Ex-ante socio-economic impact assessment for a social science research infrastructure: The case of EuroCohort

The aim of this paper is to present the ex ante Socio-Economic Impact Assessment of a European Social Science Research Infrastructure (RI) named EuroCohort that will provide, over the next 34 years, a longitudinal study of the well-being of children and young people across Europe

As the single European birth cohort longitudinal survey, EuroCohort will complement information provided by other existing European longitudinal surveys like the European Social Survey (ESS), the Survey of Health, Ageing and Retirement in Europe (SHARE), and the Generation & Gender Programme (GPP) collecting detailed and accurate life history data from European children and young adults.EuroCohort has already been the topic of the FP7 Measuring Youth Well-Being (MYWeB) project,2 which has provided the proof of concept for the development of a Europe-wide longitudinal survey of child and youth well-being, and of the Horizon 2020 European Cohort Development Project (ECDP) which has created a specification and business case for this European RI. 3 A new European Project related to the preliminary phase of EuroCohort, the H2020 Coordinate Project, has started in April 2021.
There is a large literature showing the policy relevance of children's well-being for society as a whole.The promotion of children's well-being is important not only for children to enjoy a good childhood, but also since it may increase their future well-being over time.For instance, the accumulation of early skills by children has a direct correlation to the development of more advanced skills (Cunha & Heckman, 2007).In this respect, policy interventions and social investments in this area may also be economically efficient (Heckman & Tim, 2014).
In 2019 the OECD published a Reference Framework for Assessing the Scientific and Socio-Economic Impact of Research Infrastructures (OECD, 2019).This framework draws on the experience of a number of existing research infrastructures, including several that are directly relevant to EuroCohort (among them there were the ESS and the Consortium of European Social Science Data Archives (CESSDA)).Moreover, there is a small but growing literature on the socioeconomic impact of research infrastructures (see, for instance, Kounduri et al., 2014;EC, 2019b;Catalano et al., 2021). 4In this paper, we take stock of these works in order to evaluate the ex-ante socio-economic impact of a social science RI such EuroCohort.
Finally, it is important to specify that the importance of a European Longitudinal survey on the well-being of children and young adults like EuroCohort has been enhanced by the COVID-19 pandemic because the data that it will supply may contribute to the design of evidence-based social policies aimed at reducing the long-term consequences of the pandemic on the well-being of children and young adults.In this regard, it should be stressed that among new and serious challenges faced by for the economies and societies of all high-income countries, "childrenas dependants-are among those at greatest risk of seeing their living standards fall and their personal well-being decline" (Richardson et al., 2020).
The remaining part of this paper is organized as follows.The next section contains a short literature review.Section 3 presents a general description of EuroCohort.Section 4 is devoted to the socio-economic impact evaluation of this longitudinal survey.Section 5 presents a riskassessment exercise.Section 6 concludes.

LITERATURE REVIEW
In this section we are going to present a short literature review of different projects and published works addressing socio-economic evaluations or cost-benefit analysis (CBA) of policies aimed at improving children and young people's well-being.In general, following Vining and Weimer (2010), it is useful to point out that the application of economic evaluation methods to social policy raises a number of issues.Among those, the main ones are: the need for a comprehensive approach (taking account of all valued effects in predicting net benefits) approach to assessing social policies; the need to recognize and explicitly address the great uncertainty in prediction and valuation involved in applying economic evaluations in most social policy areas; the need to consider those behaviors, which occur frequently in social policy, that do not satisfy the assumptions of neoclassical welfare economics, and the application to policies that often have strong distributional goals and consequences. 5os et al. ( 2004) provide a literature review on prevention and early intervention programs for youth evaluation conducted, generally in the United States, since 1970.The conclusions are mixed.Some programs gave taxpayers a good return for their money.However, some others failed to generate more benefits than costs.This source evaluates the benefit-cost ratio of a wide set of prevention and early intervention programs for youth and concludes that the value of this indicator ranges from 0 to over 100.It is worth noticing that 23 out of 61 programs examined had a benefit-cost ratio less than 1, while the value of this indicator for all but four of these programs did not exceed 28.Barnett and Masse (2007) present the results of a CBA conducted for the Abecedarian program, which offered educational experiences of up to 10 h/day (approximately 250 days per year) for children from early in the first year of life until they entered kindergarden.Under this program a group of 111 disadvantaged North Carolina children born between 1972 and 1977 were randomly assigned as infants to either the early educational intervention group or the control group.Those results were obtained from a randomized trial with longitudinal follow-up through to the age of 21.They show that an important benefit of the Abecedarian program was the labor market success of participants' mothers.According to Barnett and Masse (2007), the Internal Rate of Return of the Abecedarian program exceeds 7%.Karoly (2016) reviews the CBA results for several US preschool programs.Among those programs the one displaying the lowest benefit-cost ratio is the Oklahoma's universal preschool program (Tulsa program) serving children one year before entering kindergarten.The benefit-cost ratios for both the full-day and the part-day programs are presented.Furthermore, for each of the above values of this indicator for children in each of three following income groups (freelunch students, reduced-price lunch students and full-price lunch students) are provided.These benefit-cost ratios range from 4.08 for part-day free lunch students to 2.82 for full-day full-price lunch students, which represent the lowest value of this indicator among the programs evaluated by this study.Schweinhart (2016) takes into consideration several US Early Childhood Care and Education (ECCE) programs and, coherently with Aos et al. (2004), concludes that only high-quality ECCE programs display a return of investment which is large enough to allow us to affirm that they are socially desirable from the point of view of the efficient use of the available resources.Saminder and Gillian (2016) provide some estimates of the monetary value of benefits and costs generated by the services supplied by the Children's Centre Services in England. 6The services they supply include childcare and early education programs, health services, parenting classes and specialized family support services.According to this source, the benefit-cost ratio of those services ranges from 0.5 to 6.5.Cannon et al. (2018) consider 115 US early childhood programs related to children or parents of children from the prenatal period to age 5.This study also discusses programs such as home visiting, parent education, government transfers providing cash and in-kind benefits.The findings show that most of the programs considered have positive effects on at least one child outcome and those with an economic evaluation tend to show positive economic returns.
Obviously, the above conclusions about improving the efficiency of the expenditure on children and young people's well-being in the US by switching resources cannot be applied directly to EU member states because of the differences between the US and European countries' social conditions.However, the results presented above, particularly the ones presented by Aos et al. (2004) and Schweinhart (2016) and Cannon et al. (2018) suggest that most likely there is room for improving this efficiency also within the EU.In this respect, Doyle (2020) presents a very interesting analysis of the impact of investment in parenting from pregnancy to formal schooling in a disadvantaged community in Dublin, Ireland.The results of that paper show the positive and substantial impact of investment in parenting from pregnancy to formal schooling on children's cognitive, social and behavioral development and constitute an important benchmark for future research in this area in other European countries.
EuroCohort is a RI that will provide longitudinal survey data to be used for better targeting social policies.In this respect, survey data are of benefit only to the extent that they are used in the policy process and affect change in policy, government expenditure and individual programs that might contribute to children and young people's well-being. 7he impact of a longitudinal study on policy may depend upon several variables.One of them is the format in which survey data are presented.Technopolis Group ( 2017) is a study assessing the policy, academic and teaching impacts that have been achieved through the ESS.According to this source, data presented in formats that are comprehensible and can be easily shared and used will have a higher probability of being used by policymakers than data presented in a less userfriendly form.The preparation of a tabular volume of ESS data for the Austrian ministry of social affairs is an example of translation of ESS data that forms a pipeline to a major use of the data. 8oreover, it is possible to observe that there is no single, straightforward or commonly agreed method for estimating government expenditure on children and young people's well-being.This reflects both the debates on the nature and conceptualization of well-being and the challenges of cross-country comparison of government social spending.It is beyond the remit of this paper to address the methodological and empirical debate in this area.Rather, we draw on the substantive work undertaken by the OECD and on data from the OECD and from Eurostat in this field to provide (see Appendix) an estimate of social expenditure in the 21 EU Member States (Austria, Belgium, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Luxembourg, Netherlands, Poland, Portugal, Slovak Republic, Slovenia, Spain, Sweden and United Kingdom) which were also OECD Member States in 2013. 9We also use the available data to provide a rough bottom line estimate of total spending on children and young people's well-being for the remaining EU Member States (Bulgaria, Croatia, Cyprus, Latvia, Lithuania, Malta and Romania).This evaluation can provide a benchmark of the scale of the improvement in the effectiveness of this expenditure that EuroCohort could generate.
Our literature review shows that it is possible to improve the efficiency of the expenditure on children and young people's well-being by switching resources from lower efficiency to higher efficiency programs.According to the sources quoted above, the benefit-cost ratio of the different child and young adult intervention programs is highly volatile.Therefore, there is plenty of room for potentially improving the efficiency of the expenditure related to children and young people's well-being using evidence-based policies informed by the data provided by Eurocohort.

