Policy Research Working Paper 8802 What Works to Reduce Inequalities in Higher Education? A Systematic Review of the (Quasi-)Experimental Literature on Outreach and Financial Aid Estelle Herbaut Koen Geven Education Global Practice April 2019 Policy Research Working Paper 8802 Abstract Policy makers are increasingly searching for ways to allow but not when they only provide general information on more disadvantaged students to access and complete higher higher education. For financial aid, the paper finds that education. The quickly growing (quasi-)experimental liter- need-based grants do not systematically increase enrollment ature on policy interventions in higher education provide rates but only lead to improvements when they provide the opportunity to identify the causal effects of these inter- enough money to cover unmet need and/or include an early ventions on disadvantaged students and discuss inequality commitment during high school. Still, need-based grants mechanisms at the last stage of the educational system. The quite consistently appear to improve the completion rates of paper reviews 75 studies and rigorously compares more than disadvantaged students. In contrast, the evidence indicates 200 causal effects of outreach and financial aid interven- that merit-based grants only rarely improve the outcomes tions on the access and completion rates of disadvantaged of disadvantaged students. Finally, interventions combining students in higher education. The paper finds that out- outreach and financial aid have brought promising results, reach policies are broadly effective in increasing access for although more research on these mixed interventions is disadvantaged students when these policies include active needed. counseling or simplify the university application process, This paper is a product of the Education Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/research. The authors may be contacted at kgeven@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team What Works to Reduce Inequalities in Higher Education? A Systematic Review  of the (Quasi‐)Experimental Literature on Outreach and Financial Aid  Estelle Herbaut   Sciences Po, Observatoire sociologique du changement (OSC)  27, rue Saint‐Guillaume, 75337 Paris Cedex 07, France.  estelle.herbaut@sciencespo.fr    Koen Geven  Economist, World Bank  1818 H St NW, 20009, Washington DC, USA.   kgeven@worldbank.org                     Keywords:  higher  education;  social  inequalities;  (quasi‐)experimental  methods;  literature  review; outreach; financial aid  JEL codes: I22 (Educational Finance), I23 (Higher Education) I24 (Education and Inequality), I28  (Government Policy)      What Works to Reduce Inequalities in Higher Education? A Systematic Review  of the (Quasi‐)Experimental Literature on Outreach and Financial Aid  1. Introduction  In recent years, equity in higher education has emerged as a central political issue in many  countries,  and  policy  makers  are  increasingly  seeking  policy  instruments  to  support  disadvantaged students in their access to, and completion from, higher education. Building on  recent research in economics, psychology and sociology that has identified the causal effects  of policy interventions in higher education, this systematic literature review aims to provide  an overview of the effects of various interventions on the enrollment and completion rates of  disadvantaged students. It also provides the opportunity to discuss and shed new light on the  mechanisms driving social inequalities in the last stage of the educational system.   This review has three distinctive features. First, we are exclusively concerned with outcomes  of disadvantaged students. Earlier reviews in this field (Heller, 1997; Leslie & Brinkman, 1987)  or  a  more  recent  meta‐analysis  (Sneyers  &  Witte,  2018)  have  assessed  the  effects  of  interventions  on  outcomes  of  any  young  person  in  higher  education.  In  contrast,  we  only  include  studies  that  estimate  an  effect  on  disadvantaged  groups.  We  use  the  term  ‘disadvantaged students’ to refer to a broad class of lower socio‐economic status groups. The  literature alternatively defines these groups as low‐income, non‐white, working‐class, or first‐ generation college students. While there are differences between these groups, there is also  a substantial overlap and a broad definition allows us to capture the relevant literature on  equity in higher education, including the different dimensions of social disadvantage.  Secondly, we focus on both enrollment in and completion of higher education. In recent years,  the  literature  has  increasingly  recognized  that  getting  more  youth  into  higher  education  is  insufficient and that interventions should also ensure that they ultimately graduate (Bettinger,  2004; Castleman & Long, 2013). We thus present effects on both access and graduation in  higher education.   Thirdly, we present a systematic overview of the (quasi‐)experimental literature on this topic.  While a number of research syntheses have summarized empirical evidence on interventions  in higher education, the large majority relies on cross‐sectional evidence. Only a few reviews  2    have  specifically  summarized  the  (quasi‐)experimental  literature  and  their  scope  was  narrower. For instance, Page & Scott Clayton (2016) focus only on college access in the United  States, while Deming & Dynarski (2009) only discuss financial aid. In addition, these reviews  discuss the conclusions of the literature in a narrative form without systemically providing the  estimates on which they are based. The present overview conveys the results in a narrative  form but also rigorously gathers, provides, and compares the causal effects on both access  and completion.    The  present  review  discusses  75  studies  that  provide  causal  estimates  of  the  impact  of  outreach  and  financial  aid  interventions  on  access  or  completion  rates  of  disadvantaged  students in higher education.  Outreach interventions are defined as policies that target youth  in  secondary  education  and  aim  to  raise  participants’  aspirations  and  readiness  for  higher  education. These include interventions that provide information, counseling, and/or focused  academic tutoring in order to increase and facilitate transition to higher education. Financial  aid includes monetary help provided to students to meet, at least partially, their financial need  for  higher  education.  In  this  category,  we  discuss  universal,  need‐based,  merit‐based,  and  performance‐based  grants,  loans  and  tax  incentives.  Finally,  we  discuss  the  effects  of  interventions  which  have  combined  outreach  and  financial  aid.  In  addition  to  outreach  interventions  and  financial  aid  policies,  a  number  of  other  interventions  may  help  reduce  inequalities  in  higher  education  but  the  available  (quasi‐)experimental  evidence  on  their  efficiency is currently insufficient for a literature review and these results are not discussed  here.    2. Barriers faced by disadvantaged students in higher education   Outreach and financial aid may help disadvantaged students to access and complete higher  education if these interventions efficiently address some of the barriers met by disadvantaged  students in higher education. We summarize the most common hypotheses discussed in the  current literature on education inequality mechanisms. These include (1) financial barriers, (2)  lack of academic preparation, (3) lack of information and, (4) behavioral barriers. While there  may  be  additional  mechanisms  that  prevent  disadvantaged  students  from  succeeding  in  higher education (e.g. negative self‐identities or discrimination), these mechanisms are not  specifically addressed by financial aid or outreach programs and are not discussed here.  3    2.1 Unmet financial need  Financial barriers are often at the core of the concerns about higher education opportunities  for disadvantaged students who are eligible for it. The total financial cost of higher education  studies includes both direct costs such as tuition fees and living costs, study materials, and  health coverage, and indirect costs such as foregone earnings.  In some countries, the direct  costs  of  higher  education  attendance  have  risen  dramatically  over  the  last  years  and  have  raised public concern about affordability. In the U.S., between 1985 and 2015, average tuition  and  fees  in  public  four‐year  institutions  increased  more  than  threefold  in  real  terms  (Ma,  Baum, Pender, & Bell, 2015). And this trend is not restricted to the United States. Between  1995 and 2010, in 14 of 25 industrialized countries, governments have reformed the structure  of tuition fees (OECD, 2012). With some exceptions (e.g. Germany), this meant that tuition  fees went up.  Low‐income students seem to be particularly sensitive to the  price of higher education for  both enrollment decisions (Heller, 1997; Kane, 1994) and year‐to‐year persistence (Paulsen &  St.  John,  2002).  Large  unmet  financial  need  makes  students  more  likely  to  work  and  for  a  substantially higher number of hours (Scott‐Clayton, 2012). In turn, investing many hours in  paid work reduces the time students can devote to study and has been shown to be associated  with  longer  time  to  graduate  and  with  a  higher  probability  of  dropout  before  graduation  (Choitz & Reimherr, 2013; King, 2002).   2.2 Unsuitable academic preparation  A  lack  of  academic  preparation  may  be  a  major  barrier  for  disadvantaged  students’  educational attainment  (Carneiro & Heckman, 2002). A large share of these students may  drop out from school, but even among students eligible for higher education, lower levels of  academic preparation and performance can constitute a major hurdle. For example, Greene  and Forster (2003) estimate that in the public high school class of 2001 in the U.S., half of all  black  and  Hispanic  students  graduated  from  high  school  but  only  20%  and  16%    of  them,  respectively, had the minimum qualifications for applying to four‐year colleges. This lack of  academic preparation clearly limits students’ options in terms of accessing selective forms of  higher education (i.e. highly ranked universities).   This  lower  level  of  initial  academic  credentials  can  also  hinder  graduation  from  higher  education.  For  example,  in  the  U.S.,  a  larger  proportion  of  students  coming  from  4    disadvantaged backgrounds need to take remediation courses during their higher education  studies (Sparks & Malkus, 2013). Since there is a lack of evidence about the effectiveness of  remediation, this may reduce these students’ chances of completing their degrees (Attewell,  Lavin, Domina, & Levey, 2006; Scott‐Clayton & Rodriguez, 2014).     2.3 Lack of information  The  lack  of  accurate  information  about  higher education  among  disadvantaged  students is  another plausible mechanism highlighted in the literature. First, students from disadvantaged  backgrounds may underestimate the returns to higher education and overestimate the costs  of enrollment, leading them to underestimate the net returns of a higher education degree.  Focusing  on  the  literature  which  evaluates  expectations  about  earnings  before  students  decide  to  enter  higher  education  (usually  high  school  seniors),  results  on  the  accuracy  of  earning benefits associated with a tertiary degree and on the influence of social background  is mixed (for a detailed summary of the available empirical evidence, see Abbiati & Barone,  2017). For example, in the U.K., high school students were found to make accurate estimations  of the returns of a university degree, independently of their social background (Williams &  Gordon, 1981) and, similarly in Switzerland, no clear patterns of the effect of father’s  level of  education could be identified (Wolter, 2000). In contrast, other studies find that estimated  earnings after a university degree are overestimated by high school students, independently  of  social  origin  (Avery  &  Kane,  2004),  or  that  overestimation  of  returns  is  stronger  among  students coming from advantaged social backgrounds (Abbiati & Barone, 2017).   Regarding  the  estimated  cost  of  higher  education,  the  empirical  literature  has  consistently  shown  that  high  school  students  tend  to  overestimate  higher  education  costs  (Abbiati  &  Barone, 2017; Avery & Kane, 2004; Loyalka, Song, Wei, Zhong, & Rozelle, 2013) and suggests  that  incertitude  or  overestimation  of  the  costs  are  more  common  among  disadvantaged  families (Grodsky & Jones, 2007; Olson & Rosenfeld, 1985; Usher, 2005).   A related problem is the lack of information on how to access financial aid. Financial aid and  its application process are often complex, particularly in the US‐context. Students need to fill  out the Free Application for Federal Student Aid (FAFSA), which, with over 100 questions, has  been criticized for being “long and cumbersome” and deterring disadvantaged students from  applying for financial aid (Long, 2008). In 2000, around 850,000 students who did not file the  FAFSA were actually eligible for financial aid (King, 2004) and lower middle income, white and  5    male candidates were found to be less likely to complete the FASFA even when they were  eligible for it (Kofoed, 2017). Although the complexity of the aid application process has been  mainly highlighted in the United States, the non‐take‐up of financial aid may be a problem  relevant to other national contexts. In Germany, for example, a recent simulation estimates  that around 40% of the eligible low‐income students do not take up their entitlements (Herber  & Kalinowski, 2016).  2.4 Behavioural deficits  Recently,  the  field  of  behavioral  economics,  building  on  findings  from  cognitive  sciences,  neurobiology and psychology, has brought attention to behavioral barriers as an explanation  for suboptimal choices and behaviors in education (Lavecchia, Liu, & Oreopoulos, 2015). These  barriers include present bias, cognitive overload, and routine or status quo bias.   The present bias may explain why some students or families do not invest in education in the  most optimal way. Education is a domain where costs are salient in the present, while benefits  are  more  uncertain  and  time  distant.  If  some  students  give  more  priority  to  immediate  rewards, this may negatively impact enrollment decisions, time devoted to study and dropout  behavior  (Lavecchia  et  al.,  2015).  In  sociology,  the  relatively  short  time  horizon  of  working  class students has been put forward to explain why these students are diverted away from  academic  tracks  in  postsecondary  education  and  choose  lower‐status  tracks  which  are  typically shorter in duration and offer more concrete rewards on the job market, e.g. entering  a specific occupation (Hillmert & Jacob, 2003).  In addition, students may make suboptimal choices regarding their educational career due to  cognitive overload. The paradox of choice highlights that a large set of options is not always  better as people may be overwhelmed by the number of alternatives which are cognitively  costly  to  compare  (Jabbar,  2011).    This  may  be  especially  relevant  in  the  case  of  higher  education  where  the  lack  of  structure  makes  it  especially  difficult  to  navigate  for  students  (Scott‐Clayton, 2011).   Thirdly, the status quo bias suggests that people rely heavily on routine and on the default  option,  not  engaging  in  the  optimal  behaviors  despite  appropriate  information.  In  higher  education, one powerful example of the importance of the default option in shaping behaviors  is provided by a small change in the cost of sending test scores in college applications in the  6    United States in 1997.  When the ACT increased the number of reports that could be send for  free from three to four, the proportion of test‐takers sending four reports rose from 3% to  74%, although the price to send a fourth report before the change was only US$6. This change  in the default option for applications mainly benefited low‐income students who were able to  enroll in more selective colleges (Pallais, 2013).    There is currently little evidence confirming that these behavioral barriers particularly affect  disadvantaged students. It may be that disadvantaged students are more bounded in their  decision‐making  processes  (by  the  lack  of  resources,  information  sources,  lower  reference  points, etc.)  or that they are more affected by the consequences of suboptimal choices (Scott‐ Clayton, 2011). However, the emerging literature suggests that these mechanisms are helpful  to  design  interventions  which  efficiently  trigger  behavioral  changes  among  disadvantaged  students (Ross, White, Wright, & Knapp, 2013).   3. Method  3.1 Inclusion criteria  Three  main  criteria  have  been  used  to  select  relevant  articles  and  reports.  First,  we  only  selected  studies  that  look  specifically  at  the  impact  of  an  intervention  on  disadvantaged  students.  We  only  included  studies  evaluating  interventions  that  were  either  targeted  specifically at these groups or were broader in scope but investigated the heterogeneity in the  effect of the interventions and provided estimates on these groups. Second, we only included  studies  with  a  (quasi‐)experimental  design.  A  “naïve”  comparison  between  educational  outcomes of students participating in an intervention, and those who do not, is likely to lead  to biased estimates, especially in the case of interventions targeted at disadvantaged students  who  differ  from  other  students  in  many  observed  and  unobserved  characteristics.  Thus,  selected  studies  build  either  on  randomized  controlled  trials  (i.e.  formal  experiments),  or  quasi‐experiments  that  analyzed  a  counterfactual  using  appropriate  matching  techniques,  instrumental variables, difference‐in‐differences or regression discontinuity methods. Finally,  we  only  selected  evaluations  of  interventions  which  provided  estimates  on  students’  behaviors in higher education (enrollment or graduation). We excluded all studies which only  evaluated  an  intervention  in  light  of  changes  in  students’  aspirations  or  intermediate  outcomes (persistence, GPA in higher education, etc.).   7    3.2 Literature search  Several strategies were used to find relevant studies. We first reviewed all titles and abstracts  of search results in the following electronic databases: JSTOR, ERIC, WEB OF SCIENCE and the  Pathways to College Online Library.1 We also searched the websites of organizations working  on higher education policies, most notably the Institute of Education Sciences (IES), the policy  research  organization  MDRC,  the  National  Center  for  Postsecondary  Research  (NCPR),  the  non‐profit organization ACT and The National Bureau of Economic Research (NBER). Once we  had  reached  a  starting  set  of  papers  matching  all  our  inclusion  criteria,  we  systematically  reviewed all their references and identified and checked all the studies citing them. We limited  the search to articles or reports in English and published by May 2018. Overall, we reviewed  titles and abstracts of thousands of academic articles, working papers and policy reports. This  yielded an initial set of 296 studies which we carefully read and systematically reviewed on  our inclusion criteria, leaving us with 87 studies which met all the inclusion criteria. However,  12  studies  which  evaluate  interventions  for  which  the  (quasi‐)experimental  evidence  is  currently too scarce to be discussed in a literature review are not presented here. We thus  further  focus  on  the  findings  of  75  studies  which  specifically  evaluate  outreach  programs,  financial aid policies or a combination of the two. The list of the selected studies is presented  in Table A.1 in the Appendix.  Figure 1 shows the distribution of the type of publications, the interventions evaluated, the  (quasi‐)experimental  designs,  and  the  countries  where  the  interventions  were  evaluated  among  these  75  studies.  Randomized  experiments  are  the  most  common  methodology  implemented,  followed  by  regression  discontinuity  and  difference‐in‐differences  design.  In  addition,  the  (quasi‐)experimental  literature  on  outreach  and  financial  aid  comes  overwhelmingly from North America and no less than 60 studies evaluate an intervention from  the  United  States.  The  lack  of  diversity  in  the  educational  contexts  where  interventions  or  policies are tested is already an important result from this review and should be kept in mind  when interpreting the results of these studies.                                                            1  The following search terms were used: (College OR “Higher Education” OR “Tertiary Education” OR University)  AND (Inequality OR Stratification OR Access OR Drop‐out OR Retention OR Persistence) AND (Experiment OR RCT  OR Policy OR Intervention OR Reform OR Effect OR Impact).   