally important. Growth that is unevenly distributed by geography, age or other characteristics will inevitably put greater pressure on particular regions or policy areas. Consider the following graph which gives the change in the Indigenous population estimate by broad age group and State/Territory.

Population change by State/Territory

There are two main points to note from the above figure. First, the highest rate of growth over the last intercensal period was amongst those aged 55 years and over. That is, the Indigenous population is aging. Secondly, growth was fastest in the east-coast states and slowest in the Northern Territory.

Some of the causes of this rapid population increase are likely to continue to have an effect into the future. Although it is very difficult to predict future changes in identification or enumeration, we can use current data to obtain a set of population projections for what the Indigenous (and non-Indigenous) population might look like into the future if currently high rates of fertility and paternity continue. This is not a forecast in the usual sense in that one doesn’t make predictions based on judgement and probabilities. Rather, population projections tend to be simply a reflection of what the future would look like if previously observed data and trends continued.

The most common method to project the Indigenous population (and the one used by the ABS) is the cohort-component methodology. At the national level, the Indigenous population at a particular future time period (t+Δ) is equal to the population at a previous point in time (t), plus births to Indigenous mothers and births to non-Indigenous mothers and Indigenous fathers that occurred over the period, minus Indigenous deaths, plus international net migration. For regional population change, net internal migration should also be added. These components of population change are examined separately and rates for these are applied to cohorts of a base population as appropriate, resulting in a set of projections for a set time period. This process is iterative across age groups over the projection period.

Such methods require a set of assumptions about the particular populations and rates. The assumptions I used to obtain an initial set of population projections are given at the end of this post. In essence, I use the most recently available data for birth rates and death rates, apply these to the 2011 base population, and assume a net international migration rate for the non-Indigenous population of 180,000 people per year. Using these assumptions, the following figure gives the projected Indigenous population from 2016 to 2031, as well as the estimated Indigenous population in 2011. The dotted line (plotted against the right-hand axis) gives the per cent of the total Australian population projected to be Indigenous.

What the above figure demonstrates is that if rates of fertility, paternity and mortality continue as they have been, then the Indigenous population is projected to grow from around 670,000 in 2011 to around 1,040,000 by 2031. While this population growth isn’t as fast as was observed between 2006 and 2011 (because no identification or enumeration change was incorporated into the projections), it is still projected that the Indigenous population will grow from around 3.0 per cent of the Indigenous population in 2011 to 3.8 per cent by 2031.

A growing Indigenous population has a number of policy implication. With a given Indigenous affairs budget, fewer resources will go to each Indigenous person, at least in percentage terms. Alternatively, resources may need to be drawn from other budgetary areas or from additional taxation, which clearly has a different set of implications. Whatever the response, it is perhaps even more policy relevant to note that this growth in the Indigenous population is not projected to be evenly distributed across the age distribution.

For example, although the total Indigenous population is projected to grow by 59% between 2011 and 2031, the population aged 0 to 24 is only to projected to grow by 47%. This is still faster than the equivalent projection for the non-Indigenous population in that age group, but is much slower than the 65 plus Indigenous population which is projected to grow by 301%. What this means is that the Indigenous population in 2031 is projected to be much older than the Indigenous population in 2011, as demonstrated by the following age pyramid (which gives the per cent of the 2011 and 2031 Indigenous population in each age/sex combination).

The current Indigenous population is estimated to be slightly older than the population was as of the 2006 Census. The above figure shows, however, that this aging is likely to continue with 6.4% of the Indigenous population projected to be aged 65 years and over in 2031 compared to around 3.4% at the moment. This could have financial implications as low rates of employment are likely to mean that Indigenous retirees have far less savings than their non-Indigenous counterparts. It is, however, also likely to have implications for health and disability policy.

To understand why, consider the following distribution of profound or severe disability for Indigenous and non-Indigenous males and females from the 2011 Census.

Clearly, and perhaps not surprisingly, rates of disability increase substantially across the age distribution. However, it is worth noting that across the age distribution Indigenous males and females have a higher rate of disability than their non-Indigenous counterparts. When you apply these disability rates to the 2011 population estimates, you obtain an estimate of around 19,500 Indigenous males with a profound or severe disability and 17,500 Indigenous females. This represents 5.9% and 5.2% of the respective populations.