EUROCOHORT
A short description of the main organizational aspects of EuroCohort is provided below.We consider here that all 28 European countries (27 European Union Member States and the UK) will participate in EuroCohort.

The EuroCohort's RI and its longitudinal design
EuroCohort will have a central research hub which will be responsible for and manage the functioning of the survey.This survey hub will be located in a single institution.In addition, Euro-Cohort will develop an online portal where users can access information including, for instance, survey data, methodological reports and published research.We also consider that each country will have a national coordinating team.Those 28 national coordinating teams will manage the fieldwork (via data collection agencies) and will co-operate with the central team in other national level issues.In particular, each national team shall guarantee, among others, organizational capabilities such as technical and scientific expertise (for a wider discussion see Wilska & Salmela-Aro, 2019).Moreover, the micro-data collected at the national level from children and young people and their families/carers may to be linked with national administrative datasets to provide further opportunities for researchers and policy makers in different policy areas (e.g., education, health).
Figure 1 illustrates SHARE's governance arrangements which represent a relevant example for the future governance model of EuroCohort.
As for the longitudinal design of this survey, it is planned to follow two cohorts, called respectively Age 8 cohort and Age 0 cohort (see Figure 2).

Sampling
Following the analysis presented in the ECDP project, we define a large country as a country with a population above 10 million people, and a small country as a country with a population the topic of the second part of the FP7 MYWeB project, which has provided the proof of concept for the development of a Europe-wide longitudinal survey of child and youth well-being.Furthermore, from January 2018 to December 2019 EuroCohort has also been the topic of the Horizon 2020 European Cohort Development Project (ECDP), which has created a specification and business case for this European RI.
below 10 million people (see the relevant table in the Appendix). 11Each large country starts with an age 0 cohort effective sample size of 10,000 children (this corresponds to a nominal sample size of 14,300), while each small country starts with an age 0 cohort effective sample size of 5,000 children (7,200 nominal sample size).For the cohort of 8-year-olds, each large country starts with an age 8 cohort effective sample size of 8,000 children (11,400 nominal sample size), while each small country starts with an age 8 cohort effective sample size of 4,000 (5,700 nominal sample size) (Lynn, 2019).12This survey will be based on Computer-Assisted Personal Interviewing (CAPI) at wave 1 for both cohorts.From wave 2 on the survey will be based on a Computer-Assisted Web Interviewing (CAWI)-CAPI mixed-mode for both cohorts.We consider that at wave 2 we will have 50% CAWI and 50% CAPI, while in the next two waves the percentage of CAWI will increase by 10% at each wave until reaching 70%, and it will remain constant thereafter.
As it is customary in these cases, before the beginning of the very first wave of interviews Euro-Cohort will undertake a series of piloting phases from pre-pilot cognitive interviewing through to a full-scale dress rehearsal in each country for each cohort.This will allow for fine-tuning of the functioning of the research instruments and field processes before the start of each cohort fieldwork wave.
With regard to piloting, we consider a 250 interviews per country pilot and a full-scale dress rehearsal before each cohort fieldwork wave.Dress rehearsal size will be equal to 400 interviews in small countries and 600 interviews in large ones.Pilot and dress rehearsal interviews will follow the same rules stated above for the wave they refer to and will take place the year before it (see Babarović et al., 2019).The costs of piloting and dress rehearsal reported in the preparation phase will be reduced during the implementation and operation phases, given that the cognitive interviewing and pilot targeted sample sizes remains the same but the dress rehearsal sample will be subject to attrition.

