Archive for September, 2012

Changing the exam system

September 27, 2012 1 comment


Author: Simon Burgess

It will take a while to fully understand the scope of the proposed changes to the exam system in England.  The exams are obviously very important for students, but also for schools and teachers. Here are some initial thoughts.

A key pledge is to ‘restore rigour’: exams are going to be more rigorous (of course, England has always had at least one exam noted for its rigour. More rigorous might mean different things: it might simply mean more failures or more broadly, greater differentiation between students.

It seems inevitable that one implication of this (in the short run at least) is greater educational inequality. One outcome of making an exam more rigorous is that more students will fail it. It seems hard to see how that could not be true. This is not fatalism; it may well be that over time, teaching brings more students above the line, but there will be a number of cohorts first with more failures. A tougher exam will not be credible if more people pass it. This coupled with a hint of no retakes seems particularly harsh. ‘No second chances’ is an uncompromising view of how to run an education system, and perhaps at odds with much of the rest of public life.

An alternative sense of ‘more rigorous’ is that the new exam will allow for greater differentiation between students. This is as much to do with the marking and reporting as the nature of the exam itself. Rather than proliferating letters and multiple stars, one way to do this that has been mooted is to simply report percentages. More or less differentiation – separating or pooling in economics terms – has a number of implications. On the plus side, it may enable more efficient matching of workers and jobs, raising average productivity; it is likely to incentivise greater effort at school as students currently more-or-less assured of an A* grade push on to shoot for 90%+. On the down side, there will be greater inequality of attainment and consequently greater inequality of earnings. It is not obvious how sizeable these pros and cons are, but there is no guarantee that the outcome will be a happy one.

We have a school system described by the Department for Education as increasingly autonomous.  Diversity of supply is the maxim, and there is reduced oversight of what schools actually do. But what brings the system together is the common exam system. Autonomous schools can do their own thing with less outside ‘interference’ … as long as its students achieve good grades. Schools’ freedom to operate is constrained by close public scrutiny of their performance in common high stakes exams. If there is less monitoring of the process of education, then regulation of the outcome (attainment measures) becomes even more important.

School performance tables, the ‘league tables’, and Ofsted are the central performance components of the regulatory framework. While Ofsted is important, it is largely driven by the school performance tables reporting exam scores. So getting the exams right is very important. Schools will react to changes in the regulatory environment as they will react to any incentives (implicit or explicit) embedded in the system. So in addition to its implications for students, the exam system will very strongly influence how schools operate. Changing the exam system is a big opportunity to get things right or to get things wrong.

One possibility would be to use the reform to move away from the current sharp threshold recording attainment, the fraction of students in a school getting at least 5 good passes. If the students are given percentage scores not letter grades (see below), the metric for the school could simply be the mean score. There is a lot to be said for not having a sharp threshold, or for locating it at the point that policy makers truly do want schools to focus. But this could be done now under the present system – GCSEs have continuous scores that are then converted to letter grades (controversially this year). Switching from a threshold to a continuous average is a separate issue to the rest of the proposed changes, and is not dependent on them.

The proposal to have one exam board per subject seems like a good idea. While monopolies may well raise costs, the nature of the competition between the exam boards was dysfunctional. And the idea of ‘competition to be a monopoly’ is well established (train companies, power transmission and other infrastructure companies) and reasonably well understood.


Focussing now on the students, what is the best form of tests for students? The proposals favour a return to a single exam for each subject at the end of the course, and the end of coursework, retakes and modules. Doing well in such exams will require a particular type of skill, so these are what students will try to master and what schools will try to teach.

What sort of skills should education foster? This is not a straightforward question. One plausible answer is that education should test and certify the levels of competence that students have achieved in the skills that employers want.

No one knows what skills business will need in ten years time, and so we can only speculate as to what we should grade students on. But given the ubiquity of the web, it seems very unlikely that ‘remembering large amounts of information and writing it down quickly’ will figure high up on the list. It is hard to see this as a highly prized capability. Some other much-lauded education systems  (keen on rigour) test abilities more likely to be of value. A great deal of research now focusses on different cognitive and non-cognitive skills, how they are built up and how they relate to inequality. I don’t know if there is any evidence whether the ability to remember large amounts of information is more or less socially graded than broader ranges of skills.