If the above disability rates were applied to the 2031 Indigenous population, however, then it is estimated that there would be 34,300 Indigenous males with a profound or severe disability in 2031 and 33,600 Indigenous females. This is not only an 83% increase in the Indigenous population with such a disability, but also represents an increase in disability as a per cent of the Indigenous population – up to 6.6% of Indigenous males and 6.3% of Indigenous females.

These population projections are in many ways preliminary and may change somewhat as new data becomes available and as different techniques are applied to them. They are indicative though and it is quite likely that the Indigenous population will continue to grow at a faster rate than the non-Indigenous population and that there will be an ongoing aging of the population.

For many reasons, Indigenous population increase is a good thing. It can signal a greater confidence in identifying as such, an improvement by our statistical agencies in establishing the true population or, perhaps, a greater proportion of Indigenous Australians surviving to adulthood and beyond. But, it doesn’t come without complications and issues. Ideally, these issues should be thought through and planned for by policy makers and researchers well in advance in order to ensure that Indigenous Australians receive the type of policy support required for them to live the lives that they desire and are in many ways entitled to.

_________________________________________

The following are the assumptions that I use to generate a set of preliminary projections:

Base population – Use the ABS, 2011-based preliminary population estimates for each five-year age group with the State/Territory population allocated to Indigenous Regions based on the region population count;
Births of Indigenous children to Indigenous and non-Indigenous mothers – Use the State/Territory based fertility rates (from 2009), applied to each Indigenous Region and age group;
Deaths of Indigenous Australians – Use the most recent (2005-2007) mortality rates for each State/Territory applied to each Indigenous Region and age group;
International migration – Assume zero international migration for the Indigenous population and 180,000 net non-Indigenous migration per year (allocated to each Indigenous Region based on its share of recent migrants over the 2007-2011 period);
Internal migration – Assume that the net Indigenous Region migration rates over the 2006-2011 intercensal period are replicated for each subsequent period, capped at -5% and +5%

Posted in Indigenous populations | Leave a comment

Advantage and disadvantage across the country

Posted on April 2, 2013 by nicholasbiddle

On March 28th, the Australian Bureau of Statistics (ABS) released a new round of data based on the 2011 Census. One part of this latest release that caught the eye of the media was the 2011 Socio-Economic Indexes for Areas (SEIFA), a set of measures of relative socio-economic disadvantage and advantage, calculated for almost all geographic areas in Australia.

Some of the headlines during the week were: ‘Peppermint Grove Australia’s most affluent suburb‘, ‘Suburban advantage survey plots where the true battlers live‘ and ‘Western Australia’s wealth a migration magnet‘.

Unlike most other data that the ABS releases from the census, SEIFA contains a significant amount of value-add. Some of the best and brightest analysts within the ABS combine information from a number of census data items to summarise different aspects of the socio-economic conditions in an area into four indices:

Index of Relative Socio-Economic Disadvantage (IRSD)
Index of Relative Socio-Economic Advantage and Disadvantage (IRSAD)
Index of Economic Resources (IER)
Index of Education and Occupation (IEO)

The indices are created using a technique known as Principal Components Analysis which, in essence, summarises as much of the variation as possible in the underlying data items into a single index number. As an example, the IRSAD includes 10 variables that indicate that the area is relatively advantaged (e.g. percentage of people with high income, employed as a manager or professional, households with three or more cars) and 15 variables that indicate an area is relatively disadvantaged (e.g. no education, single parents, unemployed).

A range of geographic areas can then be ranked based on their SEIFA score for each of the four indices. This can be done for neighbourhoods (Statistical Area Level 1), suburbs/towns (Statistical Area Level 2), Local Government Areas, Postal Areas and even neighbourhoods within Commonwealth/State Electoral Divisions.

With regards to the latter, it is interesting to note, that only 0.3% of the Opposition Leader Tony Abbott’s electorate of Warringah live in the most disadvantaged 10% of neighbourhoods (based on the IRSAD) whereas 20.0% live in the most advantaged 10% of neighbourhoods. In Prime Minister Julia Gillard’s electorate of Lalor, on the other hand, 8.0% live in the most disadvantaged 10% of neighbourhoods compared to only 2.0% of her electorate who live in the most advantaged 10% of neighbourhoods. In and of itself these statistics tell us very little about the respective policy views of the Prime Minister and the Opposition Leader or their commitment to improving the lives of their constituents. But, it might tell us a little bit about the priorities of the people who they represent.