8    Figure 1: Characteristics of studies included   Publication              Intervention                  Evaluation design                                 70 63 60 50 42 40 40 35 33 28 30 20 17 15 10 10 6 4 6 5 0 Note: For the interventions, articles evaluating more than one type of intervention are counted more    3.3 Coding  For  each  of  these  articles,  we  coded  the  experimental  design,  the  characteristics  of  the  intervention  (place,  duration,  content),  the  nature  of  the  sample  (eligibility  criteria  for  participation,  assignment  to  control  and  treated  group,  etc.),  and  the  outcomes  selected  (effect  size,  standard  errors,  timing  of  measurement,  model  used  and  baseline  in  control  group). The selection and coding of the studies was first carried out by one coder (allocated  at  random)  and  a  second  coder  then  reviewed  the  initial  codes.  In  cases  of  conflict,  we  discussed  the  disagreement.  In  all  cases,  we  managed  to  resolve  our  differences  after  deliberation.   3.4 Estimate selection  Most studies reported more than one estimate of the effect of an intervention on access or  graduation  rates.  In  order  to  report  only  the  most  comparable  estimates,  we  defined  four  main  rules  to  select  them.  First,  we  reported  the  effect  on  enrollment  rates  which  are  measured immediately after high school graduation or after participation in the program since  it  was  most  often  provided.    Conversely,  we  selected  the  longest  time‐frame  available  regarding graduation rates. Since this review focuses on how to improve graduation rates of  disadvantaged  students,  we  compare  estimates  that  evaluate  whether  students  ultimately  earned a degree in higher education. In addition, we only reported the estimates referring to  the  most  disadvantaged  participants.  For  example,  when  the  effect of  an  intervention  was  provided for participants with different income levels, we selected the lowest level.  Finally,  9    we  only  reported  estimates  related  to  enrollment  or  graduation  in  public  institutions,  if  a  distinction between public and private was made.   3.5 Analysis  We  decided  against  a  formal  meta‐analysis  that can  estimate  an  average effect size  of  the  interventions. There is a large diversity of studies involved, with different interventions and  different  estimation  strategies,  with  their  own  assumptions,  which  are  important  for  the  interpretation of the estimated effect. As a result, there are too few studies in each category  to do a meaningful formal meta‐analysis. Instead, we opt for a systematic review that presents  the selected findings and implications in a narrative form. We clustered the studies based on  the characteristics of the interventions and we provide all selected estimates and the details  of the different interventions in the Appendix.  We  also  compare  the  raw  unstandardized  estimated  effects  and  decided  not  to  calculate  standardized effect sizes. While acknowledging that standardized effect sizes would facilitate  the comparison of our estimates with external benchmarks, we argue that standardized effect  sizes are not absolutely necessary given the characteristics of our review and their calculation  would have some important limits in this case. We only included studies which provide the  effect of an intervention on the exact same outcomes, enrollment and graduation rates. Even  for a meta‐analysis, it is recognized that raw mean differences can be used directly when all  studies use the same outcome and report the effect a meaningful scale (Borenstein, 2009).  Second, among the 75 selected studies, only three reported standardized effect sizes and they  were already calculated with two different methods. For all the other studies, we would need  to use different methods to calculate them based on the information available in each study  and at the price of many assumptions. Given that all the selected studies focus on the same  meaningful outcomes and that we do not aim to obtain an average effect of the interventions,  we  thus  report  and  mainly  discuss  the  estimated  marginal  effect  of  the  intervention  in  percentage points. Still, we systematically report in the Appendix the baseline means, when  available. In addition, for the interventions where many studies are available, we provide a  graphical overview of the available evidence by plotting the selected estimated effects and  the calculated relative risks to make the comparisons across studies easier.     10    4. Outreach programs  We grouped outreach interventions in three types that may affect students differently. The  first group consists of low‐intensity interventions that address information barriers faced by  high school students. These interventions of short duration mainly deliver general information  on financial aid, college costs and returns to higher education or college application. A second  group of interventions is designed to complement information with personalized assistance  and aims to guide students during the steps of the enrollment procedures. These interventions  are more often spread over a longer period, provided by tutors who engage in a personalized  exchange with participants and often include proactive strategies to ensure that participants  engage  in  the  program.  Recently  though,  some  low‐cost  nudging  interventions  have  been  designed to provide guidance to students through automated procedures. The third group of  outreach programs offer academic tutoring during upper secondary education, in addition to  information and counseling. Lasting several years, these interventions include extensive after‐ school activities and aim to increase students’ academic readiness for higher education.   We found 28 studies which provide causal effects of the effect of outreach interventions on  access to higher education for disadvantaged students but only 4 which provide estimates on  graduation rates (Table 1). The lack of evidence on graduation may be consistent with the aim  of  outreach  interventions,  which  primarily  aim  to  facilitate  access  to  higher  education.  Nevertheless,  it  is  crucial  to  know  whether  disadvantaged  students  who  entered  higher  education after participating in an outreach program were able to eventually graduate and  this should clearly be addressed more often in the future. Finally, outreach interventions are  usually  evaluated  through  experimental  designs  and  have  been  tested  in  six  different  countries. However, we also note that the diversity of educational contexts is only found for  interventions providing additional information only. The large evidence on the interventions  classified as “information & support” comes exclusively from the United States and Canada,  and testing such interventions in other contexts would also be necessary in the future.       11    Table 1: Available evidence on the impact of outreach interventions     Access Graduation Number of studies by type of interventions        Not specified (Any outreach programme)  1  0  Information  8  0  Information & support  18  3  Information, support & tutoring  3  1  Total number of studies  28  4  Studies' characteristics         RCT design (in % of total studies)  82%  50%  Diversity of national contexts (nb of country)  6  2  National‐scale interventions (in % of total studies)  25%  25%  Single‐institution interventions (in % of total studies)  11%  0%  Source: Tables B.1, B.2, B.3, B.4, B.5 in Appendix.          4.1 Impact on access to higher education  Only one study provides a quasi‐experimental evaluation of outreach programs in general, not  limited  to  one  specific  intervention.  Domina  (2009)  uses  longitudinal  data  to  compare  the  efficiency of outreach programs and found an increase in enrollment (+5.5 p.p.) in any higher  education institution, but this was not statistically significant (Table B.1 in Appendix). Since no  information was available on the type of services offered, it is possible that different program  designs have very different impacts on college enrollment.   The evaluations of specific outreach interventions indeed suggest a great variety of effects on  enrollment,  depending  on  the  characteristics  of  interventions.  As  shown  by  figure  2,  interventions  providing  disadvantaged  students  with  additional  information  only  on  higher  education  seem  to  have  very  little  impact  on  access  patterns,  while  interventions  which  complemented information with assistance or individualized guidance on college or financial  aid applications seem to be more efficient. Among the 18 studies included, the range of the  estimated effects is wide, but most found a statistically positive effect on the enrollment rates  of disadvantaged students and more than half found  an increase in enrollment rates by at  least 10%.       12    Figure 2: Selected estimates for the impact of outreach on access to higher education       Note:  Refer  to  estimates  on  access  to  any  type  of  higher  institution,  whenever  available.  If  not  provided,  estimates on access to four‐year institutions or to university are used instead. See Appendix B for further details.    Whether they focus on financial aid information or costs and returns to higher education, most  of the interventions providing disadvantaged students with additional information had a very  small or null impact on enrollment rates of disadvantaged students (Table B.2 in Appendix B).  Interestingly, such interventions have been tested in very different contexts and consistently  brought  little  improvement  in  widening  access  to  higher  education  for  disadvantaged  students. In the U.S., providing information on aid eligibility and application in tax preparation  offices (Bettinger, Long, Oreopoulos, & Sanbonmatsu, 2012) or sending high school seniors  text  messages  on  the  financial  benefits  of  financial  aid  (Bird,  Castleman,  Goodman,  &  Lamberton,  2017)  did  not  increase  enrollment  of  disadvantaged  students.  In  Finland,  an  information session on returns to higher education did not have any impact on transition rates  of disadvantaged students (Kerr, Pekkarinen, Sarvimäki, & Uusitalo, 2014) similarly to what  was found in Colombia (Bonilla, Bottan, & Ham, 2017). In Chile, where students consulted web  pages on returns to higher education, there was also no impact on enrollment rates (Hastings,  Neilson, & Zimmerman, 2015). In the U.S., the inclusion of an online shopping sheet to provide  13    personalized  information  about  costs  and  loan  options,  had  even  a  negative  effect  on  the  enrollment  behaviors  of  low‐income  admitted  students,  although  this  effect  was  not  statistically significant (Rosinger, 2016). Even a more intensive intervention which provided  personalized information on the costs, benefits and chances of success in higher education  through three meetings did not improve access of disadvantaged students in Italy (Abbiati,  Argentin, Barone, & Schizzerotto, 2017).   Among the eight studies reviewed, only one found a large positive impact on enrollment rates.  Despite  a  design  very  similar  to  interventions  previously  mentioned,  Loyalka,  Song,  Wei,  Zhong, & Rozelle (2013) found that a one‐time presentation on cost and financial aid in poor  counties in China increased enrollment by 8 percentage points. Nevertheless, the authors note  that the information intervention did not have an impact on enrollment for lower SES students  (estimates were unfortunately not provided).   How should we interpret these findings? We formulate different hypotheses building on the  literature which has investigated information biases about higher education. First, it could be  that  beliefs  about  the  costs  or  returns  to  higher  education  are  “sufficiently”  biased  to  represent  a  barrier  for  disadvantaged  students  only  in  specific  national  or  educational  contexts. If so, information campaigns  can have an impact on access rates, but only if access  to information on financial aid and costs of higher education is extremely limited. The only  study  which  found  a  large  positive  impact  for such  intervention  took  place  in  China  where  students learn about financial aid packages only after being accepted to a higher education  institution. This lack of early information on financial aid may deter disadvantaged students  to even apply (Liu et al., 2011; Loyalka et al., 2013). In other contexts, information about costs,  returns or financial aid may be more widely accessible and there would be no need to address  this  issue.  It  is  interesting  to  see,  for  example,  that,  a  recent  intervention  in  the  U.S.  that  provided  semi‐personalized  information  about  returns  to  higher  education  to  high  school  students  (through  a  web  platform)  reported  major  difficulties  in  mobilizing  schools  and  students  to  participate.  In  three  years,  only  25  schools  out  of  300  agreed  to  join  the  experiment despite active outreach, and in the participating schools, students made very little  use  of  the  developed  tool.  As  noted  by  the  authors,  this  is  a  useful  finding  in  itself  which  suggests that there may be little demand for additional information, at least in this specific  context (Blagg, Chingos, Graves, & Nicotera, 2017).  14    Another  hypothesis  would  be  that  students’  beliefs  about  higher  education  do  not  automatically  impact  their  intention  to  attend  higher  education  and/or  their  behaviors  to  apply. If so, information interventions may be efficient in changing students’ beliefs but that  would not necessarily translate to intentions and/or behaviors. For example, in the U.S.,  Avery  and Kane (2004) found that there was only a weak connection between students’ estimations  of  net  returns  from  higher  education  and  plans  to  attend  college.  However,  there  is  also  evidence that information interventions are efficient in changing beliefs about cost or returns  from higher education and intentions to attend (Bleemer & Zafar, 2018; Oreopoulos & Dunn,  2012; Peter & Zambre, 2017). One study found that providing additional information about  grants did not change college intentions but did increase college application behaviors (Ehlert,  Finger,  Rusconi,  &  Solga,  2017).  Finally,  providing  general  information  about  a  prestigious  grant changed disadvantaged students’ knowledge but did not affect their propensity to apply  to it, unless general information was combined with a meaningful role model who could show  that someone with a similar background had been successful in obtaining such grants (Herber,  2018). These results call for further research on the relationship between beliefs, intentions  and behaviors regarding higher education. In addition, it is important to recall that, in many  educational systems, enrollment in higher education goes beyond the student’s own decision.  Not only do students need to apply but they also need to be selected by the tertiary institution  to  be  able  to  enroll.  Even  when  additional  information  increases  college  intentions  and  application behaviors, it may be that the lack of support during the application process hinders  the chances of disadvantaged students making successful applications.  Finally,  further  research  would  be  needed  to  disentangle  the  effect  of  information  interventions,  depending  on  the  type  of  information  provided.  Providing  additional  information on returns from higher education in the labor market, on available financial aid,  or  on  chances  of  success  may  impact  disadvantaged  students  very  differently.  And  the  connection between beliefs, intentions and behaviors may vary depending on the nature of  the information biases and updates. It is very interesting to see, for example, that providing  students with a personalized message about their chances of graduating in a chosen program  did not increase their actual enrollment if the message was positive, but led to a large decrease  (by 14 p.p.) in enrollment in this specific program if the assessment of the chances of success  was negative (Pistolesi, 2017). This result suggests that providing additional information on  15    the  odds  of  success  may  be  more  efficient  in  changing  behaviors  when  it  is  negative  (thus  leading  to  a  decrease  in  enrollment)  but  has  little  impact  when  it  is  positive.  It  would  be  interesting  to  investigate  whether  this  would  also  be  the  case  for  the  other  types  of  information relevant for higher education decision‐making.   In contrast, the effect of the interventions which complemented information with assistance  or  individualized  guidance  on  college  or  financial  aid  application  were  found  to  increase  enrollment rates of disadvantaged students in most cases (Figure 2 and Table B.3 in Appendix  B).  Typically,  the  “information  &  guidance”  outreach  interventions  provide  personalized  advice and support on higher education applications through counselors. In some cases, the  counseling  program  can  run  over  a  few  years  in  high  school:  An  early  example  of  such  a  program  is  the  Talent  Search  program,  a  large‐scale  program  in  the  U.S.,  which  provides  information  and  support  to  disadvantaged  students  from  ninth  grade  onwards.  Using  propensity score matching, Constantine, Seftor, Martin, Silva, & Myers (2006)  estimate that  initial enrollment of Talent Search participants in a postsecondary institution was higher by  18,  4,  and  15  percentage  points,  respectively,  in  Texas,  Indiana,  and  Florida.  Similarly,  In  Canada, the “Explore Your Horizons project” provided 40 hours of after‐school activities over  three  years  in  high  school  (Ford  et  al.,  2012).  This  included  guidance  for  disadvantaged  students  and  their  parents.  The  intervention  was  successful  in  increasing  participation  of  disadvantaged students in higher education, by around 10 percentage points.   Six interventions were designed to provide counseling to disadvantaged students during the  senior year in high school only. In the US, Avery (2010) analyzed an individualized counseling  intervention of 10 hours over the school year for high‐achieving disadvantaged high school  seniors.  The  intervention  led  to  an  increase  of  8  p.p.  in  access  to  most  selective  higher  education institutions, although this large increase was not significant due to the small sample  size of this pilot study (Avery, 2010). Similarly, counseling in the senior year of high school was  found to increase the probability of enrolling in higher education for disadvantaged students  by 3 p.p. (Stephan & Rosenbaum, 2013), and up to 7 p.p. (Barr & Castleman, 2017). It also  showed  to  be  efficient  in  diverting  disadvantaged  students  from  short  programs  and  encouraging  them  to  enroll  in  four‐year  institutions  (Bos,  Berman,  Kane,  &  Tseng,  2012;  Castleman & Goodman, 2014). Finally, being enrolled in a school which offered a “GO center”  i.e. a dedicated classroom for the college application process with a full‐time counselor and  16    active outreach run by selected student peers, already increased enrollment of low‐income  students by 3.5 p.p. which should be taken as a lower bound estimate as it does not focus on  students who actually took part in the program (Cunha, Miller, & Weisburst, 2018).  There  are  several  ways  in  which  these  –  moderately  intense  –  interventions  may  have  influenced  disadvantaged  students’  enrollment  behaviors.  While  a  longer  exposition  to  information on higher education may be beneficial, these interventions also help students to  navigate among college choices. Moreover, they reduce the complexity of application tasks  which seems to be a crucial step to induce changes in application behaviors as suggested by  the behavioral theories described earlier. Additionally, it seems that early familiarization with  higher education options may be a powerful way to raise students’ educational aspirations  which in turn can raise students’ performance in high school. Indeed both the Talent Search  and Explore Your Horizons, which were spread over four and three years respectively, have  raised high school completion among disadvantaged students although they did not include  academic tutoring (Constantine et al., 2006; Ford et al., 2012). These results thus draw our  attention  to  the  role  of  anticipatory  decisions  (Erikson,  Goldthorpe,  Jackson,  Yaish,  &  Cox,  2005) on academic performance.   Although  they  are  not  likely  to  increase  educational  aspirations,  short‐term  targeted  counseling interventions to support students in the application and enrollment period also  appear to be efficient in raising access rates of disadvantaged students.  Four interventions  specifically  focused  on  students  after  upper‐secondary  graduation  and  provided  proactive  counseling  during  the  summer  months  to  low‐income  students.  The  results  highlight  the  importance of engaging students in available counseling activities as a key factor to improve  students’ outcomes. Three of these interventions had very consistent and substantial impact  (between 8 and 14  p.p.) on  immediate enrollment  and enrollment  in four‐year institutions  (Castleman, Arnold, & Wartman, 2012; Castleman, Owen & Page, 2015, Castleman, Page, &  Schooley, 2014).  