Non-response and attrition rates
Following the analysis presented in the ECDP project, we consider a response rate of 75% for the second wave of each cohort, an attrition rate of 15% for the third wave and 10% for the fourth wave.
After the fourth wave we consider a 5% attrition rate between each wave of the survey.Response and attrition rates will vary among countries.At this stage we are not able to credibly estimate these variations.Therefore, we consider here that all countries have the same response rate for the first wave of each cohort and the same attrition rate thereafter.
A way to enhance response rates may be to use monetary incentives.The use of incentives is to be decided at the country level, because both the need for them and their effect may depend upon country-specific factors.Therefore, at this stage we are not in the position to estimate either the amount of these incentives or the extent to which they will be awarded.However, in the analysis 11 According to this classification, Portugal belongs to the group of large countries, while Hungary, with a population of 9.8 million, should be considered a small country.However, this classification could be refined by considering that the size of the age group 0-24 in Hungary (2,489,938 people) is almost the size of the same age group also in Portugal (2,517,097 people).Therefore, as an exception from the general rule stated above, we decided to also include Hungary into the group of large EU countries.
conducted below we consider to award uniform monetary incentives to all participants across Europe for indicative purposes.The values of these incentives utilized in our analysis should be considered as average values of the real figures.We already know that it may be necessary to monitor the extent to which monetary incentives disproportionately encourage the participation of low-income people and thereby have an effect on the composition of the sample (see Stoop et al., 2016).

EUROCOHORT SOCIO-ECONOMIC IMPACT ANALYSIS
To assess the social desirability of EuroCohort from the point of view of the efficient use of the available resources it is possible to take as a starting point the framework for the evaluation of RIs presented by Florio and Sirtori (2016).However, these authors focus on capital intensive RIs, while surveys like EuroCohort are RIs more labor than capital intensive.Therefore, in this paper we will need to adapt their framework to the socio-economic evaluation of a survey, following, in particular, the recent European funded research projects which have analysed the socio-economic impact of research infrastructures (see, for instance, Kounduri et al., 2014;Beagrie & Houghton, 2016;EC, 2019b).

Distribution of benefits
The first step to evaluating the benefits generated by a project is to identify the set of potential beneficiaries.In the case under scrutiny this set includes: • scientists and researchers producing knowledge; • young professionals, junior researchers and students spending time working within the RI; • children and young people receiving a benefit from an improvement in the effectiveness of government expenditure and individual programs that might contribute to improve their wellbeing.It looks very likely that this group will enjoy a relevant share of the benefits generated by EuroCohort; • policymakers getting the chance to develop more efficient and/or effective policies aimed at improving children and young people's well-being.This can allow them to save scarce public resources for different uses.
Therefore, below we consider that the benefits arising from any actual or possible use of a longitudinal survey RI can be of four main types: • Use value and efficiency gains for researchers • Knowledge output • Human capital accumulation • Benefits provided to end-users (policy makers, children and young people).

Use value and efficiency gains for researchers, practitioners and policymakers
To assess the benefits for researchers generated by EuroCohort we can start by noticing that if we consider the market for scientists as perfectly competitive, the opportunity cost of their time can be assumed as being equal to their average hourly compensation.
According to the European Commission ( 2007) the average yearly salary of an EU researcher in 2006 was almost €38k.That corresponds to about € 2018 €46.6k.To estimate the average hourly wage of an EU researcher it is possible to divide this yearly salary for 1,720 hours per year of work, obtaining €22.10 per hour, which in 2018 was equivalent to €27.10 per hour.Pollock et al. (2020) estimates that the number of EU28 potential academic users with an interest in EuroCohort findings is 17,000 using as a starting point the UK's 2014 Research Excellence Framework data on the number of academics in UK institutions whose publications were externally peer reviewed by discipline, weighting each discipline 13 according to the proportion of its academics that will benefit from EuroCohort, and extrapolating the results obtained to the EU28 countries.Taking SHARE as a reference point, they also estimate that each year from 500 to 700 of those potential academic users will actually log in to EuroCohort, with approximatively 2,000 downloads.
The same algorithm also allows Pollock et al. (2020) to estimate that in the UK 43.2% of those academic users will be staff, while the remaining 56.8% will be PhD students.To obtain a conservative estimate of the hourly cost of a PhD student we consider as a floor for this cost the minimum yearly cost of an Italian PhD student, which is set by national law at €18.8k, and divide this cost for the same 1,720 hours per year of work we used to calculate the hourly wage rate for the average EU researcher, obtaining €11 per hour.From the above, assuming that the proportion between staff and PhD students among potential EU28 academic users is equal to the one of the UK calculated above, it is possible to estimate the average hourly opportunity cost of the time of an academic user of EuroCohort as €18, which corresponds to the weighted average of the hourly wages of a researcher and a PhD student.
Following the approach envisaged by Beagrie and Houghton (2016) for the European BioInformatics Institute (EMBL-EBI), the most direct measure of the value of EuroCohort to the academic community (use value) is the time (and therefore cost) users will spend accessing its database (access time) and obtaining the desired data.To quantify the monetary value of the access time of EuroCohort we timed 2,000 accesses per year by an estimated average access time of 1 hour and an average hourly opportunity cost of an academic user of EuroCohort of €18, obtaining a yearly value of the access time of EuroCohort to the academic community of €36k.
A second component of the use value of EuroCohort to the academic community is the amount of time users will spend working with data from EuroCohort (use time).To quantify this use time, following Florio et al. (2016) it is possible to consider that the average EU researcher devotes a 60% share of his or her time, corresponding to around 1,000 hours per year, to research activities.Considering that the average academic user of EuroCohort will spend 1% of his or her research time, corresponding to 10 hours per year, working with EuroCohort data, and an average of 600 academic users of EuroCohort per year, it is possible to obtain an estimated total use time of 13 To perform this exercise, those disciplines where grouped as follows: (1) Psychology, Psychiatry and Neuroscience EuroCohort data of 6,000 hours per year.Timing these 6,000 hours of EuroCohort use time per year by the average hourly opportunity cost of an academic user of EuroCohort of €18 it is possible to obtain a yearly use time value of EuroCohort of €108k.
Following once again the approach envisaged by Beagrie and Houghton (2016), we can consider that the availability of EuroCohort data will enable researchers to reduce the amount of time spent working with children and young people's well-being data by 50% with respect to what would be required in order to perform the same task using administrative data.To quantify this efficiency gain in monetary terms, we consider that without EuroCohort, 300 academic users would have allocated 20 hours per year working with children and young people's well-being data reaching the same productivity that they are able to reach working 10 hours per year with EuroCohort data.Therefore, the existence of EuroCohort data enables those researchers to save 3,000 hours per year of their working time and to use that time to do additional research which brings additional returns.Timing these 3,000 hours of saved research time by the average hourly opportunity cost of an academic user of EuroCohort of €18 we obtained a yearly efficiency gain generated by EuroCohort of €54k.Pollock et al. (2020) estimate that the number of EU28 individual policymakers with an interest in EuroCohort findings is 6,300 by extrapolating from the estimated number of UK's people with a direct interest in policy who are also potential beneficiaries of EuroCohort findings. 14Following the same approach we have just used for researchers, we can consider that the availability of EuroCohort data will enable these policymakers to reduce the amount of time spent working with children and young people's well-being data by 50% with respect to what would be required in order to perform the same activity using administrative data.To quantify this efficiency gain in monetary terms, we consider that without EuroCohort these 6,300 policymakers would have allocated 10 hours per year of their time to work with children and young people's well-being data, and, by doing so, they would be able to reach the same productivity that they are able to reach working 5 hours per year with EuroCohort data.Therefore, the existence of EuroCohort data enables those policymakers to save cumulatively 31.5khours per year of their working time and to use that time to do additional work which brings additional returns.
Quantifying in monetary terms the value of this time is a task beyond our means. 15To overcome this problem, we decided to use as a proxy for the cost of policymakers the cost of senior researchers (€150k per year in high-cost countries, €75k in medium-cost and €37.5 in low-cost ones). 16Dividing these amounts for 1,720 hours per year of work it is possible to obtain a labor cost for policymakers equal to €87.2 per hour in high-cost countries, €43.6 in medium-cost ones and €21.8 in low-cost ones.Timing these amounts by the estimated number of hours of work saved by policymakers in each country will give a total value of the efficiency gain by policymakers generated by EuroCohort equal to €1.7m.
Finally, Pollock et al. (2020) identify at the European level around 390 child advocacy organizations that may utilize the EuroCohort data.Considering an average of 3 people for each organization using EuroCohort data, that will result in 1,170 child advocacy organization staff members.
Following the same approach we have just used for policymakers, we can consider that the availability of EuroCohort data will enable these child advocacy organization staff members to reduce the amount of time spent working with children and young people's well-being data by 50% with respect to what would be required in order to perform the same activity using administrative data.To quantify this efficiency gain in monetary terms, we consider that without EuroCohort, these 1,170 child advocacy organization staff members would have allocated 20 hours per year of their time to work with children and young people's well-being data, and, by doing so, they would be able to reach the same productivity that they are able to reach working 10 hours per year with EuroCohort data.Therefore, the existence of EuroCohort data enables those child advocacy organization staff members to save cumulatively 11,700 hours per year of their working time and to use that time to do additional work which brings additional returns.To quantify in monetary terms the value of this time, we decided to use as a proxy for the cost of child advocacy organization staff members the median EU hourly net earnings for a single person without children in 2015 calculated from the data reported in the Appendix and inflated to 2018 (€9.82).Multiplying this amount by the estimated number of hours of work saved by child advocacy organization staff members in each country will give a total value of the efficiency gain by this group generated by EuroCohort equal to €115k.

