| Study | Used/reason not used in data synthesis | Treatment year (s) | Country |
|---|---|---|---|
| Achilles, 1995 | Too high risk of bias on the confounding item | 1991‐1994 | USA |
| Akerhielm, 1995 | Too high risk of bias on the confounding item | 1988 | USA |
| Angrist, 1999 | Too high risk of bias on the confounding item | 1991 | Israel |
| Angrist, 2014 | Too high risk of bias on the confounding item | 2009‐2011 | Italy |
| Annevelink, 2004 | Too high risk of bias on the confounding item | 2000‐2001 | NL |
| Blatchford, 2002 | Too high risk of bias on the selective reporting item | 1996/97 | UK |
| Blatchford, 2003a | Collection of results from British Class Size Study. Cannot assess RoB as not enough information is provided. Only one effect size reported (but not number of observations used, so cannot calculate standard errors), the rest reported as NS or a narrative description such as ‘there was found to be an effect’. | 1996/97 and maybe 1997/98 | UK |
| Blatchford, 2003b | No results reported other than graphs without CI. | 1996‐1999 | UK |
| Bonesrønning, 2003 | Too high risk of bias on the confounding item | 1998‐2000 | Norway |
| Boozer, 1995 | Too high risk of bias on the confounding item | 1988 | USA |
| Boozer, 2001a | Too high risk of bias on the confounding item | 1985‐1990 | New Zealand |
| Boozer, 2001b | Too high risk of bias on the confounding item | 1988 | USA |
| Borland, 2005 | Too high risk of bias on the confounding item | 1990 | USA |
| Bosworth, 2014 | Not enough information provided to calculate standard errors | 2001‐2002 | USA |
| Bressoux, 2009 | Used in data synthesis | 1991‐1992 | France |
| Breton, 2012 | Too high risk of bias on the confounding item | 1997 | Columbia |
| Burde, 1990 | Too high risk of bias on the confounding item | 1988 | USA |
| Carpenter, 2003 | Too high risk of bias on the selective reporting item | 1996/1997 | UK |
| Chargois, 2008 | Too high risk of bias on the confounding item | 2007 | USA |
| Clanet, 2010 | Only report the significance level and only sign of the effects that are significant | 2001‐2002 | France |
| Costello, 1992 | Too high risk of bias on the confounding item | 1995 | USA |
| Dee, 2011 | Subject specific test score, may be mathematics, reading, science or history but not specified. First difference between subjects is outcome | 1988 | USA |
| Dennis, 1986 | Too high risk of bias on the confounding item | 1985‐1986 | USA |
| Dharmadasa, 1995 | Too high risk of bias on the confounding item | 1989 | Sri Lanka |
| Dieterle, 2013 | Only have data at required level for two of three grades and do not provide useable separate results | 2003‐2004 | USA |
| Dobbelsteen, 2002 | Too high risk of bias on the confounding item | 1994/1995 | NL |
| Ecalle, 2006 | Used in data synthesis | 2002‐2003 | France |
| Galton, 2012 | Too high risk of bias on the confounding item | 2004‐2008 | Hong Kong |
| Gerritsen, 2017 | Used in data synthesis | 1994‐2005 | NL |
| Gilman, 1988a | Too high risk of bias on the confounding item | 1984‐1988 | USA |
| Gilman, 1988b | Too high risk of bias on the confounding item | 1985 | USA |
| Haenn, 2002 | Too high risk of bias on the confounding item | 1994/1995 to probably 2001 | USA |
| Hallinan, 1985 | Too high risk of bias on the confounding item | Not reported | USA |
| Hirschfeld,2016 | Too high risk of bias on the confounding item | 2016 | USA |
| Hojo, 2011 | Too high risk of bias on the confounding item | 2007 | Japan |
| Hojo, 2013 | Too high risk of bias on the other bias item | 2003 | Japan |
| Hudson, 2011 | Used in data synthesis | 1990 | USA |
| Iacovou, 2002 | Too high risk of bias on the confounding item | 1965, 1969 and 1974 | UK |
| Iversen, 2013 | Too high risk of bias on the confounding item | 2003‐2004 | Norway |
| Jakubowski, 2006 | Too high risk of bias on the confounding item | 2002‐2004 | Poland |
| Konstantopoulos, 2014 | Too high risk of bias on the confounding item | 2001 | Greece |
| Konstantopoulos, 2016 | Too high risk of bias on the confounding item | 2003 and 2007 | Cyprus |
| Konstantopoulos, 2016 | Too high risk of bias on the confounding item | 2011 | Multiple1 |
| Krueger, 2002 | Too high risk of bias on the confounding item | 1998‐1999 | Sweden |
| Lavy, 2001 | Too high risk of bias on the confounding item | 1991 | Israel |
| Levin, 2001 | Too high risk of bias on the confounding item | 1994/1995 | NL |
| Li, 2015 | Too high risk of bias on the confounding item | 2011 | Multiple2 |
| Li, 2017 | Too high risk of bias on the confounding item | 2011 | Multiple1 |
| Lindahl, 2005 | Too high risk of bias on the confounding item | 1998 | Sweden |