In these cases, counseling was available for students in the control group  but without any proactive outreach, while counselors used many means to contact students  in the treatment group. The large gap in enrollment between the two groups thus indicates  that availability of information or counseling is not sufficient and that counselors actively need  to  reach  out  to  potential  students.  This  is  achieved  using  small  financial  incentives  for  participation in another one‐month counseling intervention which also brought about large  17    increases (17 to 20 p.p.) in enrollment rates of non‐white and low‐income students (Carrell &  Sacerdote,  2013).  Only  one  summer  counseling  intervention  did  not  significantly  increase  enrollment rates of disadvantaged students in higher education (Castleman & Page, 2015).  But  even  this  intervention  led  to  an  increase  of  almost  5  p.p.  in  enrollment  in  four‐year  institutions  and  led  to  an  increase  in  enrollment  rates  of  12  p.p.  for  students  with  less‐ developed college plans. Thus, it may also be that the efficiency of such interventions depends  largely on their ability to target students who are the most at risk to fail to carry out their  matriculation after their high school graduation.   But is it possible to efficiently guide students through the application process with no contact  with counselors? Five interventions tested low‐cost interventions offering guidance through  automated or semi‐automated procedures and results are promising that these interventions  can,  to  some  extent,  improve  access  outcomes  of  disadvantaged  students.  In  the  U.S.,  Bettinger  et  al.    (2012)  tested  a  streamlined  personal  assistance  for  the  FAFSA  application  which increased college enrollment of low‐income high school students by 8 p.p. In addition,  Hoxby  &  Turner  (2013)  sent  high‐achieving  low‐income  students  semi‐customized  college  advising and college application fee waivers, by regular mail, to simplify the paperwork tasks  to obtain application fee waivers.  They concluded that treated students enrolled significantly  more in institutions matching their ability: an increase of 5 p.p., which amounted to a 20%  increase compared to the mean of the control group. With intervention costs amounting only  to  $6  per  student,  this  type  of  intervention  is  extremely  promising.  The  outcomes  of  interventions  that  provide  personalized  information  on  the  steps  that  need  to  be  taken  to  enroll  (without  the  simplification  component)  are  somewhat  smaller  but  still  lead  to  improvement in enrollment behaviors with minimal intervention costs. For example, sending  text messages to remind high school graduates of the tasks required for enrollment during the  summer had a small impact on two‐year institution enrollment (+3 p.p.) but not on overall  access  to  higher  education  (Castleman  &  Page,  2015).  However,  text  messaging  increased  enrollment of low‐income students by almost 6 p.p. and of first‐generation students by almost  5 p.p. (Castleman & Page, 2017). Finally, a large‐scale nudging experiment which sent only a  few emails and text messages to disadvantaged college‐intending high school seniors to guide  them  step‐by‐step  through  the  completion  of  the  FASFA  application  was associated  with  a  small but statistically significant increase in enrollment (+1.7 p.p.) (Bird et al., 2017). In this  18    study,  the  control  group  was  receiving  the  same  number  of  messages  but  with  general  information about financial aid, so the positive impact of the texts which included “planning  prompts”  confirms  the  importance  of  complementing  information  with  concrete  logistics  guidance to efficiently increase access to higher education.   These  results  are  encouraging  but,  as  mentioned  earlier,  the  evidence  on  “information  &  guidance”  outreach  interventions  come  exclusively  from  North‐America  and  similar  interventions  should  be  tested  in  other  contexts  to  confirm  the  efficiency  of  counseling  or  nudging outreach interventions.    Finally,  there  are  fewer  evaluations  of  intensive  outreach  programs  that  offer  intensive  academic  tutoring  during  upper  secondary  education.  These  interventions  not  only  try  to  address  information  gaps  but  also  the  lack  of  academic  preparation  of  disadvantaged  students. Although limited, the current evidence suggests that these intensive interventions  may  have  little  impact  on  overall  access  to  higher  education  (Table  B.4  in  Appendix  B).   Randomized  experiments  to  evaluate  the  “Upward  Bound”  program  and  the  “College  Possible” program, which both offer academic support in upper secondary school, did not find  a significant impact on access to higher education (Avery, 2013; Myers, Olsen, Seftor, Young,  & Tuttle, 2004; Seftor, Mamun, & Schirm, 2009). One possible explanation is put forward by  Myers et al. (2004) who suggest that the absence of impact on postsecondary enrollment is  the consequence of the large number of students who do not complete the program. Since  these interventions last over many years and include many hours of out‐of‐school activities,  many pupils usually drop out before completing them.   4.2 Impact on graduation  Table B.5 in Appendix presents the estimates of outreach programs on graduation rates but,  as  mentioned  earlier,  we  found  few  (quasi‐)experimental  studies,  only  four  studies,  which  have evaluated the impact of outreach programs on graduation rates of participants.  So far, only one study has been able to identify a positive impact of an outreach program on  graduation  rates.  Constantine  et  al.  (2006)  identified  a  substantial  increase  of  5  p.p.  in  completion  rates  at  2‐year  institutions  for  participants  of  the  “Talent  Search”  program  in  Florida.  Conversely,  the  “Upward  Bound”  program  did  not  have  any  impact  on  graduation  rates, which is consistent with the almost negligible impact found for enrollment (Seftor et al.,  19    2009). Similarly, and despite leading to a large increase in enrollment rates, the “Explore Your  Horizons”  intervention  in  Canada  failed  to  find  an  effect  on  graduation  rates.  Since  the  increase in enrollment rates was exclusively driven by enrollment in  university programs and  graduation rates measured only four years after expected high school graduation, later data  may  be  necessary  to  identify  an  increase  in  graduation  rates  (Ford,  Grekou,  Kwakye,  &  Nicholson, 2014). However, with a long‐term evaluation, Cunha et al. (2018) did not find that  the increase in enrollment for low‐income students translated in an increase in graduation by  eight years: being enrolled in a school offering outreach (GO center) seems to induce enrolling  students who are also more at risk of dropping out once in college. These results suggest that  the long‐term benefits of outreach interventions may be limited if students are not further  supported  once  in college (Cunha et al., 2018) and that more attention should be given to  graduation outcomes in evaluations of outreach programs.   5. Financial support  As financial aid has diversified over the last two decades, we may expect some heterogeneity  in their effects and separately discuss the impact of universal grants (available for all students),  need‐based aid (which uses parental financial conditions as the main eligibility criteria), merit‐ based  aid  (which  requires  high  academic  performance,  usually  at  high  school  graduation),  performance‐based aid (which is contingent on staying enrolled and making passing grades in  higher education),  loans and tax incentives.   Table 2: Available evidence on the impact of financial aid     Access Graduation Number of studies by type of interventions        Universal grants  1  1  Need‐based grants  14  12  Merit‐based grants  6  4  Performance‐based grants  4  2  Loans  2  3  Tax‐credit  2  1  Total number of studies  28  22  Studies' characteristics         RCT design (in % of total studies)  18%  23%  Diversity of national contexts (nb of country)  8  3  National‐scale interventions (in % of total studies)  43%  45%  Single‐institution interventions (in % of total studies)  7%  9%  Source: Tables C1‐C12 in Appendix C.        20    Table 2 shows that most of the available evidence deals with need‐based grants. Contrary to  outreach interventions, we could find many studies providing estimates of the impact of aid  on  graduation  outcomes.  Around  half  of  the  studies  evaluated  a  national  aid  scheme,  and  there  is  some  diversity  in  the  educational  contexts  where  the  effect  of  financial  aid  was  evaluated.  However,  the  available  causal  evidence  on  the  effect  of  some  aid  schemes  for  disadvantaged students remains extremely limited, most notably for universal grants, loans  and tax‐credits.   5.1 Effects on enrollment  One study provided causal estimates of the effect of universal grants or price reduction on the  access rates of disadvantaged students, using a difference‐in‐differences design (Table C.1 in  Appendix  C).  Large  price  reductions  in  community  colleges,  which  amount  to  at  least  60%  reduction  of  the  tuition  fees,  based  on  residency  was  found  to  successfully  increase  disadvantaged students’ enrollment in these institutions but to divert students from four‐year  institutions    (Denning,  2017).  More  quasi‐(experimental)  evidence  is  obviously  needed  to  conclude  whether  these  policies  participate  in  reducing  inequalities  in  higher  education.  It  may be that universal financial grants, which normally only include a basic application process,  are more efficient in reaching all disadvantaged students than specifically targeted programs  which  require  complex  application  forms.  Conversely,  it  may  be  that  socially  advantaged  students  react  more  to  such  opportunity  and  remain  the  primary  beneficiaries  of  these  policies.   More  studies  are  available  regarding  the  effect  of  grants  which  defined  more  stringent  eligibility  rules.  Figure  3  displays  the  collected  estimates  for  need‐based  and  merit‐based  grants.  Results  on  the  effect  of  need‐based  grants  are  mixed.  Many  studies  find  a  small  substantive  effect,  but  which  fails  to  reach  statistical  significance.  A  few  studies,  however,  found that need‐based grants had a large effect on access rates of disadvantaged students.  Results on merit‐based grants are also mixed but with a different pattern: some concluded  that merit‐based grants actually decreased enrollment rates of disadvantaged students and  only a third of the available studies found that such grants had a positive statistically significant  effect on access to higher education for disadvantaged students. Since there is such diversity  21    in these findings, it is necessary to discuss the studies and the design of the aid schemes in  more detail.   Figure 3: Selected estimates for the impact of financial aid on access to higher education     Note:  Refers  to  estimates  on  access  to  any  type  of  higher  institution,  whenever  available.  If  not  provided,  estimates on access to four‐year institutions or to university are used instead. See Appendix C for further details.  The evidence on  need‐based aid is mixed. While most studies find a small substantive effect  on access to higher education (Table C.2 in Appendix), only a third of the selected estimates  are statistically significant. Among the 14 studies reviewed, only four interventions found a  statistically significant effect larger than 5 percentage points. However, the grant programs  evaluated differ greatly from one another and it is possible to identify some of the features  that  seem  to  be  associated  with  larger  impacts  on  access  rates  to  higher  education.  Most  notably the amount and the timing of the grant seem to be central features in the efficiency  of need‐based financial aid.   For example, in the U.S., the Pell grant, which can be quite small, was not associated with any  increase in enrollment (Denning, Marx, & Turner, 2017; Kane, 1995; Rubin, 2011). Conversely,  studies analyzing grants that supplement the Pell grant are more likely to find positive effects  22    of aid, supporting the hypothesis that the size of aid matters. In a randomized controlled trial  in  the  United  States  (California),  Richburg‐Hayes  et  al  (2015)  provided  a  one‐time  $1,000  additional subsidy for enrolling in higher education which increased enrollment at any college  by 3.5 percentage points (although it was not statistically significant), and by 5 percentage  points  for  two‐year  colleges.  Using  a  regression  discontinuity  design,  Castleman  and  Long  (2013) found that an additional yearly renewable grant of  $1,300  (in 2000$) had a positive  (+3  p.p.),  but  statistically  non‐significant  effect  on  higher  education  enrollment  which  was  mainly driven by an increase in enrollment in four‐year institutions (statistically significant at  10%).  Bettinger (2015) also found  a small but statistically significant response to the Ohio  College  Opportunity  Grant:  those  who  received  around  $750  more  grant  aid  because  of  a  reform of the aid scheme were 1.5 percentage points more likely to enroll at public, four‐year  colleges.  Linsenmeier  et  al  (2006)  found  that  one  university  grant,  that  replaced  a  loan  (increasing total grant aid by an average of just over $3,000), had a small impact on attendance  among admitted students (yield rate) for  low‐income students (2 p.p.) but was able to raise  attendance by close to 9 p.p. for low‐income minority students, an estimate almost significant  at the 10% level.  Finally, interventions that offer very generous subsidies were found to have large effects on  enrollment. Dynarski (2003) found that the elimination of the Social Security Benefits program  that targeted children of deceased, disabled or retired parents decreased enrollment by 22  percentage points. Under this program, students received an average subsidy of $6,700 per  year (in 2000$), at a time when tuition averaged around $1,900 per year at public universities.  Similarly,  the  temporary  ban  on  all  types  of  federal  financial  aid,  for  students  with  drug  convictions,  decreased  immediate  college  attendance  by  22  p.p.  although  this  effect  was  mainly  the  consequence  of  delayed  enrollment  during  the  time  of  the  ban  (Lovenheim  &  Owens, 2014).  Evidence from Europe seems to confirm that the effect of need‐based aid is only identifiable  when  the  amount  of  aid  is  large  enough.  In  France,  the  main  need‐based  grant  scheme  contains different levels of aid. While a fee‐waiver (which amounted to 174 euros) had small  positive  (statistically  non‐significant)  effects,  an  additional  €1,500  per  year  increased  enrollments by almost 3 percentage points, and by almost 5 p.p. for enrollment in  the first  year  of  undergraduate  programs  (Fack  &  Grenet,  2015).  In  the  United  Kingdom,  the  23    implementation of need‐based grants of £960 (2006 prices), on average, was associated with  an increase in access to higher education of almost 4 p.p. among low‐income youths (Dearden,  Fitzsimons, & Wyness, 2014). In contrast, in Germany, a 10% increase in the federal students’  financial assistance scheme led to a small but not significant increase in enrollment rates of  low‐income students (Baumgartner & Steiner, 2006). The authors argue that this may have to  do with the small sample size, but it is also possible that the increase in aid, which amounted  to €45 per month on average, was too small to lead to any sizable increase in enrollment rates,  in line with the findings from the studies discussed above.   Together  with  the  amount,  the  timing  of  the  grants  may  also  be  important  for  efficiently  supporting disadvantaged students. In New Brunswick in Canada, Ford et al. (2014) deposited  a maximum of CAN$8,000 in high school students’ saving accounts. The amount was deposited  in tenth grade, giving students enough time to prepare their college applications. Importantly,  students  were  only  able  to  access  the  grants  for  two  years  while  in  college.  Enrollment  in  postsecondary education increased dramatically, by almost 11 percentage points, although  this was driven exclusively by an increase in short program enrollment. Another example of  financial aid with early commitment was tested in Italy (Azzolini, Martini, Romano, & Vergolini,  2018). Disadvantaged students were invited to save money for their education during their  last  two  years  of  high  school  and  their  deposits  on  this  dedicated  saving  account  were  matched at a rate of 4 to 1. The money could then only be used for educational expenses and  this led to a large increase in enrollment of almost 9 p.p. Not only were students aware of the  amount of money they had for higher education studies before the end of secondary school,  but  students  and  families  were  directly  involved  in  anticipating  and  saving  for  educational  expenses, which may be another promising way to increase educational aspirations for higher  education (Azzolini et al., 2018).  The causal evidence on merit‐based aid suggest that these types of grants can have negative  effects for disadvantaged students, and only have a positive effect when they are designed to  guarantee  that  disadvantaged  students  have  access  to  them  (Table  C.3  in  Appendix  C).  Eligibility for merit‐based aid is defined in reference to the academic ability of the students,  with  criteria  setting  minimum  high  school  grades  or  performance  in  specific  standardized  tests.  The rationale for this form of aid is that it may incentivize student performance in high  school (thus increasing academic preparation for higher education), while encouraging good  24    performers to enroll in higher education. However, since high performers are typically from  privileged  backgrounds,  it  is  possible  that  these  kinds  of  programs  are  not  accessible  to  students from disadvantaged backgrounds. If this is so, this form of aid may reward those who  would anyway enroll in college, or even increase inequality across social groups. On the other  hand, some of the merit‐based grants are made accessible only to disadvantaged students by  including  a  need‐based  eligibility  criterion  and  may  be  able  to  improve  access  to  higher  education for this group.  With one exception, merit‐based grants that did not have a need‐based eligibility criterion  often seemed to have either increased inequalities or failed to trigger any improvement for  disadvantaged students (Bruce & Carruthers, 2014; Dynarski, 2000; Sjoquist & Winters, 2015).  Only Cohodes and Goodman (2014) found a positive effect of a merit‐based grant without a  need‐based eligibility criterion. The Adams scholarship in Massachusetts added between $900  and $1,700 in annual aid to reduce tuition costs for those who score highly on the state‐wide  examinations in tenth grade and without any need‐based eligibility component. Enrollment in  four‐year institutions increased by more than 6 percentage points among non‐White students,  while  it  went  up  by  almost  4  percentage  points among  low‐income  groups.  The difference  with the negative effects identified by the previous studies may be interpreted in light of the  specific design of the Adam scholarship: the initial idea was to provide a grant to students  whose  score  would  place  them  in  the  top  25  percent  of  students  state‐wide.  However,  “Concerned  that  […]  statewide  standard  would  assign  scholarships  largely  to  students  in  wealthy,  high‐performing  school  districts”,  the  state  decided  that  a  student’s  total  score  would  need  to  fall  in  the  top  25  percent  of  scores  in  his  or  her  school  district  (Cohodes  &  Goodman, 2014). Thus, although there was no need‐based criterion for eligibility, the grant  scheme was designed to guarantee that disadvantaged students would benefit from it.   Regarding  merit‐based  grants  which  are  targeted  to  lower‐income  students,  Kane  (2003)  found  that  a  merit‐aid  program  in  California  with  a  need‐based  component  increased  enrollment  by  4  percentage  points  immediately  below  the  income  eligibility  threshold.  Similarly, Vergolini, Zanini and Bazoli (2014) found that an Italian merit grant, available only  for high performers from low‐income families increased enrollments by 6.5 percentage points,  although this finding was not statistically significant.   25    While there is limited evidence on the effect of performance‐based scholarships, which make  grant payment conditional on minimum academic achievement in higher education, the few  available studies find promising effects. Nevertheless, it should be noted that these types of  grants often focus on students who have already carried out the first enrollment steps in a  specific institution and provide them incentives to register for a minimum number of courses.   