One argument for one-shot, high-stakes, closed-book exams is that anything more open is susceptible to parental help, and thus more likely to favour the middle class. But there are other ways to make sure that the summative assessment is solely based on the student’s work alone. Furthermore, parents support their children’s education in so many ways (conversation; providing books and computers; a quiet place to study; role models; trips) and this is unlikely to be the most important. It is worth emphasising that parents helping their children to learn is a good thing, and should be encouraged where lacking, not banned where present.

This preliminary set of points suggests that the proposed reforms of assessment will not promote skills likely to be valued in the labour market. They are likely to lead to more students failing and leaving school with nothing, and/or delaying their one-shot take, or just dropping out. This spells greater inequality in educational attainment and so in life chances.


September 25, 2012 Leave a comment

Author Paul Gregg

Predistribution is a manufactured word which featured heavily in a recent speech by Ed Miliband to a Policy Network organised conference around tackling Britain’s current economic woes and longer term challenges. It was coined by Jacob Hacker a professor of political sciences at Yale and is a play on redistribution but seeks to emphasise the role of government and wider decision making in reducing inequality at source.

Economists make a distinction when discussing income inequality between inequality in the primary (or private) distribution of work, wages and occupational pensions, and inequality observed after the secondary redistribution effects of taxes and benefits. Predistribution is thus concerned with the potential for influencing the primary distribution of incomes so as to reduce the need for heavy intervention in terms of higher taxes and benefits. It is thus analogous to prevention strategies in say public health that seek to reduce smoking and improve diets so as to reduce the need for costly health interventions later.

It is perhaps not fully appreciated that the large rise in inequality in the UK in the 1980s and early 90s was not primarily driven by reductions in higher tax rates or restricting benefits to grow in line with prices rather than earnings which had been the previous norm. Rather it was driven by a growing polarisation of work into dual earner and no earner families, with some 20% of working age families having no earner by 1995 up from around 8% in the mid-1970s. This was combined with rapid divergence in wages between high and low waged earners, such that real wages for typical workers (those in the middle of the wage distribution or median) rose by 23% between 1979 and 1995, but for those near the bottom (10th percentile), the rise was just 12% and near the top (90th percentile) wages grew by an impressive 40%. For the very highest paid, the 1% of the population with the highest earnings, the growth was considerably faster still. In contrast the growing access to second stage occupational pensions over and above the state pension saw pension incomes grow rapidly and one of the poorest groups in society caught up somewhat with other families.

What surprised many economists over Labour’s period in office was that the return to near full-employment did not see a significant reversal of the wage inequality built up in the 1980s and 90s. In addition the boost in the numbers of graduates and reduction in the numbers leaving school with few qualifications only halted the growth in the wage gap between better and less educated rather than reversing the previous trend. So in Labours term of office overall inequality edged up despite major redistributional efforts under tax credits and the Pension Credit. This was primarily after 2002 when wages stagnated for ordinary workers and the share of national income going to workers, rather than in profits and self-employed incomes, started to fall quite sharply combined with continuing rapid increases in earnings among top earners.

Also in this period efforts to reduce child poverty were hampered by the fact that low earning males had wage growth not only below the higher paid but also lower paid women. This meant that having a second earner became increasingly important and single earner couples, where the sole earner is usually the father, saw increased poverty rates despite tax credits. This was combined with a continued growth in lone parenthood and so resdistributional efforts were partly baulked by continuing adverse trends in the primary distribution of work and wages.

Of course, as with many buzz words or phrases the underlying concept is not new. Governments of all colours have pursued policies to achieve such ends before. The National Minimum Wage to reduce wage inequality, welfare to work policies to reduce the number of working age families with no earner and support for widespread occupational pension schemes all seek to increase incomes for key groups so as to reduce the need for secondary top ups by government.

The current political narrative plays along the lines that the high current deficits and large overall debt level means that the room for progressives to act through spending is likely to be severely constrained. Hence reducing inequality in primary distribution both reduces the need to act through redistribution but also frees up resources to address other goals. The politics also suggests a sense that the Blair-Brown era of being relaxed about the rich getting richer but to use the proceeds of growth to address poverty through redistribution has passed.

