ObjectivesStudy substance use among second-generation immigrants in France (G2) in comparison with first-generation immigrants (G1) and majority population, and particularly the variation according to practice of heritage language.Methodsin a nation-wide probability sample, we focused on the two largest groups (from Maghreb and Southern-Europe: n = 2,736), analysing their use of alcohol, tobacco and cannabis. The main factor of interest was practice of heritage language at home at age 15 (vs. exclusive use of French), and we also considered type of parentage: homogenous (two immigrant parents from the same country) vs. mixed (only one, the other being French).ResultsIn comparison with the majority population (neither G1 nor G2: n = 19,185), G1 had lower levels of substance use, followed by G2 speaking the heritage language and by G2 with homogeneous parentage. The other G2 reported higher levels of substance use, close to those of the majority population (even higher for tobacco among Maghrebins and cannabis among South-Europeans). These associations remained robust after adjusting for potential individual and contextual confounding variables.ConclusionIntegrating those specificities may help in designing culturally grounded prevention policies benefitting the population at large.

Purpose: The "alcohol harm paradox" has been evidenced among adults, but it is still largely unexplored among adolescents. We examined in a sample of French adolescents the relation between family socioeconomic status (SES), family living arrangement and parental substance use on 1 hand, and heavy episodic drinking (HED), lifetime alcohol-induced emergency room visits (A-ERV), and number of alcoholic drinks and solitary drinking during the last episode on the other hand. Methods: A cross-sectional nationwide survey in March 2017 involved 13,314 French adolescents parents' alcohol and tobacco consumption. We used risk ratios (RRs) from modified Poisson parental substance use, and family living arrangement were independently associated with HED. Discussion: Our findings reveal an "alcohol harm paradox" in late adolescence in France. Lower (c) 2023 Society for Adolescent Health and Medicine. All rights reserved.

Background A higher risk of suicidal ideation associated with self-report of Coronavirus Disease 2019 (COVID-19)-like symptoms or COVID-19 infection has been observed in cross-sectional studies, but evidence from longitudinal studies remains limited. The aims of this study were 2-fold: (1) to explore if self-reported COVID-19-like symptoms in 2020 were associated with suicidal ideation in 2021; (2) to explore if the association also existed when using a biological marker of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) infection in 2020. Methods and findings A total of 52,050 participants from the French EpiCov cohort were included (median follow-up time = 13.7 months). In terms of demographics, 53.84% were women, 60.92% were over 45 years old, 82.01% were born in mainland France from parents born in mainland France, and 59.38% completed high school. COVID-19-like symptoms were defined as participant report of a sudden loss of taste/smell or fever alongside cough, shortness of breath, or chest oppression, between February and November 2020. Symptoms were self-reported at baseline in May 2020 and at the first follow-up in Autumn 2020. Serology-confirmed SARS-CoV-2 infection in 2020 was derived from Spike protein ELISA test screening in dried-blood-spot samples. Samples were collected from October 2020 to March 2021, with 94.4% collected in 2020. Suicidal ideation since December 2020 was self-reported at the second follow-up in Summer 2021. Associations of self-reported COVID-19-like symptoms and serology-confirmed SARS-CoV-2 infection in 2020 with suicidal ideation in 2021 were ascertained using modified Poisson regression models, weighted by inverse probability weights computed from propensity scores. Among the 52,050 participants, 1.68% [1.54% to 1.82%] reported suicidal ideation in 2021, 9.57% [9.24% to 9.90%] had a serology-confirmed SARS-CoV-2 infection in 2020, and 13.23% [12.86% to 13.61%] reported COVID-19-like symptoms in 2020. Self-reported COVID-19-like symptoms in 2020 were associated with higher risks of later suicidal ideation in 2021 (Relative Riskipw [95% CI] = 1.43 [1.20 to 1.69]), while serology-confirmed SARS-CoV-2 infection in 2020 was not (RRipw = 0.89 [0.70 to 1.13]). Limitations of this study include the use of a single question to assess suicidal ideation, the use of self-reported history of mental health disorders, and limited generalizability due to attrition bias. Conclusions Self-reported COVID-19-like symptoms in 2020, but not serology-confirmed SARS-CoV-2 infection in 2020, were associated with a higher risk of subsequent suicidal ideation in 2021. The exact role of SARS-CoV-2 infection with respect to suicide risk has yet to be clarified. Including mental health resources in COVID-19-related settings could encourage symptomatic individuals to care for their mental health and limit suicidal ideation to emerge or worsen.