Of the four available studies, three (Barrow, Richburg‐Hayes, Rouse, & Brock, 2014; Jackson,  2010; Richburg‐Hayes et al., 2015)  identified a positive significant effect on enrollment (Table  C.4 in Appendix C) and the only study that did not show any increase was targeting freshmen  students  who  already  had  a  registration  rate  of  almost  100%  in  the  control  group  (Binder,  Krause, Miller, & Cerna, 2015).   Finally, the evidence on loans suggests that these forms of aid may be efficient in improving  access rates of disadvantaged students but more experimental research is necessary (Table  C.5 in Appendix C). In Chile, the national loan program was found to increase enrollment by  20  percentage  points  for  college‐intending  students  in  the  lowest‐income  quintile  (Solis,  2013).  Similarly,  short‐term  loans  covering  tuition  fees  in  South  African  public  universities  were estimated to double enrollment rates of admitted disadvantaged students  (Gurgand,  Lorenceau, & Mélonio, 2011).  In contrast, the available evidence on  tax incentives does not  suggest any positive impact for disadvantaged groups’ access to higher education (Table C.6  in  Appendix  C)  as  two  studies  in  the  U.S.  fail  to  identify  an  effect  on  enrollment  for  disadvantaged students (Bulman & Hoxby, 2015; LaLumia, 2012)  As these tax incentives only  provide income relief about 10.5 months after enrollment, these may not be very effective in  addressing unmet financial need. Moreover, these tax incentives tend to benefit middle‐ and  upper‐income families, as lowest‐income families do not pay taxes and are thus not eligible  for them.   5.2 Effects on graduation  The literature on the effects of financial aid on higher education graduation is still quite recent  but has lately received growing attention.  Regarding an example of a “universal” grant, price  reduction  in  community  colleges,  based  on  residency,  led  to  a  small  increase  in  associate  degree graduation for black students but not for low‐income students, for whom the increase  in enrollment did not translate into more graduates (Denning, 2017).   26    The  available  evidence  further  suggests  that  need‐based  grants  are  often  efficient  in  supporting the graduation of disadvantaged students (Table C.8 in Appendix C). Alon (2011)  found that each additional $100 of Pell grant received in the first year by students coming  from  the  poorest  families  increases  degree  completion  by  0.6  percentage  points,  which  is  statistically  significant.  Similarly  an  additional  $1,000  in  annual  grant  aid  was  found  to  significantly  increase  graduation  rates  of  minority  students  enrolled  in  private  and  most  selective universities (Alon, 2007) and to increase graduation from bachelor’s degrees for the  lowest‐income students by more than 5 p.p. (Denning et al., 2017). Lovenheim and Owens  (2014) also found that convicted drug offenders were 7 percentage points less likely to earn a  bachelor’s  degree  when  they  became  ineligible  for  federal  aid,  although  this  was  not  significant.  Only  Denning  (2018)  found  an  effect  of  less  than  1  p.p.  on  completion  of  a  bachelor’s degree following an increase in the Pell grant but this was estimated on students  already in their last year of a  bachelor’s program and the larger financial aid did increase on‐ time graduation by almost 3 p.p. (Denning, 2018).   Regarding the grants supplementing federal aid in the U.S., Castleman and Long (2013) found  that the Florida FSAG increased graduation from four‐year colleges by 5 percentage points.  This  is  a  substantial  effect,  as  it  represents  an  increase  of  21%  over  the  sample  mean  probability to graduate. The Wisconsin Scholars Grant was  also found to largely increase on‐ time bachelor’s graduation (Goldrick‐Rab et al., 2016) but not completion of associate degrees  (Anderson & Goldrick‐Rab, 2016). An institutional grant meant to cover 100% of unmet need  had a small but non‐significant effect on on‐time graduation (+2.2 p.p.; Clotfelter, Hemelt, &  Ladd, 2018). Finally, Turner and Bound (2003) estimated that the GI‐Bill, which provided up to  $500 in tuition expenses and up to $120 per month in living costs to returning veterans from  WWII, increased college degree completion of black students by almost 3 percentage points,  although this effect was not statistically significant. The authors argue that the absence of a  large  effect  is  due  to  higher  education  supply  problems  in  the  South  of  the  United  States,  where school segregation was still a major issue. Indeed, they identified a larger, statistically  significant, effect of almost 6 p.p. for Blacks in the northern states.  In Canada, Ford et al (2014) found that the two‐year grant provided with early commitment  during high school increased any degree completion by 9 percentage points, which represents  a  70%  increase  from  the  baseline.  In  France,  Fack  &  Grenet  (2015)  found  that  receiving  a  27    €1,500 grant, on top of a fee‐waiver increases undergraduate degree completion by almost 3  percentage points, for those on the threshold of grant eligibility in their final year. While these  effects are slightly smaller than the enrollment effect cited above, they are still sizeable, as  this  aid  allowed  around  half  the  students  who  it  incentivized  to  enroll  to  complete  their  undergraduate degrees.   The evidence of merit‐based financial aid on degree completion is limited but current findings  are not encouraging (Table C.9 in Appendix C). Among the four reviewed studies, none was  able to identify an improvement in graduation rates for disadvantaged students (Carruthers  & Özek, 2016; Cohodes & Goodman, 2014; Sjoquist & Winters, 2015; Welch, 2014). All the  selected estimates on graduation from any degree or bachelor’s degree range from ‐4 to +0.2  percentage points and none are significant.   We would expect the effects of performance‐based financial aid on degree completion to be  larger on completion as these forms of grants are specifically designed to increase persistence  and graduation.  Performance‐based aid provides short‐term monetary incentives to maintain  a  minimum  GPA  allowing  students  to  graduate  within  a  reasonable  period  of  time.  The  evidence  on  disadvantaged  students’  graduation  or  completion  rates  is  however  still  very  limited  (Table  C.10  in  Appendix  C).  Binder  et  al.  (2015)  find  that  the  VISTA  program  for  disadvantaged students at the University of New Mexico increased degree completion within  five  years  by  4.5  p.p.,  which  was  statistically significant at  the  11%  level.  Mayer,  Patel  and  Gutierrez (2015) found that a performance‐based grant in three community colleges, raised  degree  attainment  within  two  and  within  three  years,  by  3  to  4  percentage  points.  Nevertheless, within four years, the program had increased completion by less than 2 p.p. and  was  no  longer  statistically  significant.  In  other  words,  the  program  accelerated  degree  completion,  thus  increasing  efficiency,  but  did  not  increase  overall  graduation  in  the  long  term.   Finally,  none  of  the  three studies which  provide  causal estimates  of  the  effect  of  loans  on  graduation  identified  a  statistically  significant  impact  (Alon,  2007;  Dunlop,  2013).  Only  (Wiederspan, 2016) identified a  large effect (+ 20) of receiving federal loans on graduation  from associate degrees but this was not statistically significant. We could identify only one  study assessing the effects of tax incentives on degree completion for disadvantaged students  (Elsayed, 2016) and more experimental research is obviously needed to draw any conclusions.    28      6. Mixed interventions combining financial aid and outreach  This section presents the results from studies evaluating mixed interventions that combine  outreach with financial aid. While these studies make it difficult to assess the causal effect of  a  specific  component,  they  do  allow  us  to  assess  the  effectiveness  of  a  package  of  interventions. Table 3 provides the overview of the available evidence on these interventions.  The causal evidence is still limited but covers equally access and graduation outcomes. Around  half of the available evidence comes from randomized experiments. However, we could only  find evidence from the United‐States and Canada for these types of interventions and this is  clearly one of the main limits of this literature.  Table 3: Available evidence on the impact of interventions combining outreach and financial  aid     Access Graduation Total number of studies  7  6  Studies' characteristics         RCT design (in % of total studies)  43%  50%  Diversity of national contexts (nb of country)  2  2  National‐scale interventions (in % of total studies)  0%  0%  Single‐institution interventions (in % of total studies)  14%  33%  Source: Tables D.1‐D.2 in Appendix D.    6.1 Effects on enrollment  The evidence is still limited but mixed interventions seem efficient in raising enrolment. Six  out of the seven available studies found a statistically significant positive impact for at least  one disadvantaged group. And when a positive impact was identified, effect sizes are generally  large compared to outreach or aid estimates.  The  Quantum  Opportunities  Program  (QOP)  was  one  of  the  earlier  experiments  from  the  1990s and included education (tutoring, computer‐based instruction), development activities  and community service to improve the living conditions in the community. It targeted inner‐ city low‐income youth from ninth grade through to high school. Students received a small cash  incentive to engage actively in these activities, as well as bonuses when major segments were  completed. Students received over $1,000 on average, and all funding was deposited in a fund  29    that they could access while in postsecondary education. An initial evaluation found that QOP  had a dramatic effect and increased postsecondary enrollment by 26 percentage points (Hahn,  Leavitt, & Aaron, 1994) but it should be noted that the sample of this experiment was small  (N=158  students).  A  more  recent  evaluation  with  a  larger  sample  found  smaller  but  still  sizeable effects: By the time that youth were in their mid‐twenties, participants were around  7 p.p. more likely to have ever attended postsecondary education than those in the control  group (Rodríguez‐Planas, 2012).   The  other  randomized  experiment  tested  in  Canada  a  combination  of  outreach  and  need‐ based aid (Ford et al., 2014). Students were eligible to receive 40 hours of counseling during  high school, and a maximum of CAN$8,000 in need‐based aid, deposited during high school  and  payed  while  in  college,  over  two  years.  The  impact  was  substantial  as  it  increased  enrollment in higher education by more than 10 p.p. Interestingly, this study also tested the  effect of each component of the intervention individually allowing us to compare the effect  sizes of the mixed intervention with its single components: the estimated impact on access to  higher education for the mixed intervention is not larger than the impacts of the individual  components of the intervention (see earlier in outreach and need‐based grants). However,  the combination of the interventions also increased attendance at university by almost 7 p.p.  while financial aid alone only had an impact on enrolment in short programs (Ford et al., 2014).   The  Pathways  to  Education  program  (Oreopoulos,  Brown,  &  Lavecchia,  2014)  provided  an  intensive  multifaceted  support  to  pupils  from  ninth  grade  through  high  school  in  urban  settings  in  Canada.  Participants  received  counseling,  free  daily  evening  tutoring  and  group  mentoring  activities.  Students  also  received  financial  support  throughout  the  program,  including transportation, school supplies, and a financial award of CAN$1,000 at the end of  each year of program participation. Financial support could reach a maximum of CAN$4,000  and could be used only to pay for postsecondary education expenses. At the first site where  the program was tested, the program had dramatic effects on postsecondary attendance as  program  youths  were  19  percentage  points  more  likely  to  enroll  in  any  postsecondary  education. At the second site where the program was tested, however, the results were much  more modest as the increase in postsecondary enrollment was 4 percentage points, which  was  not  statistically  significant,  although  there  was  an  increase  in  application  rates  (Oreopoulos et al., 2014).   30    All these interventions reached disadvantaged students early, in ninth or tenth grade of high  school but one intervention starting only in the senior year of high school was also efficient in  raising access rates of disadvantaged students. The Knox Achieves program which provided  outreach  and  financial  aid  for  making  an  immediate  transition  to  community  colleges  increased enrollment by more than 25 p.p. in these institutions without diverting students  from universities (Carruthers & Fox, 2016).   Only two studies (Andrews, Imberman, & Lovenheim, 2016; Page, Castleman, & Sahadewo,  2016)  did  not  identify  large  increase  in  enrollment  of  disadvantaged  students  with  interventions  combining  outreach  and  generous  financial  aid.  Interestingly,  both  were  focusing  on  high‐achieving  disadvantaged  students  only.  As  already  mentioned  when  discussing  merit‐based  aid,  high‐performing  and  motivated  disadvantaged  students  can  be  expected to enroll in higher education in any case. Thus, it is less likely that such interventions  bring large improvements for this specific population.   6.2 Effects on graduation  The  available  findings  regarding  interventions  that  combine  outreach  and  financial  aid  on  graduation  rates  of  disadvantaged  students  is  still  insufficient  but  suggests  that  these  interventions  can  have  positive  effects  on  graduation  rates  but  that  their  efficiency  is  not  systematic. Of the six studies selected, three found a large positive effect on graduation rates.  Two found smaller effects (less than 5 percentage points) and one did not find any positive  effect on graduation rates of disadvantaged students.   The Quantum Opportunities Program did not affect graduation rates for bachelor’s degrees  or associate degrees. Nevertheless, youths in the program were 7 p.p. more likely to complete  two years of college (Rodríguez‐Planas, 2012). The mixed interventions implemented by two  flagship  public  universities  in  Texas  also  brought  very  limited  improvements  in  degree  outcomes of the treated students (+1.5 p.p. increase in one case and a nil effect in the other)   but these interventions already had only a limited impact in enrollment rates in these specific  universities (Andrews et al., 2016).  Conversely, Ford et al (2014) found an increase in completion by 8 p.p. in their evaluation of  learning  accounts  and  Explore  Your  Horizons.  This  is  broadly  in  line  with  the  effect  of  the  financial  aid  alone  discussed  above.  The  Dell  program,  focusing  on  high‐performing  31    disadvantaged students, was also able to support bachelor’s graduation which was raised by  19  p.p.,  despite  its  very  small  impact  on  enrollment  (Page  et  al.,  2016).  Comprehensive  intervention implemented after enrollment in higher education may also be successful. The  ASAP program targeted disadvantaged students at three community colleges in New York. In  return  for  full‐time  enrollment,  the  program  provided  students  with  free  tuition  and  free  public  transport.  Students  also  received  a  dedicated  advisor  and  academic  tutoring.  The  participants were estimated to be 18 p.p. more likely to graduate by three years, effectively  doubling  graduation  rates  (Scrivener  et  al.,  2015).  Similarly,  combining  a  need‐based  grant  with  mentoring  and career guidance  in  one  university  raised  completion  rates  by  almost  5  percentage  points,  although  this  was  not  significant  through  the  (preferred)  regression  discontinuity estimating strategy (Clotfelter et al., 2018).  7. Conclusion  The  results  of  the  experimental  or  quasi‐experimental  literature  discussed  in  this  paper  provide  an  overview  of  the  causal  effects  of  the  most  common  interventions  or  policies  implemented to raise higher education outcomes of disadvantaged students. We were able  to identify some promising ways to reduce inequalities in higher education, even though many  interventions failed to find an effect.   Outreach interventions targeted at students in high school or recent graduates seem to be a  relatively cost‐effective tool to address inequalities in access to higher education, as long as  the  interventions  go  beyond  providing  general  information  about  higher  education.  Substantial improvements have been identified when disadvantaged students were offered  personalized  counseling  activities  or  simplification  of  application  tasks,  especially  when  counselors  actively  reach  out  to  targeted  students  to  ensure  their  participation.  However,  neither interventions which only provide additional information nor those including intensive  academic  tutoring  seem  to  efficiently  raise  higher  education  outcomes  of  disadvantaged  students.  Financial  aid  is  more  expensive,  and  the  evidence  on  its  effectiveness  for  disadvantaged  students  varies  largely  depending  of  the  type  of  aid.  The  evidence  on  need‐based  grants  suggests  that  most  grant  schemes  only  lead  to  limited  improvements  in  enrollment  rates,  unless they provide substantial amounts of money. It is possible that enrollment as a response  32    to aid follows a threshold effect and that need‐based aid is only effective when it covers a  significant  part  of  unmet  financial  need  and  determining  such  a  threshold  should  be  an  interesting question for future research. It also seems that an early commitment of aid, while  students are still in high school, leads to much larger impact on higher education access and  this  type  of  grant  could  be  further  tested.  Merit‐based  aid  is  rarely  effective  in  tackling  inequalities  in  higher  education,  except  when  it  includes  a  need‐based  component  to  specifically  support  disadvantaged  students.  Conversely,  merit‐based  aid  based  only  on  academic  results,  without  any  assessment  of  students’  financial  needs,  seems  to  have  no  effect, and was even found to raise inequality. Regarding attainment, only need‐based grants  were found to increase graduation rates of disadvantaged students quite consistently.   Interventions  that  combine  early  financial  aid  and  outreach  activities  are  even  more  demanding for the public purse. Nevertheless, the experimental literature shows promising  results on enrollment and completion of disadvantaged students. Since they support students  through  different  mechanisms,  these  interventions  seem  to  lead  to  large  increases  in  enrolment rates, more consistently than either outreach or financial aid alone. It should also  be noted that effect sizes of these interventions are in the same ballpark as some of the more  effective outreach or financial aid interventions. More needs to be known, therefore, about  the cost effectiveness of these interventions as compared to other types of interventions.   Our systematic review of the literature also allows us to identify areas for which additional  experimental  evidence  is  needed.  Overall,  there  is  still  a  lack  of  available  evidence  on  the  impact of the outreach interventions on graduation rates. As the problem of dropout in higher  education  has  received  increasing  attention,  it  is  crucial  to  provide  causal  evidence  on  the  capacity  of  interventions  to  translate  a  higher  number  of  under‐represented  students  in  higher  education  into  a  higher  number  of  graduates.  Another  shortcoming  of  the  existing  literature is that there is little variation in institutional settings. Most studies discussed here  are from the United States, and further research, in other national and institutional contexts,  is  needed  to  shed  light  on  the  pertinence  of  the  interventions.  To  make  this  literature  comparable and to be able to draw more precise conclusion on the effect of financial aid, we  also  consider  that  studies  should  systematically  report  the  amount  of  the  aid  evaluated  relative to higher education costs (tuition and living expenses) in their specific context. For the  time being, it is very difficult to compare or standardize the amount of aid evaluated as the  33    costs of higher education vary so widely across countries and institutions, and this information  would be crucial to identify a threshold that financial aid needs to cover to increase access  and graduation rates of disadvantaged students.   