Predistribution, however, opens up a new set of policy tools but also a key constraint. The policy tools over and above tax and benefits are legal regulation, such as minimum wages, and public campaigns or consumer actions around fairness such as the living wage or fair trade campaigns.

There is also the use of broader public spending to achieve social goals such as the use of conditions attached to public procurement or focusing schools on reducing the extent low educational achievement.

Finally, there is the use of competition policy to reduce prices of goods and services that affect the poor more such as rents, energy and food or framing effects such as opt out rather than opt in pensions such as NEST to raise self-protection against low incomes in old age.

The key constraint is that they are indirect effects, and indirect interventions often lack the power to overturn the deeper processes already at work. Will a Living Wage campaign backed by public sector procurement achieve the scale to overturn the steady rise in wage inequality in the UK? Will shareholder activism combined with rules around binding votes for remuneration packages of top executives halt the rise in pay unrelated to firm performance?

Furthermore, as with all preventative strategies you will need to be changing the outcomes of a much larger group than just those who would have need for treatment after the event and this can often mean the costs are higher than for waiting for events to unfold. For example, most of the lowest paid are not poor, as they live in families with other earners. Raising pay levels for all the low paid thus involves far more resources being shifted than the amounts involved in tax credits targeted to address poverty.

The case for a focus on Predistribution thus rests on three factors. First, there is a capacity to shift far larger amounts of economic resources than can be moved by redistribution in the tax and benefit system and hence this makes up for poorer targeting. Second that shifting these resources comes with at least little or no economic loss from reduced economic efficiency and third that the political space for action here is substantially greater than for tax and benefit redistribution.

In terms of specific policies, there are a number of obvious policy areas but whether they are of sufficient scale to address the forces driving rising inequality is unlikely. Reducing long-term unemployment, and especially for young people, is an obvious win as time and again it has been shown how unemployment damages future earnings and employment years after a person first returns to work. Expanding the coverage of occupational pensions through NEST type schemes backed with restrictions on management fees chargeable by fund managers to make them better value to low wage savers offer hope of reducing inequality in old age. Limiting the higher prices charged by energy firms for payment systems other than direct debits, such as Charge Keys. Living wage campaigns backed by more extensive use of public procurement conditions in contacts etc. All of these offer attractive attempts to shift inequality in work and wages, but appear limited in scale or ambition.

More challenging would be boosting employment, but in particular getting these extra jobs focused on groups that the last recovery barely reached such as people in the most deprived areas, the least educated, the disabled and the over 50s. The potential wins here are large but the key to getting employment up for older and disabled workers must be in firms offering flexibility that suits workers. As with mothers, right to return after illness and right to request part-time working are likely to be central to boosting employment but run counter to the current drive for further labour market deregulation. Likewise with training, the current high job turnover in low wage sectors discourages both firms and workers from training, pushing towards long-term and investment focused employment contracts is likely to be central to Predistribution strategy but runs against the current employment regulation model in the UK. Focusing school resources on reducing Britain’s long tail of underachievement risks alienating middle class parents but again tackling low attainment must be central to a Predistribution strategy. Thus a Predistribution strategy could represent a profound challenge to the neo-liberal economic model that has operated in the UK and US over the last 30 or a limited rationale to avoid tough choices about redistribution in tough times.

Novices and Veterans: What new data tells us about teacher turnover and school deprivation

September 20, 2012 1 comment

Rebecca Allen and Simon Burgess

A new school year has just started, new students have just arrived – what about new teachers? Are there a lot of new faces in the staffrooms? One of the stories frequently told about schools serving poor communities is that they suffer from very high and damaging staff turnover. Few teachers stay a long time, and, relative to schools in the affluent suburbs, there is a constant ‘churn’ of staff. This lack of experienced teachers reduces the chances of new teachers learning the trade on the job, and means that both students and school leaders are forever coping with new names, personalities and teaching styles.

Is this true or urban myth? For the first time, we can start to answer this question systematically, moving beyond a collection of local anecdotes. New data collected from all schools about their workforce has the potential to hugely improve our understanding of teachers and teacher labour markets.

We use this data to analyse the length of time that teachers stay in schools, i.e. their job tenure in a particular school. This is the form in which the problem faces headteachers: how many novices do they have, how many veterans and so on. While there is a good deal of research on teachers leaving the profession as a whole, the issue here is how long teachers stay in a specific school.

Like many urban myths, there is a hint of truth in this view that deprived schools experience greater teacher turnover, but not much.

Overall, it seems that teaching is not a low turnover profession. In a typical school, about a fifth of teachers have been there for less than 2 years, and over half of the teachers have been in that school for less than 5 years. On the other hand, nearly a fifth have been there at least ten years, and in fact over 5 percent have stayed over 20 years. Of course, teachers vary and we compare different groups in Table 1. There is very little difference in tenure between female and male teachers, nor between primary and secondary schools. More detail still is available here

Table 1: Years in current school (%)


All teachers





0-2 years






2-5 years






5-10 years






10 years or more












Note: classroom teachers only, excluding assistant, deputy and headteachers

Averaging over all teachers, the mean time in a school is 6.7 years. Here we need to introduce a technical issue.  The data come from teachers in schools, so this is job tenure so far, elapsed tenure. Obviously, a teacher who has just arrived at a school may go on to spend the rest of her career there. Under certain circumstances, the statistical model implies that the average completed tenure is double the average elapsed tenure.

That is the overall picture, what of the differences between disadvantaged and affluent neighbourhoods? We find systematic and statistically significant differences in turnover: schools with many poor pupils do have more short-tenure teachers and fewer experienced teachers. However, on average the differences are small: 18% of teachers in the least disadvantaged schools have tenure of 0-2 years, compared to 22% in the most disadvantaged. At the other end of the scale, 20% of teachers in schools in the most affluent neighbourhoods have tenure of at least 10 years, whereas the figure in the most deprived neighbourhoods is 17%.

Figure 1: Distribution of teacher tenure by school deprivation

Distribution of teacher tenure by school deprivation

Figure 1 gives a flavour of the results. It shows the 10th percentile of tenure in school in days (the lowest line in the figure), across the full range of communities in England, from the richest 2% in the far left-hand side point to the poorest 2% in the far right-hand point. The 10th percentile comes out at somewhat less than two years, but more interestingly, is flat. The number barely changes across the entire distribution.  There is a very slight slope in the 25th percentile and in the median values, reinforcing the point that there are systematic differences but that they are quantitatively small. There is a more noticeable difference in the 75th percentile: in schools serving poor communities, there are slightly fewer experienced teachers.

We also use the richness of the data to decompose the relationship between turnover and poverty. We show that part can be accounted for by pupil characteristics, perhaps because students in schools in more deprived areas are harder to teach. Part also is accounted for by differences in the local teacher labour market around each school. Controlling for school, student and teacher labour market factors reduces the association between school poverty and turnover, but does not eliminate it.

The remaining association between teacher turnover and poverty is largely accounted for by teacher characteristics, with the poorer schools hiring much younger teachers on average. How should we understand this?

We interpret this as either deriving from the preferences of young teachers, or as reflecting the low market attractiveness of disadvantaged schools.  There are a number of possibilities. First, it could be that this is a desired career path for young teachers. New teachers may look for their first jobs near to where they trained, which implies predominantly urban and therefore on average deprived, schools. Alternatively it could be a desired career path deriving from younger teachers possibly having more idealistic preferences, and welcoming the opportunity to work in deprived schools. Under these interpretations, the allocation reflects the desire of younger teachers to work in deprived schools, and the higher turnover in such schools derives from this.

The alternative interpretation is a matching story in which the more effective teachers sort on average into the more affluent schools, and the disproportionate number of inexperienced teachers in the poorer urban schools reflects the realities of the market that these schools face. Distinguishing between these interpretations is a task for future work; it will need further sweeps of the data and possibly attitudinal data from teachers as well.

It is now widely acknowledged that teacher effectiveness is the single most important factor in raising attainment. Attainment gaps arise in part from students’ exposure to teachers of differing effectiveness. The process by which different teachers end up at different schools in front of different children is little understood. We intend to spend the next few years utilising this new data to address this research programme.