The 2013 FECOND (Fertility, Contraception, and Sexual Dysfunction) probability telephone survey aims to monitor sexual health behaviors among fifteen to forty-nine year olds in France. We conducted a random experiment to compare a classic telephone survey (group T, n = 3,846 respondents) with two Internet-telephone mixed-mode protocols: a sequential Internet-telephone protocol (group S, n = 762, among which there were 462 Internet questionnaires), and a concurrent protocol (group C, n = 1,165, among which there were 208 Internet questionnaires). We compare telephone (T), sequential (S), and concurrent (C) samples on cooperation rates, break-off, and item nonresponse rates, sociodemographic characteristics, health behaviors, and seven sexual health behaviors and personal opinions questions. Reports on the most sensitive behaviors were expected to be more truthful and more prevalent on the Internet—and thus in the mixed-mode samples—than in the telephone sample. The cooperation rate (i.e., the response rate among the possible respondents selected during the initial telephone call) was higher in the classic telephone survey than in the sequential and concurrent mixed-mode protocols (88 percent for T versus 77 percent for S and 55 percent for C), where break-off and item nonresponse rates were also higher. Despite these lower response rates, mixed-mode samples showed better representativeness: their marginal distribution of sociodemographic characteristics was closer to that of the 2013 census, and they had higher R-indicators. A causal estimation of the measurement effect resulting from Internet administration found higher prevalence of three out of the seven sexual health behaviors and personal opinions in the sequential protocol compared to the classic telephone group; a similar pattern was found in the concurrent protocol. In addition, the variance of the weights of the mixed-mode protocols is lower, especially for the sequential design. Sequential telephone-Internet mixed-mode protocols nested in a probability telephone survey may be a good way to improve survey research on sensitive behaviors.

Prospective longitudinal studies mainly conclude on a causal role of e-cigarettes in the initiation of cigarettes in flagrant contradiction with conclusions drawn from epidemiology and other studies showing a sharp decline in cigarette use in parallel with the spread of e-cigarette use. This systematic review explores the reasons for this discrepancy. METHODS:Among 84 publications on e-cigarette/cigarette association in adolescents identified in the Medline database from 2011 to 2022, 23 concern 22 never-smoker longitudinal sub-cohorts. RESULTS:A link between e-cigarette experimentation at T1 and cigarette initiation at T2 is reported in sub-cohort analyses of never-smokers (AOR: 1.41 to 8.30). However, studies exclude 64.3% of T1 e-cigarette experimenters (because of dual-use) and 74.1% of T2 cigarette experimenters. With this study design, e-cigarettes contribute only to 5.3% of T2 cigarette experimentation, casting major doubt on the external validity of results and authors' conclusions that e-cigarettes have a significant effect on the initiation of cigarettes (Gateway effect) at the population level. This sub-cohort design prohibits highlighting any Diversion effect, which is the most likely mechanism accounting for the competition between these two products. CONCLUSIONS:While nicotine abstinence remains the best medical option, over-regulation of e-cigarettes because of misinterpretation of longitudinal study results may be detrimental to public health and tobacco control.