Nevertheless, most of the evidence discussed here is quite recent and this literature is growing  quickly. We therefore hope that more precise conclusions and policy recommendations could  be drawn in the coming years. Overall, the available evidence from the (quasi‐)experimental  literature  is  encouraging  for  the  institutional  and  political  leverage  to  reduce  inequality  in  higher education. Although some of the inequalities discussed here may arise very early in the  life  course,  our  results  highlight  the  possibility,  and  perhaps  the  necessity,  to  also  tackle  education inequalities later. Well‐designed interventions in high school and higher education  can  thus  bring  about  substantial  improvements  in  the  difficult  educational  careers  of  disadvantaged students.   References  Abbiati, G., Argentin, G., Barone, C., & Schizzerotto, A. (2017). Information barriers and social  stratification in higher education: evidence from a field experiment. The British Journal  of Sociology, (0).  Abbiati,  G.,  &  Barone,  C.  (2017).  Is  university  education  worth  the  investment?  The  expectations  of  upper  secondary  school  seniors  and  the  role  of  family  background.  Rationality and Society, 29(2), 113–159.  Alon, S. (2007). The influence of financial aid in leveling group differences in graduating from  elite institutions. Economics of Education Review, 26(3), 296–311.  Alon, S. (2011). Who Benefits Most from Financial Aid? The Heterogeneous Effect of Need‐ Based Grants on Students’ College Persistence.  Social Science Quarterly,  92(3), 807– 829.  Anderson, D. M., & Goldrick‐Rab, S. (2016). Aid After Enrollment: Impacts of a Statewide Grant  Program  at  Public  Two‐year  Colleges.  Retrieved  from  http://www.wihopelab.com/publications/Anderson‐Goldrick‐Rab‐2016‐Impacts‐of‐ Statewide‐Grant‐Program.pdf  Andrews, R. J., Imberman, S. A., & Lovenheim, M. F. (2016).  Recruiting and Supporting Low‐ Income, High‐Achieving Students at Flagship Universities (Working Paper No. 22260).  National Bureau of Economic Research.  Attewell, P., Lavin, D., Domina, T., & Levey, T. (2006). New Evidence on College Remediation.  The Journal of Higher Education, 77(5), 886–924.  Avery, C. (2010).  The Effects of College Counseling on High‐Achieving, Low‐Income Students  (Working Paper No. 16359). National Bureau of Economic Research.  34    Avery,  C.  (2013).  Evaluation  of  the  College  Possible  Program:  Results  from  a  Randomized  Controlled  Trial  (NBER  Working  Paper  No.  19562).  National  Bureau  of  Economic  Research.  Avery,  C.,  &  Kane,  T.  J.  (2004).  Student  perceptions  of  college  opportunities.  The  Boston  COACH program. In  College choices: The economics of where to go, when to go, and  how  to  pay  for  it  (pp.  355–394).  Retrieved  from  http://www.nber.org/chapters/c10104.pdf  Azzolini, D., Martini, A., Romano, B., & Vergolini, L. (2018). Affording college with the help of  asset building: First experimental impacts from Italy. Economics Letters, 169, 27–30.  Barr, A., & Castleman, B. (2017). The Bottom Line on College Counseling (p. 41). Retrieved from  https://www.bottomline.org/sites/default/files/The%20Bottom%20Line%20on%20C ollege%20Counseling%20RCTPaper_10_2017.pdf  Barrow, L., Richburg‐Hayes, L., Rouse, C. E., & Brock, T. (2014). Paying for Performance: The  Education  Impacts  of  a  Community  College  Scholarship  Program  for  Low‐Income  Adults. Journal of Labor Economics, 32(3), 563–599.  Baumgartner, H. J., & Steiner, V. (2006). Does More Generous Student Aid Increase Enrolment  Rates into Higher Education? Evaluating the German Student Aid Reform of 2001 (SSRN  Scholarly Paper No. ID 892831). Rochester, NY: Social Science Research Network.  Bettinger, E. (2004). How financial aid affects persistence. In  College choices: The economics  of where to go, when to go, and how to pay for it (pp. 207–238). University of Chicago  Press.  Bettinger, E. (2015). Need‐Based Aid and College Persistence The Effects of the Ohio College  Opportunity Grant. Educational Evaluation and Policy Analysis, 37(1 suppl), 102S‐119S.  Bettinger, E., Long, B. T., Oreopoulos, P., & Sanbonmatsu, L. (2012). The Role of Application  Assistance  and  Information  in  College  Decisions:  Results  from  the  H&R  Block  Fafsa  Experiment*. The Quarterly Journal of Economics, 127(3), 1205–1242.  Binder, M., Krause, K., Miller, C., & Cerna, O. (2015).  Providing Incentives for Timely Progress  Toward  Earning  a  College  Degree.  Retrieved  from  http://www.mdrc.org/publication/providing‐incentives‐timely‐progress‐toward‐ earning‐college‐degree  Bird, K. A., Castleman, B., Goodman, J., & Lamberton, C. (2017). Nudging at a National Scale:  Experimental Evidence from a FAFSA Completion Campaign (No. 54; p. 34).  Blagg, K., Chingos, M., Graves, C., & Nicotera, A. (2017).  Rethinking Consumer Information in  higher education (p. 59). Urban Institute.  Bleemer, Z., & Zafar, B. (2018). Intended college attendance: Evidence from an experiment on  college returns and costs. Journal of Public Economics, 157, 184–211.  Bonilla, L., Bottan, N. L., & Ham, A. (2017). Information Policies and Higher Education Choices:  Experimental  Evidence  from  Colombia  (SSRN  Scholarly  Paper  No.  ID  2546835).  Rochester, NY: Social Science Research Network.  Borenstein,  M.  (Ed.).  (2009).  Introduction  to  meta‐analysis.  Chichester,  U.K:  John  Wiley  &  Sons.  35    Bos,  J.,  Berman,  J.,  Kane,  T.  J.,  &  Tseng,  F.  (2012,  November  8).  Does  the  Source  Program  Increase Access to Financial Aid and Support Four‐Year College Enrollment? Presented  at  the  APPAM  Fall  Research  Conference.  Retrieved  from  https://appam.confex.com/appam/2012/webprogram/Paper2589.html  Bruce,  D.  J.,  &  Carruthers,  C.  K.  (2014).  Jackpot?  The  impact  of  lottery  scholarships  on  enrollment in Tennessee. Journal of Urban Economics, 81, 30–44.  Bulman,  G.  B.,  &  Hoxby,  C.  M.  (2015).  The  Returns  to  the  Federal  Tax  Credits  for  Higher  Education. Tax Policy and the Economy, 29(1), 13–88.  Carneiro, P., & Heckman,  J. J. (2002).  The Evidence on Credit  Constraints in Post‐Secondary  Schooling (Working Paper No. 9055). National Bureau of Economic Research.  Carrell, S. E., & Sacerdote, B. (2013).  Why do college going interventions work? (No. 19031).  National Bureau of Economic Research.  Carruthers, C.  K.,  &  Fox,  W.  F.  (2016).  Aid  for  all:  College  coaching,  financial  aid, and  post‐ secondary persistence in Tennessee. Economics of Education Review, 51, 97–112.  Carruthers, C. K., & Özek, U. (2016). Losing HOPE: Financial aid and the line between college  and work. Economics of Education Review, 53, 1–15.  Castleman, B., Arnold, K., & Wartman, K. L. (2012). Stemming the Tide of Summer Melt: An  Experimental Study of the Effects of Post‐High School Summer Intervention on Low‐ Income  Students’  College  Enrollment.  Journal  of  Research  on  Educational  Effectiveness, 5(1), 1–17.  Castleman, B., & Goodman, J. (2014). Intensive College Counseling and the College Enrollment  Choices of Low Income Students (SSRN Scholarly Paper No. ID 2493103). Rochester, NY:  Social Science Research Network.  Castleman, B., & Long, B. (2013). Looking Beyond Enrollment: The Causal Effect of Need‐Based  Grants  on  College  Access,  Persistence,  and  Graduation  (NBER  Working  Paper  No.  19306). National Bureau of Economic Research, Inc.  Castleman, B., & Page, L. (2015). Summer nudging: Can personalized text messages and peer  mentor  outreach  increase  college  going  among  low‐income  high  school  graduates?  Journal of Economic Behavior & Organization, 115, 144–160.  Castleman,  B.,  &  Page,  L.  (2017).  Parental  Influences  on  Postsecondary  Decision  Making:  Evidence  From  a  Text  Messaging  Experiment.  Educational  Evaluation  and  Policy  Analysis, 39(2), 361–377.  Castleman,  B.,  Page,  L.,  &  Schooley,  K.  (2014).  The  Forgotten  Summer:  Does  the  Offer  of  College Counseling After High School Mitigate Summer Melt Among College‐Intending,  Low‐Income  High  School  Graduates?  Journal  of  Policy  Analysis  and  Management,  33(2), 320–344.  Choitz,  V.,  &  Reimherr,  P.  (2013).  Mind  the  Gap:  High  Unmet  Financial  Need  Threatens  Persistence and Completion for Low‐Income Community College Students. Center for  Law and Social Policy.  Clotfelter, C. T., Hemelt, S. W., & Ladd, H. F. (2018). Multifaceted Aid for Low‐Income Students  and College Outcomes: Evidence from North Carolina.  Economic Inquiry,  56(1), 278– 303.  36    Cohodes, S. R., & Goodman, J. S. (2014). Merit Aid, College Quality, and College Completion:  Massachusetts’ Adams Scholarship as an In‐Kind Subsidy. American Economic Journal:  Applied Economics, 6(4), 251–285.  Constantine, J. M., Seftor, N. S., Martin, E. S., Silva, T., & Myers, D. (2006). Study of the Effect  of the Talent Search Program on Secondary and Postsecondary Outcomes in Florida,  Indiana and Texas. Final Report from Phase II of the National Evaluation. ED Pubs.  Cunha, J. M., Miller, T., & Weisburst, E. (2018). Information and College Decisions: Evidence  From the Texas GO Center Project.  Educational Evaluation and Policy Analysis,  40(1),  151–170.  Dearden, L., Fitzsimons, E., & Wyness, G. (2014). Money for nothing: Estimating the impact of  student aid on participation in higher education.  Economics of Education Review,  43,  66–78.  Deming,  D.,  &  Dynarski,  S.  (2009).  Into  College,  Out  of  Poverty?  Policies  to  Increase  the  Postsecondary Attainment of the Poor (Working Paper No. 15387). National Bureau of  Economic Research.  Denning,  J.  T.  (2017).  College  on  the  Cheap:  Consequences  of  Community  College  Tuition  Reductions. American Economic Journal: Economic Policy, 9(2), 155–188.  Denning, J. T. (2018). Born Under a Lucky Star: Financial Aid, College Completion, Labor Supply,  and Credit Constraints. Journal of Human Resources.  Denning,  J.  T.,  Marx,  B.  M.,  &  Turner,  L.  J.  (2017).  ProPelled:  The  Effects  of  Grants  on  Graduation,  Earnings,  and  Welfare  (Working  Paper  No.  23860).  https://doi.org/10.3386/w23860  Domina,  T.  (2009).  What  Works  in  College  Outreach:  Assessing  Targeted  and  Schoolwide  Interventions for Disadvantaged Students. Educational Evaluation and Policy Analysis,  31(2), 127–152.  Dunlop,  E.  (2013).  What  Do  Stanford  Loans  Actually  Buy  You?‐The  Effect  of  Stanford  Loan  Access on Community College Students (CALDER Working Paper No. No. 94).  Dynarski,  S.  (2000).  Hope  for  Whom?  Financial  Aid  for  the  Middle  Class  and  Its  Impact  on  College Attendance. National Tax Journal, 53(3), 2.  Dynarski,  S.  (2003).  Does  Aid  Matter?  Measuring  the  Effect  of  Student  Aid  on  College  Attendance and Completion. The American Economic Review, 93(1), 279–288.  Ehlert,  M.,  Finger,  C.,  Rusconi,  A.,  &  Solga,  H.  (2017).  Applying  to  college:  Do  information  deficits lower the likelihood of college‐eligible students from less‐privileged families to  pursue  their  college  intentions?:  Evidence  from  a  field  experiment.  Social  Science  Research, 67, 193–212.  Elsayed,  M.  A.  A.  (2016).  The  Impact  of  Education  Tax  Benefits  on  College  Completion.  Economics of Education Review, 53, 16–30.  Erikson, R., Goldthorpe, J. H., Jackson, M., Yaish, M., & Cox, D. R. (2005). On class differentials  in  educational  attainment.  Proceedings  of  the  National  Academy  of  Sciences  of  the  United States of America, 102(27), 9730–9733.  37    Fack, G., & Grenet, J. (2015). Improving College Access and Success for Low‐Income Students:  Evidence  from  a  Large  Need‐Based  Grant  Program.  American  Economic  Journal:  Applied Economics, 7(2), 1–34.  Ford, R., Frenette, M., Nicholson, C., Kwakye, I., Hui, T. S., Hutchinson, J., … others. (2012).  Future  to  discover:  Post‐secondary  impacts  report.  Ottawa:  the  Social  Research  and  Demonstration Corporation.  Ford, R., Grekou, D., Kwakye, I., & Nicholson, C. (2014).  Future to Discover: Fourth Year Post‐ Secondary Impacts Report. Ottawa: Social Research and Demonstration Corporation.  Goldrick‐Rab, S., Kelchen, R., Harris, D. N., & Benson, J. (2016). Reducing Income Inequality in  Educational  Attainment:  Experimental  Evidence  on  the  Impact  of  Financial  Aid  on  College Completion. American Journal of Sociology, 121(6), 1762–1817.  Greene, J. P., & Forster, G. (2003). Public High School Graduation and College Readiness Rates  in the United States. Education Working Paper No. 3. Manhattan Institute for Policy  Research.  Grodsky, E., & Jones, M. T. (2007). Real and imagined barriers to college entry: Perceptions of  cost. Social Science Research, 36(2), 745–766.  Gurgand, M., Lorenceau, A. J. S., & Mélonio, T. (2011). Student Loans: Liquidity Constraint and  Higher Education in South Africa (SSRN Scholarly Paper No. ID 1969424). Rochester,  NY: Social Science Research Network.  Hahn,  A.,  Leavitt,  T.,  &  Aaron,  P.  (1994).  Evaluation  of  the  Quantum  Opportunity  Program  (QOP):  Did  the  program  work.  Waltham,  MA:  Brandeis  University,  Heller  Graduate  School.  Hastings, J., Neilson, C. A., & Zimmerman, S. D. (2015).  The Effects of Earnings Disclosure on  College Enrollment Decisions (Working Paper No. 21300). National Bureau of Economic  Research.  Heller,  D.  E.  (1997).  Student  Price  Response  in  Higher  Education:  An  Update  to  Leslie  and  Brinkman. The Journal of Higher Education, 68(6), 624–659.  Herber, S. P. (2018). The role of information in the application for highly selective scholarships:  Evidence  from  a  randomized  field  experiment.  Economics  of  Education  Review,  62,  287–301.  Herber,  S.  P.,  &  Kalinowski,  M.  (2016).  Non‐Take‐Up  of  Student  Financial  Help:  A  Microsimulation for Germany. Retrieved from DIW Berlin, The German Socio‐Economic  Panel  (SOEP)  website:  http://econpapers.repec.org/paper/diwdiwsop/diw_5fsp844.htm  Hillmert, S., & Jacob, M. (2003). Social Inequality in Higher Education. Is Vocational Training a  Pathway  Leading  to  or  Away  from  University?  European  Sociological  Review,  19(3),  319–334.  Hoxby,  C.,  &  Turner,  S.  (2013).  Expanding  College  Opportunities  for  High‐Achieving,  Low  Income Students (Discussion Paper No. 12–014). Stanford Institute for Economic Policy  Research.  Jabbar,  H.  (2011).  The  Behavioral  Economics  of  Education  New  Directions  for  Research.  Educational Researcher, 40(9), 446–453.  38    Jackson,  C.  K.  (2010).  A  Little  Now  for  a  Lot  Later  A  Look  at  a  Texas  Advanced  Placement  Incentive Program. Journal of Human Resources, 45(3), 591–639.  Kane,  T.  J.  (1994).  College  Entry  by  Blacks  since  1970:  The  Role  of  College  Costs,  Family  Background, and the Returns to Education. Journal of Political Economy, 102(5), 878– 911.  Kane, T. J. (1995). Rising Public College Tuition and College Entry: How Well Do Public Subsidies  Promote Access to College? (Working Paper No. 5164). National Bureau of Economic  Research.  Kane, T. J. (2003).  A Quasi‐Experimental Estimate of the Impact of Financial Aid on College‐ Going (Working Paper No. 9703). National Bureau of Economic Research.  Kerr,  S.  P.,  Pekkarinen,  T.,  Sarvimäki,  M.,  &  Uusitalo,  R.  (2014).  Educational  Choice  and  Information  on  Labor  Market  Prospects:  A  Randomized  Field  Experiment.  Retrieved  from  Working  Paper  website:  http://www.demm.unimi.it/extfiles/unimidire/100601/attachment/pekkarinen.pdf  King,  J.  E.  (2002).  Crucial  Choices:  How  Students’  Financial  Decisions  Affect  Their  Academic  Success.  King,  J.  E.  (2004).  Missed  opportunities:  Students  who  do  not  apply  for  financial  aid.  Washington, DC: American Council on Education.  Kofoed, M. S. (2017). To Apply or Not to Apply: FAFSA Completion and  Financial Aid Gaps.  Research in Higher Education, 58(1), 1–39.  LaLumia,  S.  (2012).  Tax  Preferences  For  Higher  Education  And  Adult  College  Enrollment.  National Tax Journal, (1), 59.  Lavecchia, A., Liu, H., & Oreopoulos, P. (2015).  Behavioral Economics of Education: Progress  and Possibilities (IZA Discussion Paper No. 8853). Institute for the Study of Labor (IZA).  Leslie, L. L., & Brinkman, P. T. (1987). Student Price Response in Higher Education: The Student  Demand Studies. The Journal of Higher Education, 58(2), 181–204.  Linsenmeier, D.  M., Rosen,  H. S., & Rouse, C. E. (2006).  Financial Aid Packages and College  Enrollment Decisions: An Econometric Case Study. Review of Economics and Statistics,  88(1), 126–145.  Liu,  C.,  Zhang,  L.,  Luo,  R.,  Wang,  X.,  Rozelle,  S.,  Sharbono,  B.,  …  Glauben,  T.  (2011).  Early  commitment on financial aid and college decision making of poor students: Evidence  from a randomized evaluation in rural China.  Economics of Education Review,  30(4),  627–640.  Long, B. T. (2008).  What is Known about the Impact of Financial Aid? Implications for Policy.  An NCPR Working Paper. National Center for Postsecondary Research.  Lovenheim, M. F., & Owens, E. G. (2014). Does federal financial aid affect college enrollment?  Evidence from drug offenders and the Higher Education Act of 1998. Journal of Urban  Economics, 81, 1–13.  Loyalka, P., Song, Y., Wei, J., Zhong, W., & Rozelle, S. (2013). Information, college decisions  and  financial  aid:  Evidence  from  a  cluster‐randomized  controlled  trial  in  China.  Economics of Education Review, 36, 26–40.  39    Ma, J., Baum, S., Pender, M., & Bell, D. (2015). Trends in College Pricing 2015.  New York, NY:  The College Board.  Mayer,  A.,  Patel,  R.,  &  Gutierrez,  M.  (2015).  Four‐Year  Effects  on  Degree  Receipt  and  Employment Outcomes from a Performance‐Based Scholarship Program in Ohio (SSRN  Scholarly Paper No. ID 2594482). Rochester, NY: Social Science Research Network.  Myers, D., Olsen, R., Seftor, N. S., Young, J., & Tuttle, C. (2004). The Impacts of Regular Upward  Bound: Results from the Third Follow‐Up Data Collection.  OECD. (2012). Indicator B5 How much do tertiary students pay and what public support do  they  receive?  In  Education  at  a  Glance  (pp.  272–285).  Retrieved  from  http://www.oecd‐ilibrary.org/content/chapter/eag‐2012‐20‐en  Olson, L., & Rosenfeld, R. A. (1985). Parents, students, and knowledge of college costs. Journal  of Student Financial Aid, 15(1), 4.  Oreopoulos, P., Brown, R. S., & Lavecchia, A. M. (2014). Pathways to Education: An Integrated  Approach to Helping At‐Risk High School Students (Working Paper No. 20430). National  Bureau of Economic Research.  Oreopoulos,  P.,  &  Dunn,  R.  (2012).  Information  and  College  Access:  Evidence  from  a  Randomized  Field  Experiment  (Working  Paper  No.  18551).  National  Bureau  of  Economic Research.  Page, L., Castleman, B., & Sahadewo, G. A. (2016). More than Dollars for Scholars: The Impact  of the Dell Scholars Program on College Access, Persistence and Degree Attainment.  Page, L., & Scott‐Clayton, J. (2016). Improving college access in the United States: Barriers and  policy responses. Economics of Education Review, 51, 4–22.  Pallais,  A.  (2013).  Small  Differences  that  Matter:  Mistakes  in  Applying  to  College  (Working  Paper No. 19480). National Bureau of Economic Research.  Paulsen, M. B., & St. John, E. P. (2002). Social Class and College Costs: Examining the Financial  Nexus between College Choice and Persistence. The Journal of Higher Education, 73(2),  189–236.  Peter, F. H., & Zambre, V. (2017). Intended college enrollment and educational inequality: Do  students lack information? Economics of Education Review, 60, 125–141.  Pistolesi, N. (2017). Advising students on their field of study: Evidence from a French University  reform. Labour Economics, 44, 106–121.  Richburg‐Hayes, L., Patel, R., Brock, T., Campa, D. la, Elijah, Rudd, T., & Valenzuela, I. (2015).  Providing  More  Cash  for  College:  Interim  Findings  from  the  Performance‐Based  Scholarship  Demonstration  in  California  (SSRN  Scholarly  Paper  No.  ID  2625711).  Rochester, NY: Social Science Research Network.  Rodríguez‐Planas,  N.  (2012).  Longer‐Term  Impacts  of  Mentoring,  Educational  Services,  and  Learning Incentives: Evidence from a Randomized Trial in the United States. American  Economic Journal: Applied Economics, 4(4), 121–139.  Rosinger, K. (2016).  Can Simplifying Financial Aid Information Impact College Enrollment and  Borrowing? Experimental and Quasi‐Experimental Evidence.  40    Ross,  R.,  White,  S.,  Wright,  J.,  &  Knapp,  L.  (2013).  Using  behavioral  economics  for  postsecondary success. New York: ideas42.  Rubin,  R.  B.  (2011).  The  Pell  and  the  Poor:  A  Regression‐Discontinuity  Analysis  of  On‐Time  College Enrollment. Research in Higher Education, 52(7), 675–692.  Scott‐Clayton,  J.  (2011).  The  Shapeless  River:  Does  a  Lack  of  Structure  Inhibit  Students’  Progress at Community Colleges? CCRC Working Paper No. 25. Assessment of Evidence  Series. Community College Research Center.  Scott‐Clayton, J. (2012).  What Explains Trends in Labor Supply Among U.S. Undergraduates,  1970‐2009? (Working Paper No. 17744). National Bureau of Economic Research.  Scott‐Clayton, J., & Rodriguez, O. (2014). Development, Discouragement, or Diversion? New  Evidence on the Effects of College Remediation Policy.  Education Finance and Policy,  10(1), 4–45.  Scrivener, S., Weiss, M. J., Ratledge, A., Rudd, T., Sommo, C., & Fresques, H. (2015). Doubling  Graduation  Rates:  Three‐Year  Effects  of  CUNY’s  Accelerated  Study  in  Associate  Programs (ASAP) for Developmental Education Students (SSRN Scholarly Paper No. ID  2571456). Rochester, NY: Social Science Research Network.  