Value-added indicators are a more accurate method of assessing school performance since they eliminate more non-school factors (Meyer et al., 2017). Slovenian upper secondary schools in the General education track finishing with General Matura have been able to assess value-added measures and track changes over time since 2014. The two time points in question are achievement at the end of Grade 9, just before entering upper secondary schools, and achievement at General Matura examinations. Lower secondary schools can similarly check value-added between Grade 6 and Grade 9 (finishing grade) in different subjects since 2018. These measures are not part of any accountability scheme and are provided for schools’ self-evaluation purposes along with other achievement results.
Value added measures used by slovenian schools are calculated as average residual between actual and predicted students' achievement (Cankar, 2011). Usual method for calculation of predicted values is 'median method', where population is sorted on scores from time point 1 and divided into equal sized groups. Median scores on scores from time point 2 over all groups constitute a series of predicted values for midpoints of groups on timepoint 1. All other values are interpolated from those.
This method of calculating value-added proved robust and relatively straightforward for explaining to teachers and the general public. However, there are also indications that the calculations aren’t optimal. In this paper, we will address the issue of negative national average value-added measures. As a rule, the average value-added over all schools for a chosen year tends to be negative.

There can be many reasons, and within this presentation, we will explore the following research questions:

Could the observed negative average value be associated with school composition factors (primarily the size of the school)?
Could we associate negative average values with school background characteristics (for example, the average income of the school’s municipality)?
Could we associate negative average values with the implementation of the value-added measure (either the median method or some other step in computation)?
Since value-added models rely on differences between time points, which increases the measurement error (Papay et al., 2011), it is important to reduce additional sources of error and insights from this research might be useful to others in the field of value added models.

To address the mentioned research questions, we will use simple regression techniques or hierarchical linear regression where needed. Data on external examinations and national assessments to calculate value-added measures will come from the National Examinations Centre, while the data on municipalities will originate from the Slovenian Statistical Office. We will use value-added measures for the last five years to demonstrate the stability of findings over time. Data will we used and analyzed in a responsible manner to protect individual privacy and adhere to legal requirements. This is especially important since the data on whole cohorts of students will be used.

Value-added models can provide important information and identify underperforming schools, as demonstrated by Ferrão and Couto in the case of Portuguese schools (2014). We expect to provide insight into the problem and either identify the causes of constant negative averages or propose further steps needed to explore and resolve the issue. As value-added measures are also present in other European countries, this research will help other researchers evaluate their value-added models and contribute to a better understanding of the field.

Cankar, G. (2011). Opredelitev dodane vrednosti znanja (Izhodišča, primeri in dileme). In Kakovost v šolstvu v Sloveniji (str. 431). (2011). Pedagoška fakulteta. http://ceps.pef.uni-lj.si/dejavnosti/sp/2012-01-17/kakovost.pdf Ferrão, M., & Couto, A. (2014). The use of a school value-added model for educational improvement: a case study from the Portuguese primary education system. School Effectiveness and School Improvement, 25, 174 - 190. https://doi.org/10.1080/09243453.2013.785436. Koedel, C., Mihaly, K., & Rockoff, J. (2015). Value-added modeling: A review. Economics of Education Review, 47, 180-195. https://doi.org/10.1016/J.ECONEDUREV.2015.01.006. Meyer, R. (1997). Value-added indicators of school performance: A primer. Economics of Education Review, 16, 283-301. https://doi.org/10.1016/S0272-7757(96)00081-7. Papay, J. (2011). Different Tests, Different Answers. American Educational Research Journal, 48, 163 - 193. https://doi.org/10.3102/0002831210362589.