Advertising and Young People’s Critical Reasoning Abilities: Systematic Review and Meta-analysis

BACKGROUND AND OBJECTIVES Young people are exposed to an abundance of advertising for unhealthy products (eg, unhealthy foods, tobacco, alcohol). Because of their developing cognition, children may not be able to understand the intent of advertising. However, advertising restrictions often assume that adolescents have critical reasoning capacity and can resist the effects of advertising. This review seeks to assess whether the evidence supports this assumption. METHODS Ten databases were searched in December 2020. Inclusion criteria were participants aged 6 to 17 years, any advertising exposure, objectively measured understanding or attitudinal outcome, a comparison, control, and between-group comparison. This study included all languages and excluded studies published pre-2010. Two reviewers independently extracted data and assessed study quality. RESULTS Thirty-eight articles were included. Meta-analysis of 9 studies with attitudinal outcomes indicated that unhealthy product advertising generated more positive brand or product attitudes compared with neutral or no advertising control in all ages. There were significant effects for digital and nondigital advertising formats. We found greater understanding did not protect against the impact of advertising on brand or product attitudes. Limitations include the inability to meta-analyze the impact of advertising on understanding or the influence of age. CONCLUSIONS Evidence shows that the attitudes of young people were influenced by advertising. Critical reasoning abilities did not appear to be fully developed during adolescence and not found to be protective against the impact of advertising. Policymakers should ensure regulations to restrict marketing of unhealthy commodities protects adolescents as well as younger children.

][3][4] Advertising can lead to behavior change through direct and indirect pathways, which leads to harm through unhealthy behaviors. 5The hierarchy of effects model suggests that advertising creates awareness of and interest in a brand or product, which leads to heightened preference and then to a decision to purchase and consume. 6Much of the advertising children are exposed to is for potentially harmful products (eg, HFSS food, alcohol) which may increase unhealthy behaviors that are associated with a number of detrimental and harmful effects. 7,8irect tobacco advertising is banned in most countries, but young people are still exposed to indirect advertising, for example, through viewing tobacco use on television (TV), shown to result in smoking initiation in young people. 9lectronic cigarettes (e-cigarettes) have also grown in popularity over the last decade, and provaping advertising are prevalent on social media, with emerging evidence of harm. 10,11Research on the impacts of advertising on children over the past decade has focused particularly on HFSS food advertising. 12Young people are exposed to large amounts of food advertising through various media, which is often child-targeted and mostly for HFSS foods. 13,14eta-analyses show that food advertising increases acute calorie intake in children. 15,16ere has been a strong policy focus on tightening regulations around food advertising, although restrictions frequently only apply to children up to 12 years of age. 17There are widespread restrictions to prevent alcohol and tobacco advertising that targets children, since these products are illegal for children to purchase or use, 18 with calls to make these restrictions worldwide to address noncommunicable diseases. 19esearchers have raised concerns over the ability of children and young people to identify, understand, and apply critical reasoning in response to advertising.Thus, they are more susceptible to the influence of advertising, especially in digital formats (including embedded content on webpages, social media platforms and advergames), making this a policy target and active research area. 20ere is substantial literature on the understanding of advertising.A prominent framework has been the "Persuasion Knowledge Model," which proposes that to resist advertising, individuals must first recognize that an advert is trying to sell something (persuasion knowledge). 21Various aspects of understanding have been identified: recognizing advertising; perception of who pays for advertising and audience targeting; understanding the selling intent of advertising (ie, that advertisers are trying to sell products), persuasive intent (ie, that advertisers are trying to influence behavior via changing attitudes toward products or brands), tactics (ie, specific strategies used), and bias regarding the product (ie, discrepancies between advertised and actual product). 22Evidence suggests that "advertising literacy" (ie, knowledge and understanding of advertising intent and tactics) does not fully develop during childhood; therefore, children do not possess the necessary cognitive ability to resist advertising. 22,23For this paper, we view critical reasoning as the ability to recognize and understand advertising (advertising literacy) and how it impacts children and young people's response to advertising.Much of the work around children and advertising, and children's broader position as consumers, has been informed by Piagetian theory, which presents age-specific stages in children's development driven by cognitive ability. 24,25This suggests progressive growth in understanding, showing that as children get older, cognitive ability increases along with an increased ability to understand and resist advertising.This understanding was largely developed when TV was the main advertising medium, but the applicability to the digital age of advertising has been questioned (even for older children), as entertainment and advertising content are not clearly distinguished. 26cial-cognitive models present the effects of advertising occurring automatically without any information processing, suggesting that understanding alone is insufficient to counteract the potentially harmful effects of advertising. 25Concerning food, the Food Marketing Defense Model posits that awareness, understanding, ability (including cognitive capacity), and motivation (to resist advertising) are all required to withstand food advertising. 25Advertising, especially when digitally embedded, is designed to bypass conscious and rational decision-making and instead rely on emotional responses and unconscious processing, thereby inhibiting the ability to resist effectively. 27,28asoning abilities are not fully developed by the age of 16, older than the 12-year threshold used in many regulations; other faculties associated with decision-making also continue to develop into adulthood. 29It is established that teenagers engage in riskier behavior than both children and adults, attributed in part to changes in reward sensitivity occurring from early adolescence and the later development of self-regulatory competence. 29In addition, they may be particularly susceptible to the social influence of their peers. 30This evidence may be relevant to young people's critical reasoning of advertising, since developmentally, they may not be cognitively equipped to protect themselves from the potentially harmful effects of advertising.Studies indicate that children of all ages have difficulties identifying digital marketing. 26,27dolescents are particularly vulnerable to digital advertising because of their engagement with digital technology and media, which plays an important role in their social identity development. 17,27isting reviews and meta-analyses have shown that children of all ages are impacted by advertising, [13][14][15][16] but the notion that understanding of advertising and older age are sufficiently protective remains pervasive.This review focused on 2 areas of interest; the ability of young people to recognize and understand advertising and how they respond to advertising in terms of attitudes toward the advertised brand or product (ie, the impact on diet and attitudes).The review aimed to explore whether evidence supports the notion that critical reasoning ability affects behavioral responses and how this may differ across childhood and adolescence.Critical reasoning relates to the former, but response is likely to include broader factors that could impact on what decisions young people make and their subsequent behavior.For example, attitudes to the advertised product or brand and level of motivation to resist the impact of advertising exposure.