Beyond just playing around with numbers, SEIFA has a few more important policy and research purposes. Most obviously, the indices can be used for the targeting and planning of government and commercial services. If, for example, the Commonwealth government was planning a randomised controlled trial (RCT) to test a new service targeted at reducing unemployment and associated disadvantage, then it is unlikely to be delivered in The Ponds (highest ranking suburb in NSW), Pullenvale (Qld) or North Coogee (WA). Instead, the suburbs of Binjari (NT), Greys Plain (WA) or Wami Kata (SA) are ranked at the bottom of the IRSAD distribution and are more obvious sites for such a trial. Of course, through their members and constituents, governments already have a sense of where disadvantage and advantage lie. However, SEIFA adds some solid statistical evidence to decision making.

Another use of SEIFA is to help explain individual behaviour. There has been thousands of research papers across the health and social sciences that have shown that those individuals who live in relatively disadvantaged areas tend to have worse outcomes themselves. For example, in work that I am currently doing looking at early childhood education participation, 27% of children in the Longitudinal Study of Australian Children who lived in the most disadvantaged 20% of neighbourhoods in the year before school had their academic skills rated by their teachers as below average or far below average. For those in the most advantaged 20% of neighbourhoods, it was only 15% rated as below average or far below average.

Now, we can’t really say whether these differences are causal or not. Location is a choice and it may be that the families of children who would be rated relatively poorly by their teachers choose or a forced to choose to live in disadvantaged areas. However, the difference holds after controlling for 24 other risk factors, as well as preschool and long day care participation. At the very least, we can say that living in a disadvantaged area is predictive of low outcomes, even after controlling for important mediating factors like low family income, low birth weight, TV watching, etc.

A third use of SEIFA is simple curiosity. It enables individuals to ask, where does my neighbourhood, suburb or town fit in compared to other parts of the country? I live in the Canberra suburb of Ainslie which, based on the SA2 geography mentioned earlier, is in the 83rd percentile according to the IRSAD. More advantaged than the vast majority of suburbs across Australia, but ranking closer to the middle in terms of Canberra suburbs. The following table gives the top and bottom two suburbs for each of the eight Australian capital cities

This table, although a very small subset of areas, highlights the considerable diversity within our capital cities in terms of socio-economic advantage/disadvantage.

The fourth use of SEIFA, and the one that I find most interesting, is understanding the geographic context in which particular population sub-groups live. By definition, roughly 10% of all Australians live in the most disadvantaged 10% of neighbourhoods and roughly 10% live in the most advantaged 10%. This is not, however, true of all population subgroups.

Consider the following graph, which gives the per cent of three broad age groups who live in each of the IRSAD deciles from the most disadvantaged decile on the left to the most advantaged on the right. There is a reasonably even distribution for the population aged 0 to 14 years with slightly fewer in the 2nd to 5th decile and slightly more in the 8th to 10th. However, it is when one looks at the elderly population (65 plus) that the age distribution of area-level disadvantage becomes apparent. Around 25% of the elderly population live in the two most disadvantaged deciles compared to 16% who live in the two most advantaged ones.

IRSAD by age

This area-level disadvantage is even more pronounced when one considers the Indigenous Australian population as demonstrated below. A massive 37% of Indigenous Australians live in the most disadvantaged decile compared to only 2% who live in the most advantaged decile. It is true that some of that area-level disadvantage would have been due to the characteristics of those Indigenous Australians themselves. However, the vast majority of Indigenous Australians only make up a small per cent of the neighbourhoods in which they live. What these SEIFA results show, therefore, is that not only are Indigenous Australians relatively disadvantaged, but that they live in areas where their neighbours and friends are disadvantaged. As it is these neighbours and friends that people often use to obtain labour market, education and financial support, then it is quite possible that this area-level disadvantage contributes to individual disadvantage.

IRSAD by Indigenous

If the above exercise was carried out in the US, then it is likely that one would also find that the Native American population was more likely to live in a disadvantaged area than the rest of the population. It is also likely though that recent immigrants to that country also lived in relatively disadvantaged areas. This is not the case, however, in Australia. This is demonstrated in the following figure which gives the IRSAD distribution for two cohorts of migrants – those who arrived in Australia between 2007 and 2011 and those who arrived between 1997 and 2006. As you can see, those who arrived to Australia relatively recently (that is, in the last 15 years) are much less likely to live in relatively disadvantaged neighbourhoods and much more likely to live in relatively advantaged ones.