Knowledge output
A very imperfect measure of knowledge output is given by publications and presentations at conferences.The value of a publication VP can be calculated as where VNI = social value of producing new information; VI = social value of the degree of influence of this publication on the scientific community.
If we consider the opportunity cost of the time of researchers as being equal to their average hourly compensation, a reasonable proxy of VNI is its marginal production cost.Following Florio et al. (2016), it is possible to assume that the average EU researcher devotes a 60% share of his or her time to research activities with an average yearly productivity of three papers per year.Under these assumptions, and considering, as detailed above, an average salary of this EU researcher equal to €46.6k, the marginal production cost of an academic paper can be roughly estimated at €9k.Pollock et al. (2020) indicate that the average annual number of publications generated by SHARE results is around 120.According to this source, because of the analogies existing between SHARE and EuroCohort we can take this number as a reasonable predictor for the average number of publications which will be generated by the availability of the results of EuroCohort.
The results of a survey conducted among the users of the European BioInformatics Institute database (2016) show that 45% of respondents declared that they could not have obtained the last EMBL-EBI data they used elsewhere, nor could have they created or collected themselves.Furthermore, they also show that the majority of respondents experienced efficiency time savings by working with these data.Based on these results, below we consider that 50% of the above publications (60 papers) could not have been produced without EuroCohort, because the data they use could not have been created or collected elsewhere by their authors.We also assume that the remaining 50% of those publications (60 papers) would have been produced even without EuroCohort.However, without the availability of EuroCohort data their production costs would have been twice as high because of the loss of efficiency involved in the use of data created or collected elsewhere.
Therefore, here we consider those 60 publications per year which will be produced thanks to EuroCohort as additional research, so their value can be considered as a net benefit generated by EuroCohort itself.Timing those 60 publications by the €9k social value of each publication estimated above we obtained a monetary evaluation of this net benefit being equal to €540k.
The number of citations a paper will obtain can be used as a proxy for its degree of influence on the scientific community.The shadow price of a citation could be estimated as the opportunity cost of the time employed by a scientist to read and understand somebody else's paper and to decide to quote it.After a mutual cancellation of RI scientists' papers and scientific labor costs, we can conclude that the benefit generated by the knowledge output can be measured by the sum of the value of citations that RI scientists' papers receive and the value of the subsequent waves of citations.
According to Times Higher Education data quoted by Florio and Sirtori (2016) the average number of citations for a Social Sciences journal article is 4.67.17According to Florio et al. (2016) the average time needed to evaluate someone else's paper and to decide to cite it is one hour.Timing the above figure by the additional 60 publications per year which will be produced thanks to EuroCohort results it is possible to estimate the social value of the degree of influence of those publications in about 280 hours of research work.Taking into consideration once again the €27.1 average hourly wage of an EU researcher estimated above, it is possible to obtain a monetary evaluation of the degree of influence of EuroCohort additionally produced output on the scientific community at an amount close to €8k.