Importance The long-term consequences of COVID-19 on mental health are a critical issue given the number of people infected with SARS-CoV-2 worldwide since the beginning of the pandemic. Objective To investigate the associations between self-reported COVID-19–like symptoms or SARS-CoV-2 seropositivity and subsequent depression or anxiety. Design, Setting, and Participants This propensity score–matched cohort study began in May 2020, with follow-ups in November 2020 and July 2021. The study used data from a large, randomly selected, national population–based cohort from France, the EpiCoV (Epidémiologie et Conditions de Vie) study. Of 85 074 individuals 15 years or older who completed the questionnaires at the 3 collection times, 28 568 were excluded because they did not return a blood sample for serologic testing, 1994 because of missing data on outcomes or exposures, and 9252 to respect the temporal sequence (exposure must precede the outcome). Exposures Propensity scores based on various socioeconomic, lifestyle, and health variables were computed to match participants who experienced COVID-19–like symptoms between February and November 2020 or showed SARS-CoV-2 seropositivity in November 2020. Main Outcomes and Measures Logistic regression models were used to estimate associations between these occurrences and depression or anxiety assessed in July 2021 using the Patient Health Questionnaire 9-item and Generalized Anxiety Disorder 7-item scales, respectively. Results Among the 45 260 included participants (mean [SD] age, 51.1 [18.9] years; 52.4% women; 8.0% with depression and 5.3% with anxiety in July 2021), COVID-19–like symptoms were associated with subsequent depression (adjusted odds ratio, 1.70; 95% CI, 1.45-1.99) and anxiety (adjusted OR, 1.57; 95% CI, 1.29-1.92), whereas SARS-CoV-2 seropositivity was not. Furthermore, COVID-19–like symptoms, but not anosmia or dysgeusia alone, were associated with subsequent depression and anxiety in both the seropositive and seronegative subgroups. Conclusions and Relevance In this cohort study of more than 45 000 individuals drawn from the French general population, SARS-CoV-2 infection was not found as a risk factor of subsequent depression or anxiety. Moreover, self-reported COVID-19–like symptoms were associated with depression and anxiety assessed at least 8 months later in both seropositive and seronegative subgroups, suggesting that factors other than SARS-CoV-2 infection are implied in this association.

Background: Monitoring the prevalence of problematic cannabis use is an important public health issue. International surveys need invariant measurement tools to allow reliable comparisons across countries and between sexes. The Cannabis abuse screening test (CAST) has been developed for this purpose. This study is the first assessing its country and sex invariance in a sample of European pupils. Methods: The data come from the self-administered questionnaires completed in 2019 by pupils aged 15-16 in the European school survey project (Espad). The analytical sample was restricted to the 17 countries where at least 300 pupils reported a previous-year cannabis use (n = 8740); multigroup confirmatory factor analyses were used to assess the configural, metric and scalar invariance of the CAST toward country and sex in the 2019 Espad release. Results: Configural, metric and scalar invariance toward country hold for the 17 countries: Austria, Bulgaria, Croatia, Czech Republic, Denmark, Estonia, Finland, France, Georgia, Greece, Italia, Latvia, Poland, Portugal, Slovakia, Slovenia and Spain. Scalar invariance toward sex was met in the 17 countries as a whole and in 11 of the 12 countries where the test could be run. Scalar invariance toward country was met with 6 additional countries comprising at least 250 respondents: Ireland, Lithuania, Malta, Montenegro, the Netherlands and Norway. Conclusion: The CAST is a suitable test for comparing the measurement of problematic cannabis use amongst adolescents in Europe.

Abstract Background Adolescent heavy episodic drinking (HED) represents a serious health threat. We explore the relation of HED and of emergency room visit because of acute alcoholic intoxication during life (ERV) with family socioeconomic status (SES), family living arrangement (living in a non-intact family or living independently) and parental substance use. Setting: A cross-sectional nationwide exhaustive 12-day survey in March 2017 of French youth aged 17-18.5 participating in the national mandatory civic information day. Participants: 13,314 adolescents answering a pen and paper questionnaire about their own alcohol consumption and the consumption of tobacco and alcohol of their parents. Measurements: Outcomes were the number of past month heavy drinking episodes and ERV. Risk ratios (RR) were computed using modified Poisson regressions. Results Compared to the adolescents with the highest SES, those from the lowest SES were less likely to report 1-2 or 3-5 episodes of heavy drinking instead of none (RR=0.58, 95% CI=[0.50; 0.66] and 0.35 [0.27; 0.45]) but there was no difference for 6 episodes and above (RR=0.81 [0.59; 1.12]). At the extreme, ERV was more frequent among adolescents with the lowest SES (RR=1.86 [1.05; 3.30]). SES, parental substance use and family living arrangement were associated independently with HED; they were not confounded by each other and no interaction between these factors was significant. Conclusion Adolescents from disadvantaged families report fewer episodes of heavy drinking, but are more likely to report alcohol-related emergency room visit, showing the existence of a robust “alcohol harm paradox”.

Symptomatic COVID-19 appears to be associated with suicidal ideation but longitudinal evidence is still scarce. SARS-CoV2-induced neurological damages might underline this association, but findings are inconsistent. We therefore investigated the association between COVID-19 disease and subsequent suicidal ideation in the general population, using both self-reported symptoms and serology as well as inverse probability weighting to draw as near as possible to the direct association.Using data from the nationwide French EpiCov cohort, COVID-19 disease was assessed through 1) COVID-19 illness (self-reported symptoms of sudden loss of taste/smell or fever alongside cough, shortness of breath or chest oppression, between February and November 2020), and 2) SARS-CoV2 infection (Spike protein ELISA test screening in dried-blood-spot samples). Suicidal ideation was self-reported between December 2020 and July 2021. Inverse probability weighting with propensity scores was used as an adjustment strategy, leading to balanced sociodemographic and health-related factors between the exposed and non-exposed groups of both COVID-19 disease measures. Then, modified Poisson regression models were used to investigate the association of COVID-19 illness and SARS-CoV2 infection with subsequent suicidal ideation.Among 52,050 participants from the EpiCov cohort, 1.68% [1.54% - 1.82%] reported suicidal ideation in the first half of 2021, 9.57% [9.24% – 9.90%] had a SARS-CoV2 infection in 2020 and 13.23% [12.86% – 13.61%] reported COVID-19 symptoms in 2020. COVID-19 illness in 2020 was associated with higher risks of subsequent suicidal ideation in the first half of 2021 (Relative Riskipw [CI95%]= 1.43 [1.20 – 1.69]) while SARS-CoV2 infection in 2020 was not (RRipw = 0.88 [0.69 – 1.12]).If COVID-19 illness was associated with subsequent suicidal ideation, the exact role of SARS-CoV2 infection with respect to suicide risk has yet to be clarified. Psychological support should be offered to persons recovering from symptomatic COVID-19 in order to minimize suicidal ideation risk. Moreover, if such psychological support is to be implemented, serology status alone does not seem a relevant criterion to define persons who suffered from COVID-19 to prioritize.### Competing Interest StatementThe authors have declared no competing interest.### Funding StatementThe present work was supported by a French National Observatory of Suicide research grant attributed to AR (grant number R21094LL). The EpiCov study was supported by research grants from Inserm (Institut National de la Sante et de la Recherche Medicale), the French Ministry for Research and its department of research, studies, evaluation and statistics (Direction de la Recherche, des Etudes, Evaluation et Statistiques, Drees), the French Ministry for Health, and the Region Ile de France. Dr.Bajos has received funding from the European Research Council (ERC) under the Horizon 2020 research and innovation program of the European Union (grant agreement number 856478). This project has also received funding from the Horizon 2020 research and innovation program of the European Union under grant agreement number 101016167, ORCHESTRA (Connecting European Cohorts to Increase Common and Effective Response to SARS-CoV-2 Pandemic).### Author DeclarationsI confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained.YesThe details of the IRB/oversight body that provided approval or exemption for the research described are given below:The EpiCov study received approval from an ethics committee (Comite de Protection des Personnes Sud Mediterranee III 2020- A01191-38) and from the French National Data Protection Agency (Commission Nationale Informatique et Libertes, CNIL, MLD/MFI/AR205138).I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals.YesI understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance).YesI have followed all appropriate research reporting guidelines and uploaded the relevant EQUATOR Network research reporting checklist(s) and other pertinent material as supplementary files, if applicable.YesAnonymous aggregated data for the baseline wave are available online. The EpiCov dataset is available for research purpose concerning baseline wave, and will be available by March 2022 concerning the second follow-up wave for research purpose on CASD (https://www.casd.eu/), after submission to approval of French Ethics and Regulatory Committee procedure (Comite du Secret Statistique, CESREES and CNIL). Access to anonymized individual data underlying the findings may be available before the planned period, on request to the corresponding author, to be submitted to approval of ethics and reglementary Committee for researchers who meet the criteria for access to data.