IRSAD by Year of Arrival

Such a situation would probably not be replicated for all migrant groups, especially those who arrived under humanitarian visa categories. Nonetheless, the results say a lot about the way in which Australia has successfully been able to pursue a highly targeted migration policy that brings in highly skilled workers who are successful relatively quickly in the Australian labour market.

It should be kept in mind that there are a number of limitations of SEIFA. Because some of the indices include a wide range of input variables, it is not always possible to look at the range of disadvantage for some population subgroups. For example, as lone parents are included as an indicator of disadvantage in the IRSAD, it is not possible to use that particular index to test whether lone parents live in more or less disadvantaged areas than other types of parents.

A second limitation is that although they may be correlated, not all input variables are unequivocally good indicators of advantage or disadvantage. For example, the percentage of dwellings with three or more cars is used as an indicator of advantage in the IRSAD. In some instances, this is clearly an indicator of significant household wealth. In other cases though, it might be an indicator of relatively poor public transport options in the area.

Of course, this wouldn’t be the case if public transport availability was able to be included as an input variable. However, the construction of SEIFA is limited by what data is available on the census. There is no information in the census on crime rates in the area, whether there are areas for kids to play, whether there are broken street lights or whether there are any cafes that can make a good, brewed, soy chai. Perhaps more important than chai, there is no information on the census that relates to social relationships within the area or other subjective measures of wellbeing. While such measures correlate to a certain extent with objective measures like income, education, employment and overcrowded, this correlation isn’t perfect. It is quite likely that there are areas ranked as relatively disadvantaged that have a population with higher levels of life satisfaction than their more ‘advantaged’ counterparts.

The final limitation of SEIFA is that it tells us very little about the causes of disadvantage, either at the individual or area-level. Analysis of secondary data sources like the census is useful for identifying where relatively disadvantaged people live. However, in terms of knowing how to alleviate that disadvantage, what continues to be required is good quality longitudinal data and policy evaluations.

These limitations notwithstanding, SEIFA represents an important contribution to evidence-based policy making in Australia and has the potential to support research into some of the major policy and social issues in the country.

Posted in Census | 1 Comment

Poor (Indigenous) Economics – Part II

Posted on February 5, 2013 by nicholasbiddle

In yesterday’s post I gave a summary of the first five chapters of the recently released book by Professors Banerjee and Duflo titled Poor Economics. In the discussion, I added what I thought were some of the implications for one of the areas that I work in: Indigenous economic and social policy in Australia. In this post, I will give a summary of the last five chapters in the book, as well as what I see as the main implications of their analysis.

Chapter 6: Risk

The poor are exposed to a much greater level of risk than those who are relatively well off. The occupations that they work in (agriculture, small businesses) are much more volatile than other jobs. Furthermore, the effect of this variation is much greater without the savings and insurance that those in rich countries take for granted.

Those who face such risks tend to manage it by diversifying. This can be done through:

Diversifying across occupations;
Holding multiple plots in different parts of a village; or
Temporary migration with some parts of the family staying behind.

Other ways to manage risk are by being conservative in terms of managing firms or businesses, spreading risks across families and by having a greater number of children. However, the point made by the authors is that such strategies have costs.

Despite the potential economic benefits of insurance, uptake tends to be pretty low. This is often due to the lack of trust in these institutions as opposed to a poor understanding of the potential benefits of such strategies.

In Australia, much of the risk is covered by governments. Unemployment benefits are universal, unlimited (in terms of time on them) and only require recipients to meet an activity test. The public health system is extensive and in many contexts free or highly subsidised for the user. Furthermore, Australia was one of the pioneers of income contingent loans where individuals do not need to pay back the debt they have accrued through tertiary education until their income reaches a certain threshold.

Indeed, one of the most contested arguments within Indigenous policy is around the concept of ‘welfare dependency’ where too much risk has been taken away from individuals, blunting the incentives to work and study. I have yet to see a convincing empirical argument that high levels of unemployment amongst Indigenous Australians is caused by such a dependency. But it can’t be rejected on theoretical grounds. However, as the provision of things like education and health become more market based, there is the potential for this to disproportionately affect Indigenous and other low savings/wealth populations.

Chapter 7: Lending

The interest rates paid by the poor are many times higher than those paid by the rich. The standard argument is that this is due to high rates of default, but this doesn’t seem to be supported by the data. What the data does support though is that these low default rates require a lot of hard work on the part of money lenders.