Seftor,  N.  S.,  Mamun,  A.,  &  Schirm,  A.  (2009).  The  Impacts  of  Regular  Upward  Bound  on  Postsecondary  Outcomes  7‐9  Years  After  Scheduled  High  School  Graduation  [Mathematica Policy Research Reports]. Mathematica Policy Research.  Sjoquist, D. L., & Winters, J. V. (2015). State Merit‐Based Financial Aid Programs and College  Attainment. Journal of Regional Science, 55(3), 364–390.  Sneyers, E., & Witte, K. D. (2018). Interventions in higher education and their effect on student  success: a meta‐analysis. Educational Review, 70(2), 208–228.  Solis,  A.  (2013).  Credit  Access  and  College  Enrollment  (Working  Paper  Series  No.  2013:12).  Retrieved  from  Uppsala  University,  Department  of  Economics  website:  http://econpapers.repec.org/paper/hhsuunewp/2013_5f012.htm  Sparks, D., & Malkus, N. (2013). First‐Year Undergraduate Remedial Coursetaking: 1999‐2000,  2003‐04,  2007‐08.  Statistics  in  Brief.  NCES  2013‐013.  National  Center  for  Education  Statistics.  Stephan, J. L., & Rosenbaum, J. E. (2013). Can High Schools Reduce College Enrollment Gaps  With a New Counseling Model? Educational Evaluation and Policy Analysis, 35(2), 200– 219.  Turner, S., & Bound, J. (2003). Closing the Gap or Widening the Divide: The Effects of the G.I.  Bill and World War II on the Educational Outcomes of Black Americans. The Journal of  Economic History, 63(1), 145–177.  Usher,  A.  (2005).  A  Little  Knowledge  is  A  Dangerous  Thing:  How  Perceptions  of  Costs  and  Benefits Affect Access to Education. Toronto: Educational Policy Institute.  Vergolini, L., Zanini, N., & Bazoli, N. (2014).  Liquidity Constraints and University Participation  in Times of Recession. Evidence from a Small‐scale Programme. Research Institute for  the Evaluation of Public Policies (IRVAPP), Bruno Kessler Foundation.  41    Welch,  J.  G.  (2014).  HOPE  for  community  college  students:  The  impact  of  merit  aid  on  persistence, graduation, and earnings. Economics of Education Review, 43, 1–20.  Wiederspan,  M.  (2016).  Denying  loan  access:  The  student‐level  consequences  when  community  colleges  opt  out  of  the  Stafford  loan  program.  Economics  of  Education  Review, 51, 79–96.  Williams, G., & Gordon, A. (1981). Perceived earnings functions and ex ante rates of return to  post compulsory education in England. Higher Education, 10(2), 199–227.  Wolter, S. C. (2000). Wage Expectations: A Comparison of Swiss and US Students. Kyklos, 53(1),  51–69.        42    Appendix   Table A.1: Selected publications for the systematic literature review  Authors  Date  Title  Intervention  Type of  Design  Country  publication  Abbiati ,  2017  Information barriers and social  Outreach  Journal  RCT  Italy  Argentin, Barone  stratification in higher education:  & Schizzerotto  evidence from a field experiment  Alon  2007  The influence of financial aid in  Financial aid  Journal  IV  United  leveling group differences in  States  graduating from elite institutions  Alon   2011  Who Benefits Most from Financial Aid?  Financial aid  Journal  IV  United  The Heterogeneous Effect of Need‐ States  Based Grants on Students’ College  Persistence  Anderson&  2016  Aid After Enrollment: Impacts of a  Financial aid  Unpublished  RCT  United  Goldrick Rab  Statewide Grant Program at Public  States  Two‐year Colleges  Andrew,  2016  Recruiting and Supporting Low‐ Mixed  Unpublished  DiD  United  Imberman &  Income, High‐Achieving Students at  intervention   States  Lovenheim  Flagship Universities  Avery  2013  Evaluation of the College Possible  Outreach  Unpublished  RCT  United  program: Results from a randomized  States  controlled trial  Avery  2010  The Effects of College Counseling on  Outreach  Unpublished  RCT  United  High‐Achieving, Low‐Income Students  States  Azzolini, Martini,  2018  Affording college with the help of  Financial aid  Journal  RCT  Italy  Romano &  asset building: First experimental  Vergolini  impacts from Italy  Barr & Castleman  2017  The Bottom Line on College Counseling  Outreach  Unpublished  RCT  United  States  Barrow,  2014  Paying for Performance: The Education  Financial aid  Journal  RCT  United  Richburg‐Hayes,  Impacts of a Community College  States  Rouse, & Brock  Scholarship Program for Low‐Income  Adults  Baumgartner &  2006  Does More Generous Student Aid  Financial aid  Unpublished  DiD  Germany  Steiner  Increase Enrolment Rates into Higher  Education? Evaluating the German  Student Aid Reform of 2001  Bettinger  2015  Need‐Based Aid and College  Financial aid  Journal  DiD  United  Persistence: The Effects of the Ohio  States  College Opportunity Grant  Bettinger, Long,  2012  The Role Of Application Assistance And  Outreach  Journal  RCT  United  Oreopoulos, &  Information In College Decisions:  States  Sanbonmatsu  Results From The H&R Block Fafsa  Experiment  Binder, Krause,  2015  Providing Incentives for Timely  Financial aid  Unpublished  RCT  United  Miller, & Cerna  Progress Toward Earning a College  States  Degree Results from a Performance‐ Based Scholarship Experiment  Bird, Castleman,  2017  Nudging at a National Scale:  Outreach  Unpublished  RCT  United  Goodman &  Experimental Evidence from a FAFSA  States  Lamberton  Completion Campaign  43    Bonilla, Bottan, &  2017  Information Policies and Higher  Outreach  Unpublished  RCT  Colombia  Ham  Education Choices. Experimental  Evidence from Colombia  Bos et al   2012  The Impacts of SOURCE ‐ A Program to  Outreach  Unpublished  RCT  United  Support College Enrollment through  States  Near‐Peer, Low‐Cost Student Advising  Bruce &  2014  Jackpot? The impact of lottery  Financial aid  Journal  RD  United  Carruthers  scholarships on enrollment in  States  Tennessee  Bulman & Hoxby  2015  The Returns to the Federal Tax Credits  Financial aid  Journal  IV  United  for Higher Education  States  Carell &  2013  Late interventions matter too: the case  Outreach  Unpublished  RCT  United  Sacerdote  of college coaching in New Hampshire  States  Carruthers & Fox  2016  Aid for all: College coaching, financial  Mixed  Journal  PSM  United  aid, and post‐secondary persistence in  intervention   States  Tennessee  Carruthers &  2016  Losing HOPE: Financial aid and the line  Financial aid  Journal  RD  United  Ozek  between college and work  States  Castleman &  2014  Intensive College Counseling and the  Outreach  Unpublished  RD  United  Goodman  College Enrollment Choices of Low  States  Income Students   Castleman &  2013  Looking beyond enrollment: The  Financial aid  Unpublished  RD  United  Long  causal effect of need‐based grants on  States  college access, persistence, and  graduation  Castleman &  2015  Summer Nudging: Can Personalized  Outreach  Journal  RCT  United  Page  Text Messages and Peer Mentor  States  Outreach Increase College Going  Among Low‐Income High School  Graduates?  Castleman &  2017  Parental Influences on Postsecondary  Outreach  Journal  RCT  United  Page  Decision Making: Evidence From a Text  States  Messaging Experiment  Castleman,  2012  Stemming the Tide of Summer Melt  Outreach  Journal  RCT  United  Arnold, &  States  Wartman  Castleman, Owen  Stay late or start early? Experimental  2015  Outreach  Journal  RCT  United  & Page  evidence on the benefits of college  States  matriculation support from high  schools versus colleges  Castleman, Page,  2014  The Forgotten Summer: Does the Offer  Outreach  Journal  RCT  United  & Schooley  of College Counseling After High  States  School Mitigate Summer Melt Among  College‐Intending, Low‐Income High  School Graduates?  Clotfelter, Hemelt  2018  Multifaceted aid for low‐income  Financial aid;  Journal  RD  United  & Ladd  students and college outcomes:  Mixed  States  evidence from North‐Carolina  intervention   Cohodes &  2014  Merit Aid, College Quality, and College  Financial aid  Journal  RD  United  Goodman  Completion: Massachusetts’ Adams  States  Scholarship as an In‐Kind Subsidy  Constantine,  2006  A Study of the Effect of the Talent  Outreach  Unpublished  PSM  United  Seftor, Martin,  Search Program on Secondary and  States  Silva, & Myers  Postsecondary Outcomes in Florida,  Indiana and Texas  44    Cunha, Miller &  2018  Information and College Decisions:  Outreach  Journal  DiD  United  Weisburst  Evidence From the Texas GO Center  States  Project  Dearden,  2014  Money for nothing: Estimating the  Financial aid  Journal  DiD  United  Fitzsimmons,  impact of student aid on participation  Kingdom  Wyness  in higher education  Denning, Marx &  2017  Propelled: the effects of grants on  Financial aid  Unpublished  RD  United  Turner  graduation, earnings, and welfare  States  Denning  2017  College on the Cheap: Consequences  Financial aid  Journal  DiD  United  of Community College Tuition  States  Reductions  Denning  2018  Born Under a Lucky Star: Financial Aid,  Financial aid  Journal  RD  United  College Completion, Labor Supply, and  States  Credit Constraints  Domina  2009  What Works in College Outreach:  Outreach  Journal  PSM  United  Assessing Targeted and Schoolwide  States  Interventions for Disadvantaged  Students  Dunlop  2013  What Do Stafford Loans Actually Buy  Financial aid  Unpublished  IV  United  You? The Effect of Stafford Loan  States  Access on Community College  Students  Dynarski  2000  Hope for Whom? Financial Aid for the  Financial aid  Journal  DiD  United  Middle Class and Its Impact on College  States  Attendance  Dynarski  2003  Does Aid Matter? Measuring the Effect  Financial aid  Journal  DiD  United  of Student Aid on College Attendance  States  and Completion  Elsayed  2016  The Impact of Education Tax Benefits  Financial aid  Journal  PSM  United  on College Completion  States  Fack & Grenet  2015  Improving College Access and Success  Financial aid  Journal  RD  France  for Low‐Income Students: Evidence  from a Large Need‐Based Grant  Program  Ford et al.  2012  Future to Discover: Post‐secondary  Outreach;  Unpublished  RCT  Canada  Impacts Report  Financial aid;  Mixed  intervention   Ford, Grekou,  2014  Future to Discover: Fourth Year Post‐ Outreach;   Unpublished  RCT  Canada  Kwakye, &  Secondary Impacts Report  Financial aid;   Nicholson  Mixed  intervention   Goldrick‐Rab,  2016  Reducing Income Inequality in  Financial aid  Journal  RCT  United  Harris, Kelchen &  Educational Attainment: Experimental  States  Benson  Evidence on the Impact of Financial  Aid on College Completion  Gurgand,  2011  Student Loans: Liquidity Constraint  Financial aid  Unpublished  RD  South  Lorenceau &  and Higher Education in South Africa  Africa  Melonio  Hahn, Leavitt, &  1994  Evaluation of the Quantum  Mixed  Unpublished  RCT  United  Aaron  Opportunities Program (QOP). Did the  intervention   States  Program Work?  Hastings, Neilson,  2015  The effects of Earnings Disclosure on  Outreach  Unpublished  RCT  Chile  & Zimmerman  College Enrollment Decisions  45    Hoxby & Turner  2013  Expanding college opportunities for  Outreach  Unpublished  RCT  United  high‐achieving, low income students.  States  Jackson  2010  A Little Now for a Lot Later: A Look at a  Financial aid  Journal  DiD  United  Texas Advanced Placement Incentive  States  Program  Kane  2003  A Quasi‐Experimental Estimate of the  Financial aid  Unpublished  RD  United  Impact of Financial Aid on College‐ States  Going  Kane  1995  Rising Public College Tuition Fees and  Financial aid  Unpublished  DiD  United  College Entry. How well do public  States  subsidies promote access to college?  Kerr, Pekkarinen,  2014  Educational Choice and Information on  Outreach  Unpublished  RCT  Finland  Sarvimäki, &  Labor Market Prospects: A  Uusitalo  Randomized Field Experiment  LaLumia  2012  Tax Preferences for Higher Education  Financial aid  Journal  IV  United  And Adult College Enrollment  States  Linsenmeier,  2006  Financial Aid Packages and College  Financial aid  Journal  DiD  United  Rosen, & Rouse  Enrollment Decisions: An Econometric  States  Case Study  Lovenheim &  2014  Does federal financial aid affect  Financial aid  Journal  DiD  United  Owens  college enrollment? Evidence from  States  drug offenders and the Higher  Education Act of 1998  Loyalka, Song,  2013  Information, college decisions and  Outreach  Journal  RCT  China  Wei, Zhong, &  financial aid: Evidence from a cluster‐ Rozelle  randomized controlled trial in China  Mayer, Patel, &  2015  Four‐Year Effects on Degree Receipt  Financial aid  Unpublished  RCT  United  Gutierrez  and Employment Outcomes from a  States  Performance‐Based Scholarship  Program in Ohio  Myers et al.   2004  The Impacts of Regular Upward  Outreach  Unpublished  RCT  United  Bound:  Results from the Third Follow‐ States  Up Data Collection   Oreopoulos,  2014  Pathways to Education: An Integrated  Mixed  Unpublished  DiD  Canada  Brown, &  Approach to Helping At‐Risk High  intervention   Lavecchia  School Students  Page, Castleman  2016  More than Dollars for Scholars: The  Mixed  Unpublished  RD  United  & Sahadewo  Impact of the Dell Scholars Program on  intervention   States  College Access, Persistence and  Degree Attainment  Richburg‐Hayes  2015  Providing More Cash for College:  Financial aid  Unpublished  RCT  United  et al.  Interim Findings from the  States  Performance‐Based Scholarship  Demonstration in California  Rodríguez‐Planas  2012  Longer‐Term Impacts of Mentoring,  Mixed  Journal  RCT  United  Educational Services, and Learning  intervention   States  Incentives: Evidence from a  Randomized Trial in the United States  Rosinger  2016  Can Simplifying Financial Aid  Outreach  Unpublished  RCT  United  Information Impact College Enrollment  States  and Borrowing? Experimental and  Quasi‐Experimental Evidence  46    Rubin  2011  The Pell and the Poor: A Regression‐ Financial aid  Journal  RD  United  Discontinuity Analysis of On‐Time  States  College Enrollment  Scrivener et al.  2015  Doubling graduation rates: Three‐year  Mixed  Unpublished  RCT  United  effects of CUNY’s Accelerated Study in  intervention   States  Associate Programs (ASAP) for  developmental education students  Seftor, Mamun, &  2009  The Impacts of Regular Upward Bound  Outreach  Unpublished  RCT  United  Schirm  on Postsecondary Outcomes 7‐9 Years  States  after Scheduled High School  Graduation  Sjoquist &  2015  State Merit‐based Financial Aid  Financial aid  Journal  DiD  United  Winters  Programs and College Attainment  States  Solis  2013  Credit access and college enrollment  Financial aid  Unpublished  RD  Chile  Stephan &  2013  Can High Schools Reduce College  Outreach  Journal  DiD  United  Rosenbaum  Enrollment Gaps With a New  States  Counseling Model?  Turner & Bound  2003  Closing the Gap or Widening the  Financial aid  Journal  RD  United  Divide: The Effects of the G.I. Bill and  States  World War II on the Educational  Outcomes of Black Americans  Vergolini, Zanini,  2014  Liquidity Constraints and University  Financial aid  Unpublished  RD  Italy  Bazoli, & others  Participation in Times of Recession.  Evidence from a Small‐scale  Programme  Welch  2014  HOPE for community college students:  Financial aid  Journal  RD  United  The impact of merit aid on  States  persistence, graduation, and earnings  Wiederspan  2016  Denying loan access: The student‐level  Financial aid  Journal  IV  United  consequences when community  States  colleges opt out of the Stafford loan  program  Total  75 publications  RCT: Randomized Control Trial     RD: Regression Discontinuity     DiD: Difference‐in‐Differences     IV: Instrumental variable     PSM: Propensity Score Matching     47    Appendix B: Causal estimates on the effect of outreach interventions on disadvantaged students  Table B.1: the impact of outreach programs (any type) on access to postsecondary education  Baseline in Evaluation Authors Intervention Location/ Details of intervention Disadvantaged group Estimated Outcome control group Design (Year) (Country) Time of evaluation (duration) (Sample size) effect (p.p.) (%) PSM Domina College outreach Nationally Any type of outreach Disadvantaged high Enrolment 73.9 +5.5 (2009) programs representative sample programs school students (any) (United States) of students/ (?) (N=940) By 2 years after high Enrolment in 44.4 +0.2 school graduation 4-year institution   Table B.2: the impact of "information" outreach programs on access to higher education  Baseline in Evaluation Authors Intervention Location/ Details of intervention Disadvantaged group Estimated Outcome control group Design (Year) (Country) Time of evaluation (duration) (Sample size) effect (p.p.) (%) RCT Abbiati et al. Information Four Italian provinces Detailed and Senior high school Enrolment 39.3 -3.2 (2017) intervention (Milano, Vicenza, personalized students with low- (any) (Italy) Bologna, Salerno)/ information about: (1) educated parents Fall following high the costs of higher (N=1,364) school graduation education; (2) the Enrolment in 7.1 -0.07 occupational "strong" fields prospects of of study graduates; (3) the chances of Senior high school Enrolment 43.2 -0.6 successfully students from the (any) completing specific working class higher education (N=1,767) programmes. Enrolment in 10.3 0.4 (3 meetings during “strong" fields school year) of study 48    RCT Bettinger et al. H&R Block Fafsa Ohio and North Information on Low-income 17-year- Enrolment 34.2 -0.4 (2012) Experiment Carolina/ financial aid: olds whose (any) (United States) Year following the individualized parents/families experiment aid eligibility received treatment estimates (N=868) (one time) Low-income young Enrolment 9.5 +0.3 adults, with no prior (any) college (N=9,228) Low-income young Enrolment 26.3 +1.3 adults, with some prior (any) college (N=6,646) RCT Bird et al. Information-only National/ Messages with First-generation Enrolment 81.7 +0.8 (2017) financial aid nudge Fall following high information on college-intending high (any) campaign school graduation financial benefits of school seniors Enrolment at 12 +0.8 (United States) FASFA completion, (N=32,079) 2-year making salient the institution monetary gains Enrolment at 69.7 +0.08 (2-4 emails and 5 text 4-year messages) institution RCT Bonilla, Bottan, & Information Bogota/ Presentation by Low-income high Enrolment 44.8a +0.6 Ham presentation Year following the college graduates with school seniors in public (any) (2017) (Colombia) experiment information on returns schools to higher education, (N=6,003) Enrolment in 9.6a +2.4 financial aid and academic admission criteria degree RCT Hastings, Neilson Disclosure of National/ Consultation of web Low-SES High school Enrolment 77a 0.0 & Zimmerman information on By one year after pages including graduates applying to (any) (2015) costs and returns treatment information on costs federal student loan (Chile) and returns of (N=16,594) different tertiary programs (one time) 49    RCT & Kerr et al Information National sample of PowerPoint High school seniors Enrolment ? -1.0 DiD (2014) campaign on the schools/ presentation with from low-educated (any) returns to education One year after information on the districts -Males (Finland) treatment returns to education (45 minutes) High school seniors Enrolment ? +0.8 from low-educated (any) districts -Females RCT Loyalka et al Information Shaanxi/ Information on High school seniors in Enrolment 53 +8** (2013) campaign on 8 months after college costs and the poorest counties (any) college costs and treatment financial aid through a (N=2,256) financial aid booklet and an oral (China) presentation (20 minutes) RCT Rosinger Information in One public university/ Inclusion of a Pell-eligible students Institutional 48a -4.1b (2015) financial aid award Immediately after shopping sheet in the admitted to the enrolment notifications treatment online financial aid university (yield rate) (United States) award notification, (N=2,471) providing personalized information about costs and loan options. a: Refers to the whole control group, not specific to disadvantaged students. b: Own calculations based on interaction terms.    