Objective: The measures put in place to contain the coronavirus epidemic in France have led to fears of a massive increase in the consumption of alcohol, tobacco, and other drugs. Method: Taking advantage of an ongoing weekly Internet survey monitoring living conditions of the French population during the containment, we introduced questions on alcohol and tobacco use 1 month after the start of the first containment. Analyses were conducted with multivariate Poisson regressions. Results: Tobacco consumption remained almost unchanged. By contrast, more than a third of French people did change their weekly alcohol intake: 13% increased it, whereas 21% decreased it, indicating that the containment measures were often taken as an opportunity to reduce alcohol consumption, especially among those who drank less than 4 days per week before. Men were more likely to decrease their alcohol intake and less likely to increase it than women, probably because of a differential impact of the reduction of social activities on gendered social roles. The people most affected were smokers and those living in a dwelling in which someone had been infected, including themselves. Conclusions: The crisis tended to polarize alcohol consumption toward extremes. If the new drinking habits were to continue, the consequences could be beneficial for occasional drinkers but harmful for heavier drinkers.

BACKGROUND AND AIMS:Among European countries, France is particularly concerned by adolescent tobacco smoking, especially in disadvantaged socio-economic backgrounds (SES). We measured the respective contributions of parental smoking and family living arrangement to social disparities in smoking during adolescence.DESIGN:Secondary analysis of survey data.SETTING:A cross-sectional nation-wide exhaustive 12-day survey in March 2017 of French youth aged 17-18.5 years participating in the national mandatory civic information day.PARTICIPANTS:A total of 13 314 adolescents answering a pen-and-paper questionnaire about their own tobacco consumption and the smoking of their parents.MEASUREMENTS:Risk ratios (RRs) were computed using modified Poisson regressions, and population-attributable fraction (PAF) was used as a measure of the explanatory roles of the different factors as mediators of SES.FINDINGS:Adolescents living within very privileged and privileged SES were significantly less likely to report daily tobacco smoking (20.4 and 22.7%, respectively) than those within modest and disadvantaged ones (26.0 and 28.6%, respectively). Parental smoking and family living arrangement independently explained the smoking inequalities among adolescents. After adjusting for schooling factors, the risks associated with parental smoking ranged between RR = 1.64 [95% confidence interval (CI) = 1.50-1.79] when the father only smoked and RR = 2.17 (95% CI = 1.99-2.36) when both parents smoked, compared with non-smoking parents; the risk associated with living in a non-intact family was 1.35 (95% CI = 1.26-1.43) and that of living outside the parental home was 1.20 (95% CI = 1.10-1.30). Apprentices and adolescents out of school had higher risks than those at school (RR = 1.82, 95% CI = 1.68-1.98) and RR = 2.10, 95% CI = 1.92-2.29). The contribution of parental smoking to adolescent smoking (PAF = 32%) was greater than that of SES (PAF = 9%), family living arrangement (PAF = 17%) or schooling factors (14%). The share of SES decreased from 18 to 9% when considering these mediating factors.CONCLUSION:In France, parental smoking appears to be the factor that most influences adolescent smoking, followed by family living arrangement; the role of family socio-economic status is small in comparison.

BACKGROUND:Over the last two decades, telephone surveys based on random digit dialing have developed considerably. At the same time, however, the proportion of the population with a cell phone has increased, whereas landline frame coverage has declined, thus raising the possibility of discontinuing landline phone surveys. This paper aims to assess the impact of using a single-frame (SF) cell phone design instead of a dual-frame (DF) design with landlines and cell phones in the context of repeated health surveillance surveys in the general population. We analyze data from a random digit dialing health survey of the French population and assess differences between the DF and the counterfactual SF design that excludes the landline phone sample from the DF design. We evaluate the quality of the two survey designs in terms of survey productivity, response rates, representativeness, balancing of external covariates, and prevalence estimates of key health behavior indicators.RESULTS:Our results show that a SF cell phone survey has several advantages over a combined DF landline and cell phone survey. Cell numbers require fewer call attempts to complete an interview, leading to a substantial reduction in the mean data collection duration and weight dispersion. The global representativeness of the SF design was slightly better than its DF counterpart, although the elderly were underrepresented. After calibration, differences in health behavior estimates were small for the seven health indicators analyzed.CONCLUSIONS:Switching from a DF random telephone survey to a SF cell phone design has a number of practical advantages and would have a minimal impact on general population health surveys for monitoring health behavior at the population level. However, the different aspects of the survey quality had to be studied to make a decision. Further studies are needed to explore the scope of possibilities.