For banks, there is a lot of effort that needs to go into each loan, regardless of its size. This means that the incentive is to give out fewer loans or charge more for smaller ones. According to the authors (page 163) ‘Because the main constraint on lending to the poor is the cost of gathering information about them, it makes sense that they would mostly borrow from people who already know them.’ Micro-finance institutions (MFIs) have attempted to fill this gap. According to the authors on page 166, ‘Like traditional moneylenders, MFIs rely on their ability to keep a close check on the customer, but they do so in part by involving other borrowers who happen to know the customer.’

Microcredit has been shown to have a positive, but small effect. The effect is small because many poor people are not willing or not able to start a business. Furthermore, not everyone has the social networks that enable them to participate in a group whereas others don’t necessarily like the intrusion.

One of the implications for Australia from Chapter 7 is that many Indigenous Australians might also be credit constrained. One of the other potential benefits of microcredit may be the ability to make large, long term purchasing decisions. However, like in many developing country contexts, one of the main limitations may be the low levels of economic opportunities in the areas in which some (though not all) Indigenous Australians live.

Chapter 8: Savings

There are a number of ways in which a small bit of savings could lead to substantially higher income in the future. However, saving for the poor is quite expensive when using regular banks. For this reason, the poor often utilise a range of other savings methods including pooling income across individuals. However, even when people have access to good savings opportunities, they often fail to take advantage. Like those in relatively well-off circumstances this is in part because of hyperbolic discount rates or valuing the present much higher than the future. To get around this, people tend to use a lot of pre-commitment devices (that lock them into saving), but these sometimes come with large costs.

The authors also show that optimism and hope about the future can have a positive effect on savings. if you don’t think that the future is going to be very positive, then you are less likely to think about it and less likely to do something about making it better. This is especially likely to be the case amongst those who feel that what happens in the future is largely determined by outside forces, rather than being within their control.

One of the signature (and contentious) aspects of the Northern Territory Emergency Response (NTER or The Intervention) was income management. A potential benefit of this policy may have been on the level of savings. However, the discussion in Poor Economics shows that most people are already aware of such benefits. It may also be the case (though this has not been empirically tested) that the negative effects that the NTER might have had on peoples hopes about the future had an even larger negative effect.

Chapter 9: Entrepreneurship

A much higher proportion of the poor are self employed than those in rich countries. However, these businesses tend to be small and don’t make a lot of money. They can support a family, but rarely grow large enough to add additional employees or to accumulate capital. A naive analysis of rates of return to additional capital in these firms is that it is quite high. However, it would appear that diminishing returns set in quite quickly and the marginal return is quite low. To get large overall returns (enough for the business to grow and to accumulate assets) a poor person would need to invest a large amount of money. However, people are likely to be risk averse with such large investments and, what is more, even MFIs will not lend such large amounts.

Another interesting finding in Poor Economics is that surveys in developing countries that ask for the ambitions of people for their children show that, rather than wanting their kids to be self employed like they are, they would much rather they were a teacher or had some other government job. This is because these jobs tend to be stable. This stability in turn makes it easier to invest in one’s children.

Increasing the amount of risk in Indigenous communities may give the incentive to run small businesses. However, it can also mean that individuals in such circumstances do not invest in their children’s health or education due to the uncertainty of being self employed or a small business owner. As far as I am aware, we also know very little about Indigenous Australian’s attitudes to entrepreneurship.

Chapter 10: Politics

The final chapter in Poor Economics looks at political institutions. This has particular relevance for Indigenous policy where many of the problems in remote communities are attributed to poor governance. A quote from the authors (on page 248) is highly relevant here ‘in practice, the implementation of community participation and decentralization matters quite a lot. How exactly does the community express its preferences, given that people often have different views? How can we ensure that the interests of the underprivileged groups (women, ethnic minorities, lower castes, the landless) are represented?

Conclusion:

The authors of Poor Economics explicitly avoid a sweeping conclusion from the analysis presented in chapters 2 through to 10. This, of course, fits within the underlying theme of the book that effective change is often best made at the margins with grand narratives about what should or should not be done to achieve sustainable economic development likely to unsatisfactory in one way or another.

They do, however, talk about five key lessons that emerge from their research and the research of others. These conclusions are based on the lives of those living on less than $1 a day, and therefore do not always have relevance for policy in Australia. In particular, the first two conclusions that the poor often lack important information and that the poor bear responsibility for many more aspects of their lives than the rich are of less relevance to Indigenous policy. Nonetheless, the implication that well designed policy takes careful consideration of default options and small nudges is relevant to all social settings.