Human capital accumulation
EuroCohort will contribute to human capital accumulation in several different ways.On the one hand, students and young scientists who will spend a period working within EuroCohort will accumulate higher human capital relative to their peers.This human capital will take the form of both technical and scientific abilities (hard skills) and personal ones, like communication, negotiating and organizational capabilities (soft skills).On the other hand, the EuroCohort database will supply an important resource to the academic community both because it will be used for creating new teaching materials and because it will be used directly for guided learning and independent dissertation work.Both uses will generate further human capital accumulation by students who will use these materials for their classes, and doctoral candidates who will use the EuroCohort database for dissertation-related purposes.The present value of human capital accumulation private benefits produced by the RI is the sum of the expected increase in lifelong salary that each of the above beneficiaries will earn over her career compared with the without-the-project scenario.To get an understanding of the magnitude of this benefit from the above it is possible to estimate the number of PhD students which will use the EuroCohort database in 350 units per year, corresponding to the 58% of the 600 projected average academic users of EuroCohort (Pollock et al., 2020).
To supply a very rough estimate of the expected increase in the salaries of those PhD students due to their involvement in EuroCohort, it is possible to consider that Catalano et al. (2021a, b) TA B L E 1 Yearly benefits generated by EuroCohort

Benefit Amount (€)
Child advocacy organizations staff members' time 115k Policymakers' time 1.7m Researchers' access time 8k Researchers' use time 54k Social value of producing new information 540k Social value of the degree of influence of this publication on the scientific community 8k Human Capital Accumulation 400k TOTAL 2.8m estimate that performing research and acquiring skills at CERN could lead to a salary premium effect which can vary between 5% and 13% more than the normal return to education in the relevant field.Taking into account that EuroCohort will be a social science, non-material RI, to err on the side of caution, we can halve the lower limit of this interval and assume that performing research and acquiring skills within the EuroCohort framework will lead to a salary premium of 2.5% more than the normal return to education. 18 As reported above, we can consider the average yearly salary of an EU researcher equal to €46.6k, so we can quantify the yearly salary premium that each of the 350 PhD students involved with EuroCohort will command in €1.17k.Timing this amount by the 350 PhD students involved with EuroCohort each year it is possible to estimate the total value of the human capital accumulation generated by EuroCohort in €400k per year.
From the previous analysis, it is then possible to conclude that the projected benefits of Euro-Cohort can be considered as corresponding to €2.8m per year.The table below summarizes our findings.
However, we need to consider that there will be a lag between the timing of the survey's costs and the benefits.This lag will arise because EuroCohort will start generating costs from the very beginning of its operations, while benefits will not materialize for several years.The reasons for the latter are twofold: • The information content of a longitudinal survey grows with each wave of interviews.Therefore, it may need several years to reach the threshold level at which it will be able to start being used by the scientific community in a significant way and to influence the decision-making process.
• The decision-making process can have a relevant inside lag.Once EuroCohort has reached the information threshold level necessary to influence the decision-making process, it may take time until actual decisions informed by this longitudinal survey will be taken.
In order to account for the above, we consider a lag of six years between the beginning of EuroCohort's fieldwork and the time when its benefits will start to materialize (that is from 2033 onwards). 18Human capital accumulation will be favored by services like webinair, master classes, Summer schools, policy briefs (as consultants), and new research EuroCohort expects to be able to offer to interested users, like PhD students.

General remarks
To be able to provide more realistic estimates of the costs in each country, we have divided the 28 countries into high, medium and low-cost states.The results of this exercise can be found in the Appendix.This categorization was primarily based on Eurostat data about annual net earnings for a single person without children across Europe.However, PPP adjusted GDP per capita was also included in our algorithm.Considering the unweighted country average of annual net earnings for a single person without children it is possible to see that the value of this indicator for mediumcost countries is about 50% lower than that for high-cost states, and, in turn, the value of the same indicator for low-cost states is 50% lower than that for medium-cost states.Therefore, our national level cost estimates will follow these proportionate reductions and we consider that, in mediumcost states, costs will be 50% lower than in high-cost states, and, in turn, costs in low-cost states will be 50% lower than in medium-cost states.The complete description of this categorization algorithm can be found in the Appendix.

Pilot, interviews and incentives
To estimate the average cost of a CAPI interview in a high-cost country we started from the distribution of the average costs per interview in the ESS's third wave of interviews (ESS3) contained in Stoop et al. (2010).Based on this source, we obtained an ESS3 average cost of a CAPI interview in a high-cost country equal to €220.Then we converted this figure in € 2018 using the Harmonized Index of Consumer Prices provided by Eurostat for the EU28 countries, obtaining a € 2018 value of the average cost of a CAPI interview in a high-cost country of about €247.We also consider that since 2010 the real unit cost of an interview has increased by €50.As a result, below we consider that in a high-cost country the cost of each CAPI interview with a parent is €300, and that the cost of interviewing a child under 16 is €150.We consider that the cost of interviewing young people over 16 rises to €300.
As for national team costs, we consider a 50% reduction in interviews cost per sample member in medium-cost states and a further 50% reduction in cost in low-cost ones.
For the different waves we consider the following: • Waves 1-2-3 age 0 cohort: an interview with one parent.The cost per sample member is €300.
• Waves 4-5-6 age 0 cohort: interviews with one parent and the child so we will have two interviews per child per wave.The cost per sample member is €450.• Waves 7-8-9 age 0 cohort: interviews with the young person so we will have one interview per young person per wave.The cost per sample member is €300. Similarly: • Waves 1-2-3 age 8 cohort (children aged 8-9 to 14-15): interviews with one parent and the child so we will have two interviews per child per wave.The cost per sample member is €450.
• Waves 4-5-6 age 8 cohort (young people aged 17-18 to 23-24): interviews with the young person so we will have one interview per young person per wave.The cost per sample member is €300.
On the basis of the discussion contained in Villar and Fitzgerald (2017) and taking a precautionary approach, at this stage we decided to quantify the average cost of a CAWI interview as being 50% of the cost of a CAPI one.
With regards to monetary incentives we assume that in every country a €10 cash incentive will be awarded to parents (both for a two-parent or a single-parent household) which will be sampled for participating in the survey until the child reaches 8 years old.This will include the first three waves for the age 0 cohort.In the next three waves of both cohorts, thus until the child will be 14-15 years old, a €10 cash incentive will be awarded to parents and a further €5 one directly to the child.In the last three waves, so when the young adult will be 17-18 and older, a €10 cash incentive will be given directly to him.The above figures are to be considered as average incentives.Most likely, the actual incentives which will be awarded to sample members will be differentiated among countries and maybe also among sample members according to a scheme which at this stage has not been envisaged yet.

Implementation and operation phases
The best available way to estimate the cost of the EuroCohort central team is probably to consider the costs related to a similar European accelerated longitudinal survey.In that respect, SHARE central team costs approximately €3m per year for international coordination.19However, these costs are rising (in 2015 they were about €2.4m).20 Therefore, we will consider here that during the implementation and the operation phases the EuroCohort Central Hub will cost €4m per year, due also to the greater complexity of sampling of EuroCohort.
Taking as a starting point SHARE's financial data, we consider that in the sampling years of the implementation phase the national team will cost around €1.15m per year in high-cost states.This amount will cover both the National Node costs21 (€250k) and the fixed costs of the national agency (€900k.) We also consider that the cost of national agencies in a sampling year of the operation phase will be reduced to 70% of the full cost starting in the first year of the operation phase with further reductions of 5% every five years (that is 65%, 60%, 55% and 50% of the full cost).
This reduction in the costs of national agencies is plausibly justified in relation to ongoing work which examines the ways in which the collection of survey data can be made more sustainable through a combination of the use of new technologies and changing patterns of human behavior (Coupar et al., 2018) as well as increasing levels of cooperation between international survey Research Infrastructures (Emery et al., 2019).Specifically, these cost reductions will be linked to: a. the benefits of learning economies, starting from the end of the implementation phase; b. the likely increase in the competitiveness of the market of European agencies in the aftermath of the COVID-19 pandemic; c. the introduction of 5G technology and its applications in the functioning (sampling etc.) of longitudinal surveys and their management.This allows us to think that most likely we will be able to negotiate, in the Preparatory phase of EuroCohort, a significant reduction of the yearly fixed cost of national agencies with respect to what we budgeted in the design phase of EuroCohort.
In the years when there is no sampling we consider a cost of national agencies of 10% of the full cost, whereas in the years when we have piloting or dress rehearsal we consider a 30% of the full cost, across both the implementation and operation phases.

Preparation phase
Across the preparation phase, the cost of the Central Hub will rise from 5% of its €4m full cost at the start of the phase until 25% of its €4m full cost at the end of the phase.In the first year of the preparation phase (2022) we estimate the Central Hub costing 5% of the above amount (€200k) which will support a central team to coordinate the project infrastructure.This will rise from the second to the fourth year (2023 until 2025 included) to 10% (€400k) to give more intensive support to national fund raising and to testing in five countries, then to 25% (€1m) in the final year (2026) when the pilot and dress rehearsal will take place in each of the 28 countries.Based on these estimates, the total cost of the Central Hub in the preparation phase will be €2.4m.
In the last year of the preparation phase when the piloting and dress rehearsals take place, the National Nodes will cost 25% of their full operating costs, that is €62,500 in high-cost countries, €31,250 in medium-cost countries, and €15,625 in low-cost countries.In the first year (2022), those costs will be reduced to 5% of the full operating costs (12.5k, 6.25k, and 3.125k respectively).During the following three years (from 2023 to 2025 included) those costs will be reduced to 10% of the full operating costs (25k, 12.5k, and 6.2k respectively).
During the preparation phase the project is also scheduled to conduct cognitive interviewing in five countries: Croatia, Finland, France, Ireland and the United Kingdom.As for the above definition of the level of costs, Croatia is a low-cost country, while Finland, France, Ireland and United Kingdom are high-cost countries.We consider that in each of these five countries the project will stage 60 pre-pilot cognitive CAPI interviews during the second year of the preparation phase (2023).

Termination costs
The termination costs of EuroCohort will materialize in 2054 and 2055.During this phase there will be no interview costs.During the termination phase, we assume that the Central Hub cost will reduce by 50% (€2m) in the first year ( 2054) and by a further 50% (€1m) in the second year (2055), before the project ends (adding to a total of €3m).Moreover, the National Nodes cost will reduce to 10% of their operating costs, to assist with the national decommissioning of the project.The total cost is estimated to be €850k.Adding these two figures, the total cost of the termination phase is estimated to be €3.85mwhich when discounted at 4%, corresponds in present value terms to €1.4m in 2021.

The total cost of EuroCohort
The financial total cost of EuroCohort during its whole 34 years projected lifespan, obtained as detailed above, is summarized in the table below.Here we added to the above the assumption that 80% of the whole cost of EuroCohort will be met by public funds while the remaining 20% will be met by private ones.Under this assumption EuroCohort would need to raise about €160m of private funds over its 34 years life span.For the sake of completeness we need to notice that raising €160m of private funds will generate fundraising costs that, according to Rodriguez (2018), could amount to €24m.The figures displayed below are undiscounted values expressed in € 2018 (Table 2).Figure 3 shows the curve of the undiscounted cost of EuroCohort throughout its lifecycle.It is common for research infrastructures such as surveys to build up slowly in the beginning, then arriving at the highest levels of investment costs during implementation phase.The shape of the curve over time reflects the annual variations in costs due to the years in which there is intensive fieldwork taking place, producing this typical wave-like structure.Besides the Central Team's financial total cost of €113.4m,we have produced Member State level financial cost estimates, which are displayed in the Appendix, where it is also possible to find Member State level yearly financial cost estimates during the operation of EuroCohort.

Results
The evaluation of benefits for users and beneficiaries and the analysis of the costs of the RI (following the methodologies presented in StRESFRI (ESFRI, 2019) conducted in the previous sections supports the idea that EuroCohort is a socially desirable RI from the point of view of the efficient use of the available resources.To summarize, improvements in the effectiveness of European countries' expenditure related to children and young people well-being (due to the availability of new longitudinal survey data provided by EuroCohort) of a measure of around 1 over 15,000 would be sufficient for the benefits of such a survey to outweigh its costs.
As we have already seen in our literature review, the conclusions of Aos et al. (2004), Schweinhart (2016) and Cannon et al. (2018) show that in the US the efficiency of expenditure on children and young people's well-being could be improved by switching resources from low-return to highreturn well-being expenditure programs.Obviously, the above conclusions about improving the efficiency of the expenditure on children and young people's well-being in the US by switching resources between projects cannot be applied directly to European countries because of the differences between the US and the 28 European countries' social conditions.However, the recent work by Doyle (2020) is a very interesting example of important results obtained by social policies targeted to children's well-being.
To complete this discussion, it should also be remembered that, according to StRESFRI (2019), two economic effects of a RI are decommissioning costs and its residual value.In the case of EuroCohort, decommissioning costs will relate to the need to archive the huge dataset that will be produced by this survey, while its residual value will be the value that this dataset will still have for the decision makers, the scientific community and the general public.We have not considered these effects in our work, leaving them for further research.
Finally, it is worth noticing that the social expenditures considered above in the OECD database will have effects on the health of their beneficiaries, because child well-being outcomes are clearly interconnected, and this aspect is of the utmost relevance in consideration of the long-term consequences of the Covid-19 pandemic.For instance, education and personal health are clearly correlated.As Barnett and Masse (2007) assessment of the Abecedarian program puts it: Better-educated individuals can make more informed and better decisions regarding their personal health.Education increases the ability to be an effective consumer of health care services and producer of personal health.Education also increases income, allowing one to buy higher quality and quantity of health services and to establish healthier living conditions.
According to the same Barnett and Masse (2007), a proof of this ability to make better personal health decisions can be found in the fact that a follow-up regarding the program participants at age 21 showed that the rates of smoking for the control group and the early educational intervention group were 55% and 39%, respectively.Campbell et al. (2014) confirm the interconnection between education and personal health showing that according to biomedical data, children randomly assigned to the Abecedarian program intervention group had significantly lower prevalence of risk factors for cardiovascular and metabolic diseases once they reach their mid-30s compared to the ones assigned to the control group.