| Ma, 2006 | Too high risk of bias on the confounding item | 1994/1995 | NL |
| Maier, 1997 | A Regular classroom refers to a classroom with one teacher. Most regular classrooms have 15 or fewer students, but a few exceed 15. A 2‐Teacher Team classroom is a class where two teachers work collaboratively to teach as many as 30 students. A Shared‐Space classroom is a classroom that has been fitted with a temporary wall that creates two teaching spaces, each with one teacher and about 15 students. A Floating Teacher classroom is a room consisting of one teacher and about 30 students, except during reading, language arts, and mathematics instruction when another teacher joins the class to reduce the ratio to 15:1. Only analyse effect of type of classroom within SAGE schools. | 1995‐1996 | USA |
| Maples, 2009 | Too high risk of bias on the confounding item | 2006‐2007 | USA |
| McGiverin, 1989 | Too high risk of bias on the confounding item | 1984‐85 | USA |
| Merritt, 2011 | Too high risk of bias on the other bias item | 2010 | USA |
| Milesi, 2006 | Used in data synthesis | 1998‐1999 | USA |
| Molnar, 1998 | See Maier, 1997 | 1997‐1998 | USA |
| Molnar, 1999a | See Maier, 1997 | 1998‐1999 | USA |
| Molnar, 1999b | See Maier, 1997 | 1996‐1998 | USA |
| Molnar, 2001 | See Maier, 1997 | 2000‐2001 | USA |
| Moshoeshoe, 2015 | Too high risk of bias on the confounding item | 2000 | Lesotho |
| Munoz, 2001 | Used in data synthesis | 1999‐2000 | USA |
| Murdoch, 1986 | Only report p values from a multivariate model (8 outcomes) with CS, age, gender and school, separated by grade | 1984‐1985 | USA |
| Maasoumi, 2005 | No method/results we can use (first or second order stochastic dominance tests) | 1988 | USA |
| Nandrup, 2016 | Too high risk of bias on the confounding item | 2009/2010‐2011/2012 | Denmark |
| NICHD, 2004 | Not enough information provided to calculate standard errors | 1990‐1991 | USA |
| Otsu, 2015 | Relevant results are presented graphically and no ES and SE can be extracted. (Uses selected data of Angrist and Lavy (1999); schools with either one or two classes in grade 4) | 1991 | Israel |
| Pollard, 1995 | Too high risk of bias on the confounding item | 1990‐1992 and 1996‐1997 | USA |
| Pong, 2001 | Too high risk of bias on the confounding item | 1994‐1995 | Multiple3 |
| Sanogo, 1994 | Reproduction of STAR and Indiana PRIME Time results (Word et al. 1990 and Tillitsky, Gilman, Mohr, and Stone, 1988). Do not report what type of classes are included in the PRIME Time results | 1985‐1989 and 1984‐1987 | USA |
| Shapson, 1980 | They do not report outcomes for all groups for all years, so we cannot determine the effect of being randomized to one of the four arms. | 1977‐1979 | Canada |
| Tienken, 2009 | Too high risk of bias on the confounding item | 2001‐2006 | USA |
| Tillitsky, 1988 | Too high risk of bias on the confounding item | 1984‐1987 | USA |
| Uhrain, 2016 | Too high risk of bias on the confounding item | 2012‐2013 | USA |
| Urquiola, 2006 | Too high risk of bias on the other bias item | 1993 | Bolivia |
| Watson, 2016 | Too high risk of bias on the confounding item | 2008‐2012 | Australia |
| Wenfan, 2005 | Too high risk of bias on the confounding item | 1998‐1999 | USA |
| West, 2006 | Too high risk of bias on the confounding item | 1994‐1995 | Multiple4 |
| Wiermann, 2005 | Difference between mathematics and physics test scores (the chemistry/biology and the reading/biology differences scores 5) | 2000 | Germany |
| Wößmann, 2006 | Too high risk of bias on the confounding item | 1994‐1995 | Multiple5 |
| Wößmann, 2003 | Too high risk of bias on the confounding item | 1994‐1995 | Multiple6 |
| Wößmann, 2005a | Too high risk of bias on the confounding item | 1995 | Japan and Singapore |
| Wößmann, 2005b | Too high risk of bias on the confounding item | 1995 | Multiple5 |
1:
Austria, Lithuania, Croatia, Malta, Czech Republic, Portugal, Denmark, Romania, Germany, Slovak Republic, Hungary, Slovenia, Italy and Spain
2:
Austria, Lithuania, Croatia, Malta, Czech Republic, Portugal, Denmark, Romania, Germany, Slovak Republic, Hungary, Slovenia, Italy, Spain, Hong Kong, Singapore, Japan and Chinese Taipei
3:
USA, Canada, Australia, France, Germany, Iceland, Singapore, Korea and Hong Kong.
4:
Belgium Fr., Canada, Czech Rep., Franc, Greece, Iceland, Portugal, Romania, Singapore, Slovenia and Spain.
5:
USA, Australia, Belgium (Fl), Belgium (Fr), Canada, Czech Rep., France, Greece, Hong Kong, Iceland, Japan, Korea, Portugal, Romania, Scotland, Singapore, Slovenia and Spain.
6:
USA, Austria, Belgium, Denmark, England and Scotland, France, Germany, Greece, Iceland, Ireland, the Netherlands, Norway, Portugal, Spain, Sweden, and Switzerland.