The third conclusion that the authors make is that there are good reasons for why some markets are missing for the poor. In a developing country context, this often means insurance or financial markets. This is probably true, though to a lesser extent, in remote Indigenous communities. What is often missing in the communities that Indigenous Australians live in though are well functioning labour market and property markets. With the exception of resource rich areas, in the absence of significant government investment such markets are unlikely to spring up by themselves. However, governments intervene in markets all the time and there is no reason why this shouldn’t continue to occur in remote areas. The point though, is that such interventions need to be rigorously costed and evaluated.

The fourth conclusion that Banerjee and Duflo make is that quite often the main reason why policies fail is not because of some grand conspiracy or intractable incompetence on the part of policy makers and bureaucrats. Rather, it is because careful attention is not paid to the detailed design of policy. Governments often get the big picture more or less right – improve early childhood education, reduce infant mortality/morbidity, ensure people have access to a job. It is the small picture that is often found wanting. For example, should governments invest in an extra high school teacher or instead provide laptops for every student in a school? Fortunately, the answer to such small-scale questions can be effectively evaluated through carefully designed randomised controlled trials (RCTs) of the type the authors make extensive use of.

However, such testing of the details of policy delivery rarely occurs in policy debates in Australia and especially not in Indigenous policy. There are legitimate limitations of randomised controlled trials. In many cases, withdrawing or withholding treatment is not always feasible. The somewhat tongue-in-cheek example given by Andrew Leigh is you cannot use RCTs to test the effectiveness of parachutes. Not many people are going to agree to be in the control group in that experiment. There are also scalability and spill-over effects that lead to uncertainty around whether the results found in the trial will be replicated for the total population. A further limitation is that there are long lead times from when the trial is conceived to when policy conclusions can be made.

A final limitation of RCTs is that they are not really useful for testing the effect of national level policies and interventions. However, one might argue that in the case of things like constitutional recognition of the Indigenous population and the Apology to Australia’s Indigenous Peoples should be enacted because they are the right thing to do, not because of utilitarian motivations. Despite these limitations though, ethical and inclusive trials with community support are the only way to find out what specific policies work to get kids to school and preschool, improve Indigenous employment prospects, reduce the burden of disease and meet all the other policy goals of government.

The fifth and final point in Banerjee and Duflo’s conclusion is that expectations about what people can and cannot do have the potential to become self-fulfilling. This is a particular issue for the Indigenous Australian population who hear constantly through the media (and academic studies, I should add) that Indigenous Australians do worse at school, are more likely to be incarcerated, take worse care of their health, are less likely to be employed, etc.

Careful studies show that most of the difference between Indigenous and non-Indigenous Australians is due to socioeconomic disadvantage exacerbated by past policy failures. But, there is a danger that such differences are internalised and seen as something inherent to the individual or the community, as opposed to wider structures and settings. There is a strong argument for a carefully designed RCT on specific policies to maximise expectations, especially amongst the young.

Ultimately, there are limits to what development economics (or any development theory) can tell us about improving the lives of minority groups in high income countries like Australia. What we can say though is that economics and public policy needs a healthy dose of humility and a recognition that we don’t really know what types of policies work nor do we always know how to implement policy effectively. Trialling an intervention and failing should be a vote winner, not a vote loser. Or, according to the authors of Poor Economics on page 272:

‘If we resist the kind of lazy, formulaic thinking that reduces every problem to the same set of general principles; if we listen to poor people themselves and force ourselves to understand the logic of their choices; if we accept the possibility of error and subject every idea, including the most apparently commonsensical ones, to rigorous empirical testing, then we will be better able not only to construct a toolbox of effective policies but also to better understand why the poor live the way they do.’

Posted in Book reviews, Indigenous populations | Leave a comment

Poor (Indigenous) Economics – Part I

Posted on February 3, 2013 by nicholasbiddle

Last Tuesday (January 29^th) I was lucky enough to have scored an invite to a conference organised by the US Studies Centre (out of Sydney Uni) and Harvard University titled ‘Evidence‐Based Policymaking: Best Practices for Policy Design and Evaluation.’ While there were a number of interesting presentations from Australian and international researchers, the highlight for me was the opportunity to hear Professors Abhijit V. Banerjee and Esther Duflo speak. They are both co-authors of perhaps the most interesting non-fiction book I have read in quite a few years – Poor Economics.