Because the effectiveness of a coronavirus disease lockdown in curbing coronavirus disease spread depends on public support, acquiring real-time information about the way populations reacted to the lockdown is crucial. In France, such public support remained fragile among low-income persons, probably because the lockdown exacerbated preexisting social inequalities and conflicts.

Background: the EpiCov study, initiated at the end of the first national lockdown in France, aimed to provide national and regional estimates of the seroprevalence of SARS-CoV-2 infection, and to analyze relations between living conditions and the dynamics of the epidemic. We present and discuss here the survey methodology, and describe the first-round fieldwork. Method: 371,000 individuals aged 15 years or more were randomly selected from the national tax register, stratified by departments, including three overseas departments, and by poverty level with over-representation of people living below the poverty line. Health, socio-economics, migration history, and living conditions were collected through self-computed-assisted web interviews or via computer-assisted telephone interviews. The first-round survey was conducted in May. A random subsample was eligible to receive material for home blood self-sample on dried blood spot (DBS), in order to detect IgG antibodies against the spike protein (Euroimmun ELISA-S), and neutralizing antibodies for non-negative ELISA-S. For the second-round conducted in November, all respondents were eligible for the antibodies detection from home DBS sample, as well as the other household members aged 6 years or more for 20% of them. Participation and adjustment for nonresponse: 134,391 respondents completed the first-round questionnaire from May 2 to June 1, 2020, including 16,970 (12.6%) respondents under the poverty line. Multimodal web/tel interviews was randomly assigned to 20% of the sample. The other were assigned to exclusive CAWI. Overall 17,441 respondents were eligible for home blood sample, among them 12,114 returned the DBS (interquartile date: May 25- June 5). The response probability was first estimated from logit models adjusted on a wide range of auxiliary demographic and socio-economic variables available from the sampling frame, and final weights calibrated to the margins of the population census permitted to correct for a large part of the non-response bias. Conclusion: The Epicov study is one of the largest national random population-based seroprevalence cohort, with both an epidemiological and sociological approaches to evaluate the spread of the COVID-19 epidemic, and the impact on health and living conditions. One of the major interests of this study is the broad coverage of the socio-economic and territorial diversity of the population.

Background: Studies have proved the positive link between truancy and substances use in school surveys. In spite of this, no adapted weighting treatment is generally provided; even when the share of missing and truant pupils is high, and all drug use estimates are biased downward. The necessary data can be collected: on one side, individual current drug use and past episodes of absence and truancy of the respondents; on the other, the count of the presents and absents the day of the survey, including truants, in each class. However, the nature of these data prevents any classical modelling of the survey response without additional assumptions. Methods: We review one method proposed in 2002 by Guttmacher and al. that uses only the individual data and propose two methods that combine both kind of data and in which we can distinguish or not between truancy and legitimate absence. We apply them to the French release of the 2015 Espad survey (European survey project on alcohol and other drugs). The theoretical number of pupils was n=7166; 981 were absent (including 359 truants), while 178 were discarded because of the poor quality of their questionnaires and 6007 were considered final respondents. Assumptions, point estimates and variances are compared. Results: Guttmacher’ method is not conceptually valid and can lead to irrelevant corrections with high variances. Our estimate of cannabis regular use is 8.6% (std=0.75) instead of 7.7% (std=0.67), that is a non-response bias of circa 14%. Conclusion: The proposed approach relies on simple and plausible assumptions; it is preferable to any speculative consideration about the magnitude of the underestimation yielded by the classical weighting procedures. Survey designers should evaluate and discuss the potential bias of their surveys and eventually correct it.

Abstract There is an ongoing debate in the survey research literature about whether and when probability and nonprobability sample surveys produce accurate estimates of a larger population. Statistical theory provides a justification for confidence in probability sampling as a function of the survey design, whereas inferences based on nonprobability sampling are entirely dependent on models for validity. This article reviews the current debate about probability and nonprobability sample surveys. We describe the conditions under which nonprobability sample surveys may provide accurate results in theory and discuss empirical evidence on which types of samples produce the highest accuracy in practice. From these theoretical and empirical considerations, we derive best-practice recommendations and outline paths for future research.