Estimates plotted in figure 2       50    Table B.3: the impact of "information and guidance" outreach programs on access to higher education  Baseline in Evaluation Authors Intervention Location/ Details of intervention Disadvantaged group Estimated Outcome control group Design (Year) (Country) Time of evaluation (duration) (Sample size) effect (p.p.) (%) RCT Avery Individualized New York/ Individualized advice on High-Achieving, Low- Enrolment in ~42 +7.9 (2010) college counseling ? the choice of college Income high school most (United States) application, completion seniors competitive of college application, (N=106) institutions financial aid and college choice (10 hours over school year) RCT Barr & Castleman Bottom Line Boston/ Individualized Low-income, first- Enrolment 82.7 +7.0*** (2017) college advising Fall after high school counseling providing generation junior or (any) model graduation comprehensive college senior high school (United States) and financial aid support students with minimum Enrolment at 70.3 +10.3*** (One-hour individual GPA of 2.5 2-year meeting per month) institution Enrolment at 12.7 -3.4** 4-year institution RCT Bettinger et al. H&R Block Fafsa Ohio and North -Information on Low-income 17-year- Enrolment 34.2 +8.1** (2012) Experiment Carolina/ Year financial aid & olds whose (any) (United States) following the - parents/families Enrolment at 17.6 +4.7* experiment Simplification/assistance received treatment 2-year with financial aid (N=788) institution application Enrolment at 15.8 +3.7 4-year (one time) institution Low-income young Enrolment 9.5 +1.5** adults, with no prior (any) college Enrolment at 6,2 +0.8 (N= 8,506) 2-year institution     51    Enrolment at 3,1 +0.5 4-year institution Low-income young Enrolment 26.3 -0.3 adults, with some prior (any) college (N=6,646) RCT Bird et al. Information-only National/ Messages with planning First-generation Enrolment 81.7 +1.7** (2017) financial aid nudge Fall following high prompts for FASFA college-intending high (any) campaign school graduation completion, with focus school seniors Enrolment at 12 +1.2* (United States) on logistics and step-by- (N=32,079) 2-year step guidance for institution completion (2-4 emails and 5 text Enrolment at 69.7 +0.45 messages) 4-year institution RCT Bos et al. Student Outreach Los Angeles, Outreach from advisors Junior high school Enrolment at 40.4 +10.6*** (2012) for College California/ to support, counsel, and students whose primary 4-year Enrollment 18 months after high oversee the college and language is Spanish institution (SOURCE) school graduation financial aid (N=1,129) program identification, Junior high school Enrolment at 49.3 +6.1*** (United States) application, and students whose parents 4-year admissions process did not attend college institution (over one year) (N=2,037) RCT Carell & Sacerdote Mentoring program New Hampshire Weekly meetings to help Non-white high school Enrolment 51.8a +17.1***b (2013) with financial completing FASFA and seniors (any) incentives college applications with (N=419) Enrolment at 22.7a +15.4***b (United States) financial incentives: 4-year application fee waivers institution and a $100 cash bonus Low-income high Enrolment 51.8a +20.2**b for completing the school seniors (any) process (N=419) Enrolment at 22.7a +17.3**b (over one month) 4-year institution     52    RCT Castleman & Page Outreach during Dallas, Boston, Text messaging Low-income college- Enrolment 69.6 +1.9 (2015) summer after high Lawrence & campaign reminding intending high school (any) school graduation Springfield, students of tasks graduates Enrolment at 20.2 +3** (United States) Philadelphia/ required by intended (N=5,753) 2-year Fall after high school college and to connect institution graduation them with counsellor- based support Enrolment at 38.6 -1.8 (10 texts sent over the 4-year summer) institution Peer-mentor Low-income college- Enrolment 67.6 +2.3 interventions with intending high school (any) proactive outreach graduates Enrolment at 14.2 -0.4 during summer (N=3,276) 2-year (over 2 months) institution Enrolment at 38.8 +4.5* 4-year institution RCT Castleman & Page Outreach during Massachusetts and Text messaging Low-income college- Enrolment 66.4 +5.7*** (2017) summer after high Florida/ campaign reminding intending high school (any) school graduation Fall after high school students of tasks graduates Enrolment at 24.3 +5.1** (United States) graduation required for college (N=2,010) 2-year enrolment and offering institution help from counselors. Enrolment at 42.1 +0.5 Texts sent to students or 4-year to both students and institution parents. First-generation Enrolment 63.8 +4.5* (14 texts sent over the college-intending high (any) summer) school graduates Enrolment at 20.8 -0.3 (N=1,448) 2-year institution Enrolment at 42.9 +4.8* 4-year institution     53    RCT Castleman, Arnold Summer Providence, Rhode Proactive outreach from All graduates from Enrolment ? +13* and Wartman individualized Island/ counselors during the high schools with (any) (2012) counseling Fall after high school summer focusing on predominantly non- Enrolment at ? -4 (United States) graduation financial aid package, white and low-income 2-year information barriers & students institution social/emotional barriers (N=162) Enrolment at 26 +14* to enrolment 4-year (over 2 months) institution RCT Castleman, Owen Summer college University of New Proactive outreach from Hispanic high school Enrolment 84 +9.5** & Page matriculation Mexico, Albuquerque/ a high school- or graduates admitted to (any) (2015) support Fall after high school college-based counselor, university -Males (United States) graduation during the summer, (N=290) focusing on help to complete required Hispanic high school Enrolment 93 -1.1 summer tasks (financial graduates admitted to (any) aid, loan options, university-Females procedural tasks...) (N=513) (over 2 months) RCT Castleman, Page & Summer counseling Boston (MA)/ Proactive outreach from Lowest-income Enrolment 76.3 +12.3*** Schooley intervention Fall after high school counselors during the college-intending high (any) (2014) (United States) graduation summer with school graduates information on college (N=487) affordability, enrolment process and social barriers (2 months) Fulton County (GA)/ Proactive outreach from Lowest-income Enrolment 63.4 +8.5* Fall after high school counselors during the college-intending high (any) graduation summer school graduates (2 months) (N=586)     54    RCT Ford et al. Explore Your Manitoba/ After-school project Low-income and first- Enrolment 53.7 +9.4 (2012) Horizons program 2 years after high activities with enhanced generation high school (any) (Canada) school graduation career education and students (from 10th Enrolment at 17.4 +11.4* focused information on grade) college (short) post-secondary studies. (N=873) (40 hours over 3-year Enrolment at 33.8 +0.8 period) university RCT Ford et al. Explore Your New Brunswick/ After-school project Low-income and first- Enrolment 38.5 +10.1*** (2014) Horizons program 4 years after high activities with enhanced generation high school (any) (Canada) school graduation career education and students (from 10th Enrolment in 21.8 +1.5 focused information on grade) college (short) post-secondary studies. (N=1,033) (40 hours over 3-year Enrolment at 18.2 +7.7*** period) university RCT Hoxby & Turner ECO National level/ Materials sent by mail High-performing low- Enrolment in a 28.6 +5.3** (2013) Comprehensive One year after high combining Application income high school "peer college": Intervention school graduation Guidance, Net cost seniors matching (United States) information in selective (N=6,000) students' score colleges, and Fee Waiver to apply to selective colleges RD+IV Castleman & "Bottom Line" Boston and Outreach during senior Low-income college- Enrolment at 29 -35,5** Goodman (United States) Worcester, year to encourage ready students in senior 2-year (2014) Massachusetts/ students to apply to a set year of high school institution Fall after high school of target colleges: (N=2,881) graduation regular meetings with a Enrolment at 50 +17.3 counselor to help 4-year navigate the college institution application process (Over one year)     55    PSM Constantine et al. Talent search Texas/ Information about Primarily targeting Enrolment 40 +18*** (2006) program 4, 5 or 6 years after college, financial aid, low-income, (any public (United States) 9th grade assistance for financial potentially first- institution) aid applications and generation students in college application high school Enrolment at 26 +12*** process (from 9th Grade) 2-year public (nearly half of Talent (N=34,346) institution Search participants Enrolment at 19 +8*** received 10 hours per 4-year public year of services or institution fewer) Indiana/ Idem Idem Enrolment 52 +4*** 4 or 5 years after 9th (N=10,927) (any) grade Enrolment at 13 +3*** 2-year institution Enrolment at 32 +3*** 4-year institution Florida/ Idem Idem Enrolment 36 +15** 4 or 5 years after 9th (N=14,721) (any public grade institution) Enrolment at 29 +10** 2-year public institution Enrolment at 9 +5** 4-year public institution     56    DiD+PSM Cunha, Miller & GO Center Project Texas/ A dedicated classroom Low-income high Enrolment 67a +3.5** Weisburst (United States) One year after high for the college school students in (any) (2018) school graduation application process with selected schools Enrolment at ? +1.8* a full-time counsellor (N=43,230) 2-year and active outreach run institution by selected student peers Enrolment at ? +2.2* 4-year institution DiD Stephan & College coach Chicago/ One coach per high Disadvantaged High Enrolment 53 +3* Rosenbaum program Fall after high school school to provide help in school seniors (any) (calculated (2013) (United States) graduation completion of FAFSA, (primarily African from OR) scholarship, and college American, Latino and Enrolment at 2 20 +1.3 applications low-income) year-institution (calculated (Over one year) (N=35,777) from OR) Enrolment at 24 +4.1** less selective (calculated 4-year from OR) institution vs. 2-year a: Refers to the whole control group, not specific to disadvantaged students. b: Own calculations based on interaction terms.    Estimates plotted in figure 2       57    Table B.4: the impact of "information, guidance and academic tutoring" outreach programs on access to higher education  Baseline in Evaluation Authors Intervention Location/ Details of intervention Disadvantaged group Estimated Outcome control group Design (Year) (Country) Time of evaluation (duration) (Sample size) effect (p.p.) (%) RCT Avery College Possible St Paul(MN)/ After-school curriculum High school students Enrolment 63.8 +1.7 (2013) Program Fall after high school with mostly of color with (any) (United States) graduation -Extensive tutoring with below median family test preparation services income and GPA > 2.0 & (from 11th grade) -College admissions and (N=238) Enrolment at 34.4 +15.1** financial aid consulting, 4-year guidance in the transition institution to college (320 hours over 2 years) RCT Myers et al. Upward Bound National sample of Vary but always Low -income or first- Enrolment 71 +3 (2004) program schools/ academic tutoring, generation high school (any) (United States) by 2 to 4 years after preparation for college students (from 9th or Enrolment at 24 -5 expected high school entrance exams, cultural 10th grade) 2-year graduation activities and (N=2,292) institution information on financial Enrolment at 44 +6** aid 4-year (average of 477 sessions institution attended over 21 months) RCT Seftor, Mamun Upward Bound National sample of Vary but always Low -income or first- Enrolment & Schirm program schools/ academic tutoring, generation high school 79.1 +1.5 (any) (2009) (United States) by 7 to 9 years after preparation for college students (from 9th or Enrolment at expected high school entrance exams, cultural 10th grade) 2-year 22.4 -2.9 graduation activities and (N=2,102) institution information on financial aid Enrolment at (average of 477 sessions 4-year 51.9 +1.3 attended over 21 months) institution       58    Table B.5: the impact of outreach programs on graduation in higher education  Baseline in Evaluation Authors Intervention Location/ Details of intervention Disadvantaged group Estimated Outcome control group Design (Year) (Country) Time of evaluation (duration) (Sample size) effect (p.p.) (%) RCT Ford et al Explore Your New Brunswick/ After-school project Low-income and first- Any post- 12.5 +1.2 (2014) Horizons program 4 years after high activities with enhanced generation high school secondary (Canada) school graduation career education and students-from 10th degree focused information on grade (by 4 years) post-secondary studies. (N=1,033) (40 hours over 3-year period) RCT Seftor, Mamun & Upward Bound National sample of Vary but always Low -income or first- Any post- 34.8 +2.26 Schirm program schools/ academic tutoring, generation high school secondary (2009) (United States) by 7 to 9 years after preparation for college students-from 9th or degree expected high school entrance exams, 10th grade Associate 9.1 -2.18 graduation cultural activities and (N=1,724) degree information on financial aid Bachelor's 21.6 0.14 (average of 477 degree academic and activity sessions attended over 21 months) PSM Constantine et al. Talent search Florida/ Information about Primarily targeting Associate 8 +5*** (2006) program by 4 years after end of college, financial aid, low-income, degree (United States) intervention assistance for financial potentially first- (by 8 years) aid applications and generation students in college application high school-from 9th process Grade (nearly half of Talent (N=14,721) Search participants received 10 hours per year of services or fewer)     59    DiD+PSM Cunha, Miller & GO Center Project Texas/ A dedicated classroom Low-income high Any post- 21.7a -1.5 Weisburst (United States) by 8 years after high for the college school students in secondary (2018) school graduation application process selected schools degree with a full-time (N=43,230) (by 8 years) counsellor and active Associate 7.5a -0.6 outreach run by degree selected student peers (by 8 years) Bachelor's 13a +0.8 degree (by 8 years) a: Refers to the whole control group, not specific to disadvantaged students.       60    Appendix C: Causal estimates on the effect of financial aid on disadvantaged students  Table C.1: The effect of universal financial aid on access to higher education  Disadvantaged Baseline in Evaluation Authors Program Estimated Details of program group Outcome control group Design (Year) (Country) effect (p.p.) (Sample size) (%) DiD Denning Community Discount in tuition fees in community colleges Economically Enrolment at (2017) College Tuition based on residency: Annexion of municipalities disadvantaged high community 27a +5.2***b Reductions, Texas making residents eligible for reduced tuition at a school graduates college (United States) community college (in-district tuition); community (N=204,448) Enrolment at 4- colleges in Texas charged 63 percent more, on year institution 25a -3.1b average, to out-of-district students relative to in- district students Black high school Enrolment at graduates community 27a +4.8***b (N=204,448) college Enrolment at 4- year institution 25a -3.4***b a: Refers to the whole control group, not specific to disadvantaged students. b: Own calculations based on interaction terms.       61    Table C.2: The effect of need‐based financial aid on access to higher education  Disadvantaged Baseline in Estimated Evaluation Authors Program Details of program group Outcome control effect Design (Year) (Country) (Sample size) group (%) (p.p.) RCT Azzolini et al ACHAB Dedicated savings account for high school students with Low-income Enrolment 67.1 +8.7*** (2018) experiment compulsory savings between 5-50€/month and deposits matched at high school (any) (Italy) a rate of 4 to 1. Maximum savings of €2,000 matched for a students (last 2 maximum of €8,000. Money could be spent only on education- years) related expenses (N=716) RCT Ford et al. New Annual grant of CAN$4,000 for maximum two years, with early Low-income Enrolment 38.6 +10.7*** (2014) Brunswick commitment (deposited while student is in high school and and first- (any) Learning provided conditional on high school completion) generation high Enrolment at 21.6 +9.8*** Accounts school college (short) (Canada) students-from 10th grade Enrolment at 17.9 +0.9 (N=1,145) university RCT Richburg- California One-time scholarship of $1,000 for enrolling in postsecondary Low-income Enrolment 84.4 +3.5 Hayes et al Cash for education high school (any) (2015) College seniors Enrolment at 2- 43.2 (CFC) (N=3,560) year institution +5.2* (United Enrolment at 4- 42.8 States) -2.9 year institution DiD Baumgartner BaFöG Increase in federal need-based aid by roughly 10 percent (on Low-income Enrolment at 64 +1.5 & Steiner (Germany) average 45€ more per month) high school university (2006) graduates (N=456) DiD Bettinger Ohio College Increase of about $750 in total grant aid Low-income Enrolment at 4- ? +1.5*** (2015) Opportunity first-year year institution Grant students in (United public States) institutions (N=83,259)     62    RD Castleman & Florida An additional $1,300 in grant aid (in 2000 dollars), yearly Low-income Enrolment 61 +3.2 Long Student renewable high school (any) (2013) Access Grant graduates Enrolment at 2- 34 +0.1 (United (N=6,917) year public States) institution Enrolment at 4- 26 +3.2* year public institution DiD Dearden, Maintenance Implementation of a grant of £960 on average (in 2006 prices) Low-income Enrolment 15.5 +3.8** Fitzsimmons, grants 18-19- year- (any) Wyness (United olds (2014) Kingdom) (N=11,286) RD+IV Denning, Maximum An additional $1,000 in first year grant aid due to eligibility to Lowest-income Enrolment at 4- 76 +0.4 Marx & Pell grants maximum Pell grant university year public Turner (United entrants institution (2017) States) (EFC=0) (N=36,697) DiD Dynarski Social Annual renewable grant of $6,700 on average (in 2000 dollars) High school Enrolment 35.2 +21.9* (2003) Security seniors with (any, by age 23) Student father deceased Benefit during Program childhood (United (more likely to States) be low-income and/or black) (N=3,986) RD Fack & Bourses sur Fee waiver for public university fees, averaging €174 per year for Low-income Enrolment 77.3 +0.3 Grenet Critères undergraduate students grant (any) (2015) Sociaux applicants (France) (N=50,388) Annual cash allowances of €1500, in addition to fee waivers Low-income Enrolment 78.6 +2.7*** grant (any) applicants (N=194,513) Enrolment in 73.4 +4.9*** 1st year (N=16, 467)     63    DiD Kane Federal Pell Annual renewable grant of maximum $3,544 (in 1991 dollars) Black 18-19- Enrolment ? -1.5 (1995) grant year-old (any) (United females Enrolment at 2- ? +1.2 States) (N=12,163) year public institution Lowest income Enrolment ? +0.5 quartile 18-19- (any) year-old Enrolment at 2- ? +2.4 females year public (N=12,163) institution DiD Linsenmeier Institutional University grant of about $4,000, replacing a loan of the same Admitted low- 51.9 +2.0 Institutional et al. grant, amount income enrolment (2006) replacing loan students (yield rate) (United (N=13,701) States) Admitted 47.1 +8.9 minority low- Institutional income enrolment students (yield rate) (N=3,523) DiD Lovenheim Ineligibility Ineligibility for federal financial aid due to HEA98 for up to two Convicted drug Enrolment 35.8 -22** & Owens of federal years offenders (any, by two (2014) financial aid (majority of years) (United disadvantaged Enrolment 40.1 -8 States) males) (any, ever (N=7, 401) enrolled) RD Rubin Federal Pell Pell grant around the eligibility threshold (average $400) Low-income Enrolment 86a -1.35 (2011) grant high school (any, on-time) (logit (United graduates estimates) States) a: Refers to the whole control group, not specific to disadvantaged students.    Estimates plotted in figure 3       64    Table C.3: The effect of merit‐based financial aid on access to higher education  Disadvantaged Baseline in Evaluation Authors Program Estimated Details of program group Outcome control group Design (Year) (Country) effect (p.p.) (Sample size) (%) RD Bruce & HOPE- Annual grant of max $3.000 (for 2-year colleges) Enrolment (any) 85.9a -0.0 Carruthers scholarship, or max $6.000 (for 4-year colleges) to cover (2014) Tennessee tuition Pell-grant eligible Enrolment at 2- 28.5a -2.9** (United States) -Students must submit FAFSA to receive HOPE high school year public (but do not have to be eligible) graduates institution -Eligibility with near-average high school GPA (N=17,145) Enrolment at 4- 42.3a +2.4** and ACT scores year public institution Enrolment (any) 85.9a -2.6** Non-white high Enrolment at 2- 28.5a -2.8* school graduates year public (N=10,609) institution Enrolment at 4- 42.3a +1.1 year public institution RD Cohodes & Adams Between $910-$1714 in annual renewable tuition Non-white high Enrolment at 4- 71.6a +6.3*** Goodman Scholarship, aid (roughly a 20% reduction in costs) school seniors year institution (2014) Massachusetts -Not need-based (N=88,152) (United States) -Eligibility with top 25% score in own school district in 10th grade (MCAS test) Low-income high Enrolment at 4- 71.6a +3.7** school seniors year institution (N=88,152)     65    DiD Dynarski HOPE- Tuition and fee waiver, averaging $1900 per year Low-income 18-19- Enrolment (any) 30a -1.4 (2000) scholarship, but amount offset by other aids received (not year-olds Georgia cumulative with Pell) (N=3,380) (United States) -Not need-based but application differs by parental income (easier for middle/high-income) -Eligibility with at least a 3.0 GPA (B) in high Black 18-19-year- Enrolment (any) 30a -2.7 school graduation olds -Renewable conditional on maintaining a 3.0 GPA (N=1,837) (B) in college RD Kane Cal Grant, Fee subsidy of maximum $9,036 - $9,420 per year 17-20-year-old Enrolment (any) ~87 +4.2** (2003) California -Need-based: income and assets below specific grant low-income (United States) limits applicants - Minimum high school GPA around 3.1 (N=5,558) DiD Sjoquist & State-wide merit Strong merit aid - defined as not having too Non-White or Enrolment (any) 63.5a -1.99*** Winters aid programs, restrictive eligibility requirements and providing Hispanic men (2015) (United States) relatively large awards Non-White or Enrolment (any) 63.5a -0.97 Hispanic Women RD Vergolini, Trento 5B grant Annual grant of €1,200-€4,800 Low-income, high University ~70a +6.5 Zanini & (Italy) -Need-based performing students enrolment Bazoli -Final grade in high school above 93/100 (N=5,535) (2014) a: Refers to the whole control group, not specific to disadvantaged students.    Estimates plotted in figure 3       66    Table C.4: The effect of performance‐based financial aid on access to higher education  Disadvantaged Baseline in Evaluation Authors Program Estimated Details of program group Outcome control group Design (Year) (Country) effect (p.p.) (Sample size) (%) RCT Barrow et Opening Doors Additional grant in first year of enrolment of Low-income parents Enrolment at 76.7 +5.3* al. Louisiana $1,000 per semester, conditional on: accepted in institution (2- (2014) (United States) -being enrolled for at least 6 credits community colleges year) after -maintaining a C GPA. (N=1,019) drop/add period RCT Binder et VISTA at Additional grant of $1,000 per semester for 4 Low-income Enrolment at 99.4 -1.3 al. University of New consecutive semesters, conditional on: incoming freshmen institution (4- (2015) Mexico -being enrolled in at least 12 credit hours in 1st (N=1,081) year) (United States) semester, and 15 credit hours in subsequent semesters -Maintaining a GPA of 2.0 (C) or higher -Meeting with advisers at least twice per semester RCT Richburg- California CFC- Additional grants ranging from $1,000 to $4,000, Enrolment (any) 84.4 +4.9*** Hayes, et PBS for one semester or up to 2 years, conditional on: al. (United States) -Enrolment Low-income high Enrolment at 2- 43.2 (2015) -Completion of at least 6 credit hours per semester school seniors year institution +4.7*** -Maintaining a "C" average GPA or higher (N=4,642) Enrolment at 4- 42.8 year institution 0 DiD Jackson Texas Advanced Financial incentives for teachers and students Low-income Enrolment (any, ? +5.0* (2010) Placement based on scores in advanced placement courses in students in minority in Texas) (percent Incentive Program high school: Students receive between $100 and high schools increase) (APIP) $500 for each eligible course conditional on a (226 schools) (United States) score of 3 or above       67    Table C.5: The effect of loans on access to higher education  Disadvantaged Baseline in Evaluation Authors Program Estimated Details of program group Outcome control group Design (Year) (Country) effect (p.p.) (Sample size) (%) RD Solis National loan National loan programs covering tuition costs with Students taking the Enrolment (any) 13.3 +20*** (2013) programs interest rates ranging from 2% to 6%, conditional college admission (Chile) on: test in the lowest - Being in one of the four poorest income income quintile quintiles; (N=84,605) - Score at least 475 points in the national college admission test (PSU test) RD+IV Gurgand, Eduloan Short-term loans to cover tuition fees for students Admitted applicants Enrolment at 44.3 +41.9* Lorenceau (South Africa) admitted in a public university (have to be repaid to public public university & Melonio during the studies) universities with income below first quartile (N=1,397) a: Refers to the whole control group, not specific to disadvantaged students. b: Own calculations based on interaction terms.       68      Table C.6: The effect of tax credit incentives on access to higher education  Disadvantaged Baseline in Evaluation Authors Program Estimated Details of program group Outcome control group Design (Year) (Country) effect (p.p.) (Sample size) (%) IV Bulman & American AOTC allowed tax-payers to deduct yearly up to Low-income 19- Enrolment (any) ~32 No effect Hoxby Opportunity Tax $2,500 for up to four years of higher education. year-olds (2015) Credit (AOTC) AOTC is partly refundable: a taxpayer who owes (United States) zero taxes can receive a check of up to $1,000. Fixed- LaLumia Hope Tax Credit HTC allowed tax-payers to deduct yearly up to Non-white men, Enrolment (any) 3.4a +2.0 effect IV (2012) (HTC); $1,500 of college expenses for up to 2 years; aged 33-50 Lifetime Learning LLTC allowed tax-payers to deduct yearly up to Non-white women, Enrolment (any) 6.7a +1.1b Tax Credit $2,000 of college expenses an unlimited period of aged 33-50 (LLTC); time; Tuition and Fees TD allowed tax-payers to deduct up to $4,000 of Parents had no Enrolment (any) 3.4a +0.9 Deduction (TD) college expenses from adjusted gross income; college, men aged (United States) 33-50 Parents had no Enrolment (any) 6.7a -1.7b college, women aged 33-50 a: Refers to the whole control group, not specific to disadvantaged students. b: Own calculations based on interaction terms.       69    Table C.7: The effect of universal financial aid on graduation in higher education  Disadvantaged Baseline in Evaluation Authors Program Estimated Details of program group Outcome control group Design (Year) (Country) effect (p.p.) (Sample size) (%) DiD Denning Community Discount in tuition fees in community colleges Economically Associate degree (2015) College Tuition based on residency: Annexion of municipalities disadvantaged high (by 4 years) Reductions, Texas making residents eligible for reduced tuition at a school graduates 4.1a +0.3b (United States) community college (in-district tuition); community (N=204,448) colleges in Texas charged 63 percent more, on average, to out-of-district students relative to in- Black high school Associate degree district students graduates (by 4 years) (N=204,448) 4.1a +0.9**b a: Refers to the whole control group, not specific to disadvantaged students. b: Own calculations based on interaction terms.   Table C.8: The effect of need‐based financial aid on graduation in higher education  Disadvantaged Baseline in Evaluation Authors Program Estimated Details of program group Outcome control group Design (Year) (Country) effect (p.p.) (Sample size) (%) RCT Anderson Wisconsin Annual grant, complementing Pell grant, of $1,800 Low-income 2-year Associate degree 30 -1 & Scholars Grant and renewable for up to five years freshmen (by 3 years) Goldrick- (United States) (N=3,153) Rab (2016) RCT Ford et al. New Brunswick Annual grant of CAN$4,000 for maximum two Low-income and Any 12.5 +9.1*** (2014) Learning years, with early commitment (deposited while first-generation high postsecondary Accounts student is in high school and provided conditional school students- degree (Canada) on high school completion) from 10th grade (by 4 years) (N=1,145)     70    RCT Goldrick- Wisconsin Annual grant, complementing Pell grant, of $3,500 Low-income 4-year Bachelor's degree 16.3 +4.7** Rab et al. Scholars Grant and renewable for up to five years freshmen (by 4 years, on- (2016) (United States) (N=1,500) time) IV Alon Any federal, state An additional $1,000 in annual grant aid Black freshmen in Bachelor's degree 76 +3.2b*** (2007) or college grant private and most (by 6 years) (United States) selective universities (N=15,196) Hispanic freshmen Bachelor's degree 83 +3.2b*** in private and most (by 6 years) selective universities (N=15,196) IV Alon Any need-based Each additional $100 received in the first year University students Bachelor's degree 48 +0.6** (2011) grant in the lowest- (by 6 years) (United States) income quartile (N=1,937) RD Castleman Florida Student An additional $1,300 in grant aid (in 2000 dollars), Low-income high Associate degree 17 -0.3 & Long Access Grant yearly renewable school graduates (by 5 years) (2013) (United States) (N=6,917) Bachelor's degree 25 +5.2** (by 7 years) RD Clotfelter, Carolina Need-based grant covering the financial costs of Low-income Bachelor's degree 76 +2.2 Hemelt & Covenant college attendance through a mix of grant and students admitted to (by 4 years) Ladd (United States) work-study awards a public flagship (2018) university (N=1,133) RD+IV Denning, Maximum Pell An additional $1,000 in first year grant aid due to Lowest-income Bachelor's degree 43 +5.7* Marx & grants eligibility to maximum Pell grant university entrants (by 7 year) Turner (United States) (EFC=0) (2017) (N=17,109) RD Denning Any financial aid Increase in financial aid (on average + $374 in Low-income (Pell Bachelor's degree 71.2 +0,9 (2018) (United States) grants) associated with being declared financially recipients) students (by 5 year) independent in 4th year of bachelor's program (N=33,844) 71    RD Fack & Bourses sur Annual cash allowances of €1500, in addition to Bachelor's degree 25.5 +2.1 Low-income grant Grenet Critères Sociaux fee waivers (by 3 years, on- applicant entering (2015) (France) time) the first year of a bachelor's degree (N=10,951) Bachelor's degree 58.7 +2.9*** Low-income grant (same year) applicants entering the final year of a bachelor's degree (N=40,789) DiD Lovenheim Ineligibility of Ineligibility for federal financial aid due to HEA98 Convicted drug Bachelor's degree 7.4 -7.2 & Owens federal financial for up to two years offenders (majority graduation (2014) aid due to HEA98 of disadvantaged (United States) males) (N=7,401) RD Turner & GI Bill Renewable tuition subsidy of $500 + monthly Black war veterans Any 6 +2.7 Bound (United States) stipend of up to $120 (1984$) for World War II postsecondary (2003) veterans degree a: Refers to the whole control group, not specific to disadvantaged students. b: Own calculations based on interaction terms.       72    Table C.9: The effect of merit‐based financial aid on graduation in higher education  Disadvantaged Baseline in Evaluation Authors Program Estimated Details of program group Outcome control group Design (Year) (Country) effect (p.p.) (Sample size) (%) RD Carruthers HOPE- Loss of hope scholarship after first year in college College freshmen Any ? +1.4 & Ozek scholarship, because of GPA below the threshold for renewal. with family income postsecondary (2016) Tennessee Annual grant up to $4,000 (in 4-year institutions) below median degree (United States) and up to $2,000 (in 2-year institutions), (N=7,248) (on-time) conditional on near-average high school GPA and ACT scores and maintaining a 2.75 or 3.0 GPA in college RD Cohodes & Adams Between $910-$1714 in annual renewable tuition Non-white high Bachelor's degree 43.3a -2.4 Goodman Scholarship, aid (roughly a 20% reduction in costs) school seniors (by 4 years, on- (2014) Massachusetts -Not need-based (N=88,152) time) (United States) -Eligibility with top 25% score in 10th grade (MCAS test) Low-income high Bachelor's degree 43.3a -1.5 school seniors (by 4 years, on- (N=88,152) time) DiD Sjoquist & State-wide merit Strong merit aid - defined as not having too Associate degree 38.8a +0.66 Winters aid programs restrictive eligibility requirements and providing Non-White or or higher (2015) (United States) relatively large awards Hispanic men Bachelor's degree 30a -0.4 or higher Associate degree 38.8a -0.45 Non-White or or higher Hispanic women Bachelor's degree 30a 0.23 or higher RD Welch HOPE- In 2005, Annual grant up to $1,500 per year at a Community college Associate degree 6.6a -0.4 (2014) scholarship, community college and up to $3,000 in 4-year freshmen with (by 3 years) Tennessee institutions, renewable for up to five years, family income (United States) conditional on: below median -near-average high school GPA (3.0) and ACT (N=10,639) Bachelor's degree 7.2a -3.8 scores (21) (by 5 years) -Maintaining a 2.75 or 3.0 GPA in college a: Refers to the whole control group, not specific to disadvantaged students. b: Own calculations based on interaction terms. 73    Table C.10: The effect of performance‐based financial aid on graduation in higher education  Disadvantaged Baseline in Evaluation Authors Program Estimated Details of program group Outcome control group Design (Year) (Country) effect (p.p.) (Sample size) (%) RCT Binder, VISTA at Additional grant of $1,000 per semester for 4 Low-income Bachelor's degree 33.2 +4.5 Krause, University of consecutive semesters, conditional on: incoming freshmen (by 5 years) Miller & New Mexico -being enrolled in at least 12 credit hours in 1st (N=1,081) Cerna (United States) semester, and 15 credit hours in subsequent (2015) semesters -Maintaining a GPA of 2.0 (C) or higher -Meeting with advisers at least twice per semester RCT Mayer, Ohio Additional grant of $900 per semester, or $600 per Low-income parents Any 32.9 +1.6 Patel & Performance- quarter, up to a maximum of $1800, conditional in community postsecondary Gutierrez Based Scholarship on: colleges degree (2015) Program -Achieving a “C” or better in 12 or more credits (N=2,285) (by 4 years) (United States) -or a part-time award of $450 per semester/$300 per quarter for achieving a “C” or better in 6 to 11 credits       74    Table C.11: The effect of loans on graduation in higher education  Disadvantaged Baseline in Evaluation Authors Program Estimated Details of program group Outcome control group Design (Year) (Country) effect (p.p.) (Sample size) (%) IV Alon Any federal, state An additional $1,000 in annual loan aid Black freshmen in Bachelor's degree 88a +0.2b (2007) or college loan private and most (by 6 years) (United States) selective universities (N=15,196) Hispanic freshmen Bachelor's degree 88a -1.1b in private and most (by 6 years) selective universities (N=15,196) IV Wiederspan Federal loan, Federal loan receipt Low-income Associate degree 9 +20 (2016) Texas community college (by 3 years) (United States) students (N=132,147) Black low-income Associate degree 5 +16.4 community college (by 3 years) students (N=84,793) IV Dunlop Federal Stafford An extra $100 in total loan High-need Associate degree 21a +0.3 (2013) loans community college (by 5 years) (United States) students (N=2,037) Black community Associate degree 21a +1.0 college students (by 5 years) (N=437) a: Refers to the whole control group, not specific to disadvantaged students. b: Own calculations based on interaction terms.       75    Table C.12: The effect of tax credits on graduation in higher education  Disadvantaged Baseline in Evaluation Authors Program Estimated Details of program group Outcome control group Design (Year) (Country) effect (p.p.) (Sample size) (%) PSM Elsayed Hope Tax Credit HTC allowed tax-payers to deduct yearly up to Black college Any 41.6a +9.7*** (2016) (HTC); $2,200 of college expenses for up to 2 years; students who had postsecondary Lifetime Learning LLTC allowed tax-payers to deduct yearly up to applied to financial degree Tax Credit $2,000 of college expenses an unlimited period of aid (by 6 years) (LLTC); time; (N=4,850) Tuition and Fees TD allowed tax-payers to deduct up to $4,000 of Deduction (TD) college expenses from adjusted gross income (United States) a: Refers to the whole control group, not specific to disadvantaged students.       76    Appendix D: Causal estimates on the effect of mixed interventions on disadvantaged students  Table D.1: The effect of mixed interventions on access to higher education  Disadvantaged Baseline in Evaluation Authors Program Estimated Details of program group Outcome control group Design (Year) (Country) effect (p.p.) (Sample size) (%) RCT Ford et al Expand Your -40 hours of after-school project activities with Low-income and Enrolment (any) 37.8 +10.5*** (2014) Horizons + enhanced career education and focused first-generation high Learning information on post-secondary studies over a 3- school students- Accounts in New year period from 10th grade Enrolment at 21.6 +5.1* Brunswick - Annual grant of CAN$4,000 for maximum two (N=1,148) college (short) (Canada) years, with early commitment (deposited while student is in high school and provided conditional Enrolment at 16.5 +6.9*** on high school completion) university RCT Hahn, Quantum -250 hours of education Low-income high Enrolment (any) 16 +26**** Leavitt & Opportunities -250 hours of developmental activities school students- Aaron Program -250 hours of service each year from 9th grade to from 9th grade (1994) (United States) high school graduation. (N=158) -$1.00 - $1.33 per hour for participating and a grant amounting total earnings for postsecondary enrolment RCT Rodriguez- Quantum -250 hours of education Low-income high Enrolment (any 55.8 +7.4** Planas Opportunities -250 hours of developmental activities school students- postsecondary) (2012) Program -250 hours of service each year from 9th grade to from 9th grade (United States) high school graduation. (N=791) Enrolment at 2- 37.7 +4.3 -$1.00 - $1.33 per hour for participating and a grant amounting total earnings for postsecondary year or 4-year enrolment institution     77    DiD Andrew, Longhorn LOS: Combination of outreach in disadvantaged High-achieving Enrolment in 2.7 +2*** Imberman Opportunity high schools, financial aid ($4,000 per year) and minority & low- targeted flagship & Scholars (LOS) & academic tutoring during college in University of income high school university (UT) Lovenheim Century Scholars Texas seniors (eligible (2016) (CS) programs FRL) (United States) (N=15,835) CS: combination of outreach in disadvantaged High-achieving Enrolment in 4.3 +0.2 high schools, financial aid ($5,000 per year for minority & targeted flagship four years) and support service during college in disadvantaged high university Texas A&M University school seniors (TAMU) (N=21,327) PSM Carruthers Knox Achieves, Combination a college coaching (outreach) and Lowest-income high Enrolment (any) 47.8a +25.7*** & Fox Tennessee financial aid program, covering the gap between school seniors (2016) (United States) the direct cost of enrollment and aid from other (eligible FRL) Enrolment at 2- 23a +25.2*** sources, offered to students for making a (N=5,197) year institution seamless, immediate transition between high school and one of the state’s public community Enrolment at 4- 29.7a +3* colleges year institution DiD + Oreopoulos, Pathways to Comprehensive program that included counseling, Low-income high Enrolment (any) 33.6 +19.2*** Matching Brown & Education academic support, social support and financial school students- Lavecchia (Canada) support. from 9th grade Enrolment at 11.9 +9.8*** (2014) Site 1: Regent’s college (short) Park Enrolment at 21.6 +9.4*** (N=1,274) university Low-income high Enrolment (any) 40.7 +4.4 school students- from 9th grade Enrolment at 14.3 +4.6 Site 2: Rexdale college (short) (N=737) Enrolment at 26.4 -0.3 university RD Page, Dell Scholars Combination of financial support (up to $20,000 High-achieving low- Enrolment at 4- 81.2 +2.8 Castleman Program of scholarship) and individualized advising, both income high school year institution & (United States) at college entrance and throughout the duration of seniors Sahadewo postsecondary enrollment (N=2,040) (2016) a: Refers to the whole control group, not specific to disadvantaged students.     78    Table D.2: The effect of mixed interventions on graduation in higher education  Disadvantaged Baseline in Evaluation Authors Program Estimated Details of program group Outcome control group Design (Year) (Country) effect (p.p.) (Sample size) (%) RCT Ford et al Expand Your -40 hours of after-school project activities with Low-income and Any 12.6 +8.0*** (2014) Horizons (EYH) enhanced career education and focused first-generation high postsecondary + Learning information on post-secondary studies over a 3- school students- degree Accounts (LA) in year period from 10th grade (by 4 years) New Brunswick - Annual grant of CAN$4,000 for maximum two (N=1,148) (Canada) years, with early commitment (deposited while student is in high school and conditional on high school completion) RCT Rodriguez- Quantum -250 hours of education Low-income high Bachelor's or 7.1 -0.3 Planas Opportunities -250 hours of developmental activities school students- associate degree (2012) Program -250 hours of service each year from 9th grade to from 9th grade (at age 25) (United States) high school graduation. (N=791) -$1.00 - $1.33 per hour for participating and a Bachelor's degree 2.0 +1.1 grant amounting total earnings for postsecondary (at age 25) enrolment RCT Scrivener Accelerated Study Combination of counselling, tutoring, special Low-income Associate degree 21.8 +18.3*** et al (2015) in Associate courses, and financial support (tuition waiver, community college (by 3 years) Programs, New MetroCard and free textbooks) based on a full- freshmen York time enrolment requirement (N=896) (United States) DiD Andrew, Longhorn LOS: Combination of outreach in disadvantaged High-achieving Graduation from 2.0 +1.5*** Imberman Opportunity high schools, financial aid ($4,000 per year) and minority & low- targeted flagship & Scholars (LOS) & academic tutoring during college in University of income high school university (UT) Lovenheim Century Scholars Texas seniors (eligible (by 6 years) (2016) (CS) programs FRL) (United States) (N=15,835) CS: combination of outreach in disadvantaged High-achieving Graduation from 3.2 -0.0 high schools, financial aid ($5,000 per year for minority & targeted flagship four years) and support service during college in disadvantaged high university Texas A&M University school seniors (TAMU) (N=21,327) (by 6 years) 79    RD Clotfelter, Carolina Combination of need-based grant covering the Low-income Bachelor's degree 82 +4.7 Hemelt & Covenant financial costs of college attendance – through a students admitted to (by 4 years) Ladd (United States) mix of grant and work-study awards – and a public flagship (2018) additional support services, such as mentoring by university faculty and peers, career advice, professional (N=1,838) development opportunities, and social events RD Page, Dell Scholars Combination of financial support (up to $20,000 of High-achieving low- Bachelor's degree 60.5 +19.2* Castleman Program scholarship) and individualized advising, both at income high school (by 6 years) & (United States) college entrance and throughout the duration of seniors Sahadewo postsecondary enrollment (N=337) (2016) a: Refers to the whole control group, not specific to disadvantaged students.       80      81