Banerjee and Duflo’s book provides a fantastically engaging summary of their research and the research of others, looking at the complexity and policy challenges of improving the lives of those living on less than 99 cents per day. The paperback version of the book that I purchased has (at my count) 28 quotes of praise from a wide range of media outlets including The Economist, The Guardian, The Wall Street Journal and The Financial Times. It also includes quotes from individual authors and academics like Steven Levitt and Matthew Yglesias. However, Nobel Prize winner Amartya Sen gives the most succinct quote: ‘A marvellously insightful book by two outstanding researchers on the real nature of poverty.’

Much of the insights in the book come from India, a country I have a particular interest in from my own travels and the fact that that is where my wife comes from. However, reading through the book, I couldn’t help but draw insights for my own research on Indigenous Australian populations. It is true that, living in a rich country with an extensive welfare system and safety net, there are few if any Indigenous Australians who would be living on less than 99c per day. However, many of the challenges faced in a development setting also have relevance for Indigenous policy. Furthermore, Indigenous policy in Australia tends to be susceptible to grand narratives based on theory or ideology. I’m thinking of self determination, passive welfare, practical reconciliation and so on.

Theory and ideology have their place. However, if there is a key insight from the book it is that grand narratives need to be rigorously tested and challenged. Banerjee and Duflo use experimental methods to do just this. Along the way, they raise a number of questions for Indigenous policy. The following is a chapter by chapter presentation of such questions beginning in this post with discussion o Part I of the book. A Poor (Indigenous) Economics, so to speak.

Chapter 2: Food

Along with shelter and warmth, having a sufficient caloric intake is one of the basic necessities of life. However many of the poorest people living in developing countries do not have a sufficient intake to be able to perform necessary tasks or, in the case of children, reach their potential physical and intellectual development. This is reasonably well established. However, what Banerjee and Duflo show is that, for the majority of the world’s poor, access to sufficient calories is not really an issue. The major problem is that the poor choose to spend what money they do have on higher quality calories (for example sweetened milky tea in India). What are most lacking in the diets of the poor are micronutrients (like iodine and iron) and deworming tablets. In order to ensure the poor are adequately fed, we need to understand why people choose the food that they do.

In terms of implications for Indigenous populations, the main question that is raised is have the Indigenous nutrition programs that are currently in place been rigorously evaluated or, as is the case with many of the programs in developing countries, are they a relatively inefficient way to improve the health of Indigenous children? More broadly, with the focus in the Northern Territory on income management and fresh food, are the diets of the Indigenous population living in remote parts of the country adequate in terms of micronutrients.

Chapter 3: Health

The main insight from this chapter is that people in rich countries (or in relatively well off social groups within those countries) are just as bad at managing their own health as those in poor ones. They don’t go to the gym as much as they feel they should, eat the wrong foods, smoke when they know they shouldn’t and don’t go to the doctor when they know they should. This is despite the many things in relatively affluent societies that make it so much easier to stay healthy – clean water comes into the house, the sewage system takes waste water out, doctors can generally be trusted, and infectious disease is rare and usually curable. Most importantly, we don’t have to worry too much about where our next meal is to come from.

The implication of this is that people everywhere suffer from poor health outcomes that are of their own making. This comes in part from what behavioural economists have called ‘hyperbolic discount rates’ where the present or the near future is valued much more highly than the medium or long term. These discount rates and biases need to be taken into account in health policy through the careful use of incentives and nudges.

From an Indigenous policy perspective, where the headline target of government is to close the gap in life expectancy within a generation, too little attention is paid to incentives and the motivations for using or not using a particular intervention. In all settings, information alone will rarely suffice in terms of changing behaviour.

Chapter 4: Education

One of the points made in Poor Economics is that schools in poor countries are more often than not available. What’s more, they tend to be free, at least at the primary level. This is particularly true in Indigenous context where for the majority of the population, access isn’t really an issue. Despite this, enrolment rates are often quite low and absentee rates very high.

There is no doubt that the supply of education matters (school buildings, teachers, books, iPads). But, most proper evaluations show that the demand for education can often matter more. For example, CCTs or Conditional Cash Transfers (where the government provides additional welfare assistance conditional on certain education or health behaviour) have been shown to increase the demand for education and hence the level of schooling. But, there is some evidence that it is the cash transfer itself, rather than the conditionality that matters. Furthermore, the long term (behavioural) evidence on CCTs is mixed.

One point in the chapter that I hadn’t thought of was that there is a perception that later years of schooling have higher returns than earlier years. This may be an example of focusing on average as opposed to marginal returns. What it means though is that parents (or the wider community) may put all their eggs in one basket, so to speak, and invest most heavily in who they see as the most talented member of the family or community. This is great for those who get the investment. However, it ignores the large potential returns in terms of health, employment prospects and income that would come from making sure the less talented receive at least a solid basic education. This has relevance for the debate in the Indigenous context around boarding schools. Few would argue that the highly talented Indigenous student should have the same chance to become a doctor, lawyer or (god forbid) economist as a non-Indigenous student. But, this should not be done at the cost of the majority of Indigenous students where the potential returns from a little bit more quality education are quite high.

The authors also provide a range of evidence that suggest that inputs into schools don’t have that big of an effect on their own. This includes textbooks, extra teachers or even laptops. Rather, they need to be accompanied by changes in pedagogy or incentives that make children want to learn and engage. Several colleagues of mine at CAEPR (including Dr. Inge Kral and Dr. Jerry Schwab) have demonstrated this quite convincingly in an Indigenous context.

There are a number of broad conclusions for education in many contexts from Poor Economics including:

Focus on basic skills and a commitment to the idea that everyone can master them;
It takes very little training to be an effective remedial teacher. That of course doesn’t take away from the highly skilled role that teachers can play in later years of schooling;
Large potential gains are to be had by reorganising the curriculum and the classroom to allow children to learn at their own pace;
Set more proximate goals for children and teachers; and
Technology can be effective, though not as effective as good teachers.

There is a fantastic quote on page 95 of the book regarding education: ‘Among all those people who drop out somewhere between primary school and college and those who never start school, many, perhaps most, are the victims of some misjudgement somewhere: Parents who give up too soon, teachers who never tried to teach them, the students’ own diffidence [that is, returns to education are almost always positive and large] … The slots that they left vacant were grabbed, in all likelihood, by mediocre children of parents who could afford to offer their children every possible opportunity to make good’

Chapter 5: Fertility

According to Banerjee and Duflo, people in poor countries usually have the capacity to make fertility decisions. Although these decisions don’t appear to be rational, they do make sense when you look at intra-household incentives. Although averages are likely to hide significant variation, males tend to want more children than females, so additional control for females within the community and (in particular) within the household can have a dampening effect on fertility rates. One incentive for having too many children is the lack of a social security system or safety net in old age.

One potentially controversial finding discussed in Poor Economics is that it doesn’t appear that large families have a negative effect on kids in the family. This is somewhat counterintuitive and differs from most development theory which presumes that with a finite set of resources, each additional child in the family will receive a lower investment in terms of health and education. However, evidence for this view has come mainly from observational data which doesn’t control for selection effects (where kids in larger families are likely to be different in unobserved ways). Instead, when ‘researchers have tried to focus on instances where the increase was in part beyond the control of the family…they found no evidence that children born in smaller families are really more educated’ (p. 108). It does appear, however, that larger families have negative consequences for the mother.

Indigenous females in Australia tend to have higher fertility rates than their non-Indigenous counterparts. However, contraceptives are widely available in Australia as is the information required to make fertility decisions. In order to understand the high fertility rates, we therefore need to think about the particular incentives to have additional children and, the lack of disincentives when the mother is young (lack of educational opportunities and lack of employment). We also do not have any research which exploits exogenous variation in fertility decisions to test the causal effect of such large families on child and maternal outcomes.

In the next post, I will look at some of the more economics-y chapters in Part II of the book, including those on risk, lending, savings, entrepreneurship and politics.

Posted in Book reviews, Economics of Education, Gender differences, Indigenous populations | 1 Comment

The Number Cruncher

Education and income inequality in Australia

Budget night and a profile of Australian Public Service officers

Attacking educational disadvantage through school funding (co-authored with Timothy Cameron)

The Economics of Education

The National Plan for School Improvement

Limitations of the National Plan for School Improvement

Indigenous population – Initial projections and implications

Advantage and disadvantage across the country

Poor (Indigenous) Economics – Part II

Poor (Indigenous) Economics – Part I

Recent Posts

Archives

Categories

Meta

Follow “The Number Cruncher”