RISK ASSESSMENT
Undoubtedly, the ex-ante socio-economic evaluation of a project that will be running until 2055, like EuroCohort, involves a significant amount of uncertainty.To deal with this uncertainty, the last part of our work presents a risk assessment.The first step of this assessment was to identify the critical variables of the project, defined as those variables which would have the largest impact on the project's costs.On the cost side, this identification was achieved through a sensitivity analysis.
The variables we considered in our sensitivity analysis were: • the cost of a CAPI interview in a high-cost country; • the fixed costs of the national agency in the sampling years of the implementation phase in high-cost countries; • the full operating costs of the National Nodes in high-cost countries; • the cost of the Central Hub during the implementation and operation phases.
The reason for our choice was that almost all of the other monetary variables contained in our analysis were expressed as a percentage of these ones.
For each of the above variables we tested the effect of a variation of ±1% of the value adopted in the base case on the projected undiscounted total cost of EuroCohort.As is possible to see in the table below, our results show that the variation on the above parameters generated by a variation of ±1% of the cost of a CAPI interview in a high-cost country is about twice as much as that generated by an equal percentage variation of the fixed costs of the national agency in the sampling years of the implementation phase in high-cost countries.In turn, the latter is about 25% higher than the variation of the same parameters of the project generated by a ±1% variation of the full operating costs of the National Nodes in high-cost countries or by the same percentage variation of the cost of the Central Hub during the implementation and operation phases (Table 3).
Based on the results of our sensitivity analysis, we conducted a probabilistic risk analysis running a 10,000 run Monte Carlo simulation considering as critical variables for EuroCohort only the cost of a CAPI interview in a high-cost country and the fixed costs of the national agency in the sampling years of the implementation phase in high-cost countries.For each run, values of the two parameters were stochastically drawn from the probability distributions described below.For our exercise we considered the financial cost of a CAPI interview in a high-cost country as being normally distributed with a mean equal to €300 and a standard deviation equal to €30, and the fixed costs of the national agency in the sampling years of the implementation phase in highcost countries as being normally distributed with a mean equal to €900k and a standard deviation equal to €90k.
Our results show that the undiscounted total cost of EuroCohort will be between €710m and €954m, with an average value of €839.7m.The probability distribution of this indicator resulting from our Monte Carlo simulation is displayed in the figure below.This distribution suggests that the probability that the undiscounted total cost of EuroCohort, calculated as specified above, will be between €700m and €750m is 2.2%, while the probability that it will be between €750m and €800m is 22.1%, the probability that it will be between €800m and €850m is 45.7%, the probability that it will be between €850m and €900m is 24.7%, the probability that it will be between €900m and €950m is 5.2%, and the probability that it will be between €950m and €1.000m is 0.1%.Moreover, the probability that the undiscounted total cost of EuroCohort will be higher than €850m (29.9%) is higher than the probability that it will be lower than €800m (24.3%) (Figures 4 and 5).
The cumulative distribution of our results shows that their distribution is skewed to the right.Therefore, conditionally on the realisation of uncertainty, it is possible that the total cost of the project will be higher than the value calculated above.
As we did for the costs, we conducted a second probabilistic risk analysis for the benefits generated by EuroCohort running another 10,000 run Monte Carlo simulation considering as variables the cost of policymakers in a high-cost country and the number of additional publications produced thanks to EuroCohort.For each run, values of the two parameters were stochastically drawn from the probability distributions described below.
For our exercise we considered the value of time of policymakers in a high-cost country as being normally distributed with a mean equal to €150,000 and a standard deviation equal to €15,000, and the number of additional publications produced thanks to EuroCohort as being also normally distributed with a mean equal to 60 and a standard deviation equal to 10.Our results show that yearly benefits of EuroCohort (as defined in Table 1) will be between €2.2m and €2.5m, with an average value of €2,847,021.The probability distribution of this indicator resulting from our Monte Carlo simulation is displayed in the figure below.This distribution suggests that the probability that the projected yearly benefits of EuroCohort, calculated as specified above, will be lower than €2.4m is 0.6%, while the probability that they will be between €2.4m and €2.6m is 8.3%, the probability that they will be between €2.6m and €2.8m is 33.6%, the probability that they will be between €2.8m and €3m is 35.8%, the probability that they will be between €3m and €3.2m is 18.2%, and the probability that they will be between €3.2m and €3.4m is 3.5%.Moreover, from the above it is possible to see that the probability that the projected yearly benefits of EuroCohort will be higher than €3m (21.7%) widely exceeds the probability that they will be lower than €2.6m (8.9%).

CONCLUSIONS
In a nutshell, the economic evaluation which is presented in this paper suggests that improvements in the effectiveness of European countries' expenditure related to children and young people well-being (due to the availability of new longitudinal survey data provided by EuroCohort) of a measure of around 1 over 15,000 would be sufficient for the benefits of such a survey to outweigh its costs.
The above result emphasises the importance of investing resources (private and public) in order to achieve a better understanding of children and young adults' well-being as a reference for policy makers and various stakeholders (academic, practitioners and so forth), both at the European and national level.
Two concluding comments related to this result.First, the improvements in effectiveness of expenditure related to children and young people well-being can be related to a better targeting of beneficiaries of public programs.The relevance of heterogeneity of beneficiaries of these programs for the design and evaluation of public policies has been discussed, for example, by Kristensen et al. (2017) and Doyle (2020).Second, the impact of Covid-19 on future social policies across Europe is obviously still very difficult to ascertain (see Thunström et al., 2020 for a preliminary analysis).
However, the importance of data about children and young adults' well-being to design better health policies represent one additional element of potential benefits which has not been formally addressed in our paper (see also Richardson et al., 2020).
Finally, two additional directions for further research with regard to the socio-economic impact of Eurocohort.First, it would be useful to complete a full user survey to elicit willingness to pay for EuroCohort (see, for instance, Catalano et al., 2018).Second, the human capital accumulation and existence value benefits generated by a capital intensive, physical RI have been examined by a growing literature (see Florio, 2019).However, to our knowledge the above effects generated by a labor-intensive social science RI have never been investigated.Therefore, to complement the analysis presented in this paper, further research is needed to quantify the human capital accumulation and existence value benefits generated by a social science RI like EuroCohort.

A C K N O W L E D G M E N T S
We would like to thank the editor, a co-editor and two anonymous referees, Toni Babarovic, Catherine Blundell, Victor Cebotari, Daniela Cocchi, Marcel Das, Orla Doyle, Tom Emery, Rory Fitzgerald, Massimo Florio, Luigi Franzoni, Christopher Fox, Francesco Giovinazzi, Kristina Gogic, Robert Grimm, Johanna Jussila Hammes, Gundi  -education; -childcare; -cash benefits and tax breaks; -other benefits in-kind.
According to OECD family database public spending on the family can be categorised as follows.Child-related cash transfers to families with children, which includes: -child allowances, which are sometimes income-tested, and with payment levels that in some countries vary with the age or number of children; -public income support payments during periods of parental leave; -income support for single parent families (in some countries).Public spending on services for families with children, which includes: -the direct financing or subsidisation of childcare and early childhood education facilities; -public childcare support through earmarked payments to parents; -public spending on assistance for young people and residential facilities; -public spending on family services, including centre-based facilities and home help services for families in need.
Financial support for families provided through the tax system which includes: evant shortcoming of this approach, because health represents an important dimension of well-being. 25o obtain data on total spending on children and young people's well-being the OECD per capita expenditure by age data for 2013 has been combined with Eurostat population data by age for the same year.This provides a broad estimate of the scale of national spending on public services for children and young people by the 21 EU Member States which were also OECD members in 2013.The expenditure included in this analysis relates to welfare spending (cash benefits, tax breaks, and other benefits in kind) and education spending (childcare and compulsory schooling).It should be stressed that the underlying spending data are aggregated by OECD from the individual national budgets and there are a number of significant limitations around these data (the limitations of this approach are presented in OECD, 2009).The total estimated 2013 public spending on children and young people's well-being by the 21 EU Member States which were also OECD members in 2013 which we obtained as described above is about 1.25 trillion USD PPP.
We have also produced a rough bottom line estimate of total spending on children and young people's well-being for EU Member States which were not OECD members in 2013 (Bulgaria, Croatia, Cyprus, Latvia, Lithuania, Malta and Romania) by multiplying the number of residents of these countries belonging to each cohort from 0 to 25 by the corresponding 2013 per capita spending on children and young people's well-being of Greece, which in that year was the OECD EU Member State displaying the lowest overall level of this indicator.The total estimated 2013 public spending on children and young people's well-being by the EU Member States which were not OECD members in 2013 which we obtained as described above is about 46.3bn USD PPP.The estimates presented above suggest a combined expenditure of around 1.3 trillion in 2013 USD PPP for the 28 EU Member States.At 2013 PPP, 26 this would amount to around €950b.(Table A1-A4) the combined spending on children and young people's well-being of the 28 EU Member States using the USD PPP for the EU as a whole.
operation phase from January 2033 to December 2053 (21 years) • termination phase from January 2054 to December 2055 (2 years). 1010 In point of fact, the real project's timeline starts before 2022 because, as we have already noticed in the introduction, in the last few years EuroCohort has been the object of an extensive design phase.In 2015-16 this survey was already F I G U R E 1 Share Governance Arrangements.Source: http://www.share-project.org/organisation/share-eric.html[Colour figure can be viewed at wileyonlinelibrary.com]F I G U R E 2 Longitudinal design of EuroCohort [Colour figure can be viewed at wileyonlinelibrary.com] (weight: 0.05); (2) Economics and Econometrics (weight: 0.05); (3) Social Work and Social Policy (weight: 0.1); (4) Sociology (weight: 0.1); (5) Education (weight: 0.2); (6) Sports and Exercise Sciences, Leisure and Tourism (weight: 0.01); (7) Communicarion, Cultural and Media Studies, Library and Information Management (weight: 0.01).

F
I G U R E 3 Annual undiscounted costs of EuroCohort [Colour figure can be viewed at wileyonlinelibrary.com]

TA B L E 3
Percentage effect of a ±1% variation in the most relevant cost variables on the total cost of EuroCohort Variable % cost of a CAPI interview in a high cost country; ±0.42 fixed cost of the national agency in the sampling years of the implementation phase in high cost countries; ±0.19 full operating costs of the national nodes in high cost countries; ±0.14 cost of the Central Hub during the implementation and operation phases ±0.14

F
I G U R E 4 Distribution of the undiscounted total cost of EuroCohort according to the Monte Carlo analysis (million €) [Colour figure can be viewed at wileyonlinelibrary.com]

F
I G U R E 5 Cumulative distribution of the undiscounted total cost of EuroCohort according to the Monte Carlo analysis (million €) [Colour figure can be viewed at wileyonlinelibrary.com]F I G U R E 6 Distribution of yearly benefits generated by EuroCohort according to the Monte Carlo analysis (million €) [Colour figure can be viewed at wileyonlinelibrary.com]

Country PPP adjusted GDP per capita 2017 (€) 1 annual net earnings for a single person without children in 2015 (€ 100 % average worker) 2 Cost level
Knies, Phoebe Koundouri, Peter Lynn,  Chiara Monfardini, Philip O'Connel, Gary Pollock, Dominic Richardson, Paula Sargeant, Jennifer  Symonds, Bindu Sunny, Aleksandra Szymczyk, Maria Theofilatou, AlbertoZanardi and participants to the 2019 Bologna ECDP workshop and the 2019 European Meeting of the Benefit-Costs Society for useful comments and observations.Financial support from the European Union's Horizon 2020 research and innovation program, INFRADEV-01-2017 (European Cohort Development Project, ECDP) under grant agreement no.777449 is gratefully acknowledged.All remaining errors and imprecisions are our own.High, medium and low-cost countries for delivering a survey TA B L E A 122The OECD family database also provides an estimate of total spending on children and young people's in 2015 for Austria, Czech Republic, Finland, France, Germany, Italy, Spain and United Kingdom.However, to ensure a better homogeneity across the data set we have chosen to not to use these data.TA B L E A 2 Size of countries.source: our calculation on eurostat data Projected country level average financial cost per phase during EuroCohort (€) TA B L E A 31 Including fundraising costs.

Finland France Germany Greece Hungary Ireland Italy Luxembourg Netherlands Poland Portugal Slovak Republic Slovenia Spain Sweden
Estimated per capita public spending by age of children and young people for OECD EU member states in 2013 (USD PPP).Source: OECD TA B L E A 4