For the past three decades, mechanisms and policies for the regulation of education have been developed and pushed forward in the education agenda worldwide, with varying degrees of intensity according to time and geography. One of the key instruments for the enforcement of regulatory pressures on education agents is the standardized national and international testing of students in varied subjects, from the outcomes of which arise key indicators for decision-making in education policy. A specific modality of external evaluation of school systems, school organizations and pupils themselves has been on the rise globally, and it has intensified in Portugal during the last parliamentary term which covered a four-year period (2011-2015) of what can be considered neoconservative education policies, since the political discourse behind them postulated the downfall of standards in public education and the need to increase selectivity to assure quality and promote excellence (Apple, 2005; Au, 2011; Ball & Youdell, 2008). The introduction of standardized tests with high-stakes for students in 4^th grade was one of the major policy initiatives tied to this rhetoric. These tests allowed the central government not only to collect student-level data on two subjects – Portuguese Language and Mathematics –, but also to compile and publish indicators of school success based on these results, as well as create rewards for schools who surpassed the national average in the form of extra hours per week allocated to spend on extracurricular study aid; furthermore, they also served to make decisions on students’ school paths, counting 30% towards the final score.

Despite the global push for increased accountability in education, extensive research on the relation between these policies and social inequalities shows that such a focus on the performance of agents and schools, when tied mainly to standardized testing of pupils, might have perverse effects on several levels. One of them is the increasing stratification of the school network, as schools are pushed to perform better, therefore searching for and attracting audiences with differing sociocultural backgrounds – the between-school effects of accountability policies (Van Zanten, 2005; Ball & Youdell, 2008). Much is already known about the ways in which school organizations and their agents – deans, teachers, pupils, parents – respond to the pressure of rankings and public perception of their “quality”. Diamond & Spillane (2004) discovered that the schools’ position in the regional ranking as well as their social composition are key factors to define their level of compromise towards equality of opportunities in the face of accountability pressures; namely, lower-performing schools with a higher concentration of disadvantaged children increase selectivity and educational triage in an effort to raise results by “betting” in a few “winning horses”. Van Zanten (1999) found that schools’ competitive search for the “ideal costumer” intensified when standardized testing regimes were in force, with much more socially selective pupil-enrollment criteria which in turn led to the entrenchment of socially homogeneous education territories. Despite this, not many studies have focused on the relation between school stratification, social composition and national testing outcomes.

It is our goal to contribute to this ongoing debate. To do so, we analyze the effects of a national policy measure of accountability – the introduction of high-stakes standardized testing in 4^th grade – on a Portuguese municipality’s school network, its social composition, and their outcomes. Is it possible to observe a social reconfiguration of the municipality’s school network that reflects more intense and concerted mechanisms for the selection of pupils in primary education? If so, how does this reflect on schools’ average outcomes? Furthermore, are the results of schools with different social compositions differently affected by accountability policies?

We employ an extensive approach to the problem presented, based on two databases produced by the Portuguese office of education statistics (DGEEC) with data for all students enrolled in 4th grade public schools in the Almada municipality (around 1600 students and 40 schools). Each database concerns a different school year: 2009/10, when a formative national assessment was at place with no accountability measures attached; and 2014/15, one year after a national test with consequences for schools and pupils was implemented. Each database has a set of student-level variables which allowed for the construction of composite school variables of social characterization, namely the average percent of parents’ educational levels and the average percent of pupils from several national origins. Standardized test outcomes varying from 1-5 were also available per student and school. We use these variables to make a comparative analysis of the statistical dispersion of parents’ educational levels, as well as schools’ average results, before and after high stakes test-based accountability policies.

From 2009/10 to 2014/15, we found that inequalities grew within the local primary school network. Despite the general rise in educational qualifications of adults in the municipality in this four-year period the gap in parents’ educational background grew between schools that are highly sought-after by socially affluent families and those which absorb mostly children from marginalized populations. This suggests that schools, parents and possibly local educational authorities are indeed responding to accountability pressure, increasing social selectivity “at the gate”. At the same time, a larger dispersion of average school results in standardized tests in 2014/15 than in 2009/10 shows that inequalities in school outcomes also intensified. The most socially disadvantaged schools suffered the most, since their chances of falling behind the municipality’s benchmark average increased after the high-stakes test was introduced. Instead, it is among socially heterogenous schools – not in schools with mostly highly qualified populations – that we observe the least dramatic drops in outcomes, and sometimes even surprising gains, against the local trend. These results corroborate some of what has already been found about the meso-level effects of high stakes test-based accountability policies, while also introducing several new important elements for future researchers and policy-makers to consider. It seems like there is indeed an increase in school competition which in turn leads to a rise in social stratification. Such an effect may be due to a strategy to contain failure in a few sites and raise outcomes in all other schools. However, we argue that this strategy does not seem to favor average outcomes in high stakes tests – they dropped for most schools during the period analyzed, bringing the municipality’s average down.

Apple, M. (2005). Doing things the ‘right’ way: Legitimating educational inequalities in conservative times. Educational Review, 57 (3), 271-293. Au, W. (2011). Teaching under the new Taylorism: High-stakes testing and the standardization of the 21st century curriculum. Journal of Curriculum Studies, 43 (1), 25-45. Ball, S., & Youdell, D. (2008). Hidden privatisation in public education. Retrieved from Campaign for Education website: http://www.campaignforeducation.org/docs/privatisation/Endogenous%20Privatization%20Stephen%20Ball_ENGLISH.pdf. Diamond, J., & Spillane, J. (2004). High-stakes accountability in urban elementary school: Challenging or reproducing inequality?. Teachers College Record, 106 (6), 1145–1176. Zanten, Van A. (1999). Les chefs d’établissements et la justice des systèmes d’enseignement en Angleterre et en France: Les pratiques et les éthiques professionnelles à l’épreuve de la concurrence entre établissements. In M. Meuret (Ed.), La justice du système éducatif (pp.135-155). Bruxelles: De Boeck. Zanten, Van A. (2005). New modes of reproducing social inequality in education: The changing role of parents, teachers, schools and educational policies. European Educational Research Journal, 4 (3), 155-169.