METHODS
We conducted our systematic review using EPPI-Reviewer 4 software. 31The study was preregistered with PROSPERO (CRD42018116048), and the systematic review is reported in accordance with the Preferred Reporting Items for Systemic Reviews and Meta-analysis checklist. 32

Search Strategy, Eligibility Criteria, and Information Sources
The search strategy was created in collaboration with an information specialist (C.S.).The search was based on terms for population (children and young people), intervention (eg, marketing, advertising, advergame*), and a measure of "understanding" or "attitudes" (eg, reasoning, psychology, advertising literacy, cognition).Systematic searches of the following databases were conducted: ASSIA (Proquest), Child Development and Adolescent Studies (EBSCO), Cochrane Central Database of Controlled Trials, Medline (OVID), PsycINFO (OVID), Sociological Abstracts (Proquest), Social Policy and Practice (OVID) and SCOPUS, and Web of Sciencedatabases (Social Science Citation Index, Emerging Sources Citation Index).The full search strategy is included in a supplemental file (Supplemental Tables 2A, 2B, and 3).Searches were conducted on November 7, 2018 and updated on December 10, 2020.The search results were imported into Endnote reference manager software and duplicates removed.The remaining articles were imported into EPPI-Reviewer 4 software and duplicate records screened and removed; this software was used to manage the screening.
The focus of the review was initially broad, as the scope of the literature was unknown.Following the initial search, a mapping exercise was undertaken to determine the fulltext inclusion criteria.A decision was made to focus on experimental studies with an administered exposure (Supplemental Fig 4) for full details of this initial stage and mapping diagram).
Eligible for inclusion during full text screening were: studies with participants aged 6 to 17 years of age inclusive; intervention criteria of any form of advertising for any product (including HFSS products, tobacco, toys); and outcomes of objectively measured understanding (including recognition or identification of advertising, understanding selling, or persuasive intent) or attitudes (toward brand or product including liking or perceptions).Experimental and intervention studies, including randomized or quasi-randomized studies, were included and required to have an appropriate comparison or control group, including no advert, a neutral advert, or a between group comparison (age, gender, socioeconomic status [SES]) with an advert exposure.Neutral adverts were defined by the studies and included adverts that were not the focus of the study, eg, a toy or nonfood advert for studies with a food advertising exposure and food product outcome.Studies were included from 2010 onwards as these were considered most relevant to contemporary advertising practices.There were no restrictions by geography or language.Exclusion criteria were date (pre-2010), intervention (any exposure that evaluated health promotion prevention programs, charity advertising, creation, and testing of models of cognition, media training or advertising literacy, branding only), outcome measures (any nonunderstanding or attitude measures including dietary intake or purchases), study design (qualitative studies, reviews, and dissemination format [nonpeer reviewed], eg, dissertations, conference abstracts, magazine abstracts).A random sample of studies were doublescreened by 2 reviewers (H.C., and J.P.) on title and abstract using EPPI-Reviewer 4 software.All screening queries were reconciled by the reviewers.We used the machine learning capabilities of the EPPI-Reviewer software to assist with the screening because of the anticipated number of records from test searches (over 10 000).We employed an "active learning approach," where the prioritization of records was frequently refreshed so the most relevant articles were screened first.The algorithm was trained using our screening decisions.Articles screened were plotted against studies included, and this was used to indicate when to stop screening (ie, the rate of inclusion plateaued indicating that there were unlikely to be unscreened relevant articles).A classifier model was then created and applied to all unscreened records, with a score based on relevance (0-100) generated and used to double-check exclusion.For full details on the machine learning approach and updated search methods see Supplemental Table 4 and the following reference. 33Full-text screening was then independently completed by the same 2 reviewers (H.C. and J.P.) using EPPI-Reviewer 4 software and queries were jointly reconciled.

Data Extraction
Descriptive data were extracted by 1 reviewer (J.P.) and checked for accuracy by another reviewer (H.C.).Data from experimental studies for inclusion in meta-analyses were independently extracted by 2 authors (J.P., and H.C.) and any discrepancies resolved by reextraction.Corresponding authors were contacted to provide raw data where necessary; 15 authors were contacted for additional information, and 9 provided additional data and 6 did not (1 was contacted regarding understanding outcomes only).

Assessment of Quality
Risk of bias for the experimental studies was assessed by 2 reviewers (H.C., and J.P.) using Cochrane methods, 34 either RoB 2.0 for randomized trials 35 or ROBINS-I tool for nonrandomized studies. 36o assess publication bias, funnel plots were created to assess asymmetry using Egger's test. 37

Data Synthesis
For inclusion in meta-analyses for understanding of advertising or brand or product attitudes, studies were required to compare the effect of an unhealthy product (eg, food, alcohol, tobacco) advert exposure to a nonadvert control, or to a control advert (advert for unrelated products).
Studies measuring attitudinal outcomes were required to have mean values with standard deviations.Because of differences in reported outcome measures, which included a variety of different scales (eg, 1-5, 1-3, dichotomous), the DerSimonian-Laird random-effects model was used to allow for synthesis of studies and standardized mean difference (SMD) was used as the outcome for the meta-analyses.All analyses were conducted using Stata 16 (16.1,StataCorp LLC, College Station, TX, USA). 37Further details of how advert exposure conditions were combined; the outcome measures and scales and criteria for inclusion in the meta-analyses are provided in a supplemental file (Supplemental Table 5).
Two meta-analyses comparing an advert exposure to control or neutral advert were conducted, by attitude type (brand or product) and by advertising format (digital or nondigital).For this review, we define brand attitude as the attitudes toward the advertised brand and product attitude as the attitudes toward the advertised product.Digital advertising formats included advergames, webpages, social media platforms, and influencer marketing, whereas nondigital advertising formats included TV and printed adverts, and product placement on TV or in movie clips.For all studies except 1, 38 a single combined advert exposure group was calculated for each group using Cochrane methods. 34The exception was a study where 3 separate data points were included with the advert exposure of a specific product matched to the specific product attitudinal outcome measure. 38e additionally conducted metaanalyses examining the impact of advertising on attitudes by age (children #12 years, teenagers >12 years because of legislation cut-offs).

Narrative Synthesis
Findings of studies not included in the meta-analysis are reported narratively, presented by outcome (understanding or attitudinal) and by impact of age and advertising features.

Study Selection
The database searches yielded 15 656 papers, resulting in 9325 studies once duplicates were removed.A random subset of 1790 studies were screened on title and abstract to trigger the machine learning from the original search and a further 208 screened on title and abstract from the updated search.Screening on title and abstract ultimately resulted in 272 studies to be screened on full-text and assessed for eligibility.This resulted in 39 studies, from 38 articles, which met the inclusion criteria.Nine of the studies that reported an attitudinal outcome were included in the meta-analyses (Fig 1).

Study Description and Results
A summary of the descriptive data are provided in Table 1, including details on setting (country, study), participants (sample size, age details), design, advertising exposure, outcomes measures, and findings.

Narrative Synthesis of Understanding Outcomes
Meta-analysis was not possible for understanding measures, owing to the heterogeneity of exposures and outcomes for relevant studies.Many studies had control groups where, because of the nature of the questions, understanding of advertising was not able to be assessed (ie, cannot assess understanding about an advert the group did not see).

Impact of Age on Understanding
Where compared across age groups, understanding of advertising increased significantly with age (8 studies, mostly assessed as some concerns of bias and 1 as low risk of bias), 39,40,41,46,47,49,62,74 although no significant effects were found in 4 studies (mostly assessed as having some concerns of bias and one as low risk of bias), 45,48,66,67 and understanding decreased with age (assessed as low risk of bias). 75ost of these studies were conducted with children under 12 years, so evidence was limited for teenagers.Of 2 studies conducted with teenagers, 1 study assessed as having some concerns of bias directly compared children aged 9, 12, and 15 years and found that advertising recognition significantly increased as age increased 74 ; the other study assessed as low risk of bias found 12 to 14 years olds had significantly higher recognition of sponsored content in a YouTube video compared with 15 to 16 year olds, but there was no significant difference between age groups for understanding persuasive intent. 75

Impact of Advert Content on Understanding
One study with some concerns of bias reported that persuasion knowledge increased with higher brand integration (in relation to advergames), but persuasion knowledge was very low across all groups and the magnitude of differences modest. 47In relation to child "involvement" with advertising (ie, engagement with advergame), 1 study with some concerns of bias showed that children more involved with an advergame were less likely to identify commercial content. 48ne study with low risk of bias       Difference in persuasion knowledge between the conditions was not looked at differences in recognition of commercial content in advergames between a familiar HFSS brand and a fictitious or unbranded pizza game and found that recognition of the familiar brand was significantly greater than the unbranded game. 58A similar study with some concerns of bias assessed persuasion knowledge between a branded advergame and a noncommercial advergame and found no significant difference. 52even studies of mixed bias assessments (4 with some concerns, 3 low risk) measured different types of understanding; 4 found that awareness of selling intent was higher than persuasive intent in children aged 4 to 12 years (2 were significant 39,45 ; 2 did not test significance) 46,48,62 ; 2 found recognition of advertising in 7 to 16 year olds was greater than understanding persuasive intent 76 or advertising literacy 44 ; finally, 1 found skeptical attitudes toward advertising were greater than recognition of advergames as advertising, because of very low recognition in 7 to 11 year olds (62.5% to 72% vs 48.5%). 49Four studies with some concerns of bias measured the impact of advertising format and found significantly greater understanding with nondigital advertising (TV) compared with digital advertising (primarily advergames). 46,48,62verall, understanding of the persuasive intent of adverts to impact on attitudes and behaviors was generally low across studies, for example, only 40% in 11 to 12 year olds, 39 and only 1% of 7 to 9 year olds and 12% 10 to 12 year olds. 44
A meta-analysis exploring the effect of advertising by format (Fig 3) showed overall that any advert significantly increased positive attitudes, compared with no advert or neutral control, SMD 5 0.36 (P 5 .009;95% CI 0.14-0.58;I 2 5 91.2%).When examined by advertising format, both digital advertising exposure and non-digital advert exposures had a significant positive effect on attitudes, SMD 5 0.35 (P 5 .005;95% CI 0.01-0.068;I 2 5 93.2%) and SMD 5 0.36 (P 5 .005;95% CI 0.08-0.65;I 2 5 84.5%), respectively.Egger's regression analysis found no evidence of bias for either meta-analysis, although funnel plots showed some evidence of asymmetry (Supplemental Figs 5 and 6).Trim and fill analysis showed no strong evidence of missing studies for either meta-analysis (Supplemental Figs 7 and 8).Sensitivity analysis was completed running a fixed effect model; none of the findings changed in significance in either direction.
An additional meta-analysis was conducted, which looked at the impact of advertising on attitudes by age (Supplemental Fig 9).Advertising had a positive impact on attitudes compared with the control condition for both age groups (ie, >12 years and #12 years).A further meta-analysis was carried out as a sensitivity analysis (Supplemental Fig 10) to explore whether the effect held when the largest effect size was removed and the effect was still seen.

Narrative Synthesis of Attitudinal Outcomes
The majority of controlled studies not suitable for meta-analysis supported the above findings, namely that adverts brought about more positive attitudes (7 studies, mixed bias assessments: 3 low, 3 some concerns and 1 high) 41,42,47,48,53,57,59 ; however, 5 studies found no significant differences between groups (mixed bias assessment: 3 low, 2 some concerns). 51,57,58,69,73One study, assessed as having a high risk of bias, explored the impact of ecigarette adverts designed with low and high youth appeal and found the low youth appeal advert resulted in more positive attitudes than a none-cigarette control advert, but there was no difference between the high youth appeal and control adverts. 56One study, assessed as low risk of bias, found that the younger group (5-6 years) had significantly more positive product attitudes following exposure to TV advert for HFSS cereal compared with the older group (10-11 years). 50nother study with some concerns of bias found that brand preference following exposure to product placement decreased significantly with increasing age (9 vs 12 vs 15 years). 74o studies with low risk of bias examined the impact of glamorized e-cigarette advertising on perceptions of cigarette smoking or e-cigarettes, compared with neutral or no advert control.They found the adverts led to occasional cigarette smoking being perceived as less dangerous and harmful 63,71 and the use of e-cigarettes by children as being more common. 63One also found there was no difference in the appeal of e-cigarettes between adverts that glamorized e-cigarettes compared with adverts that associated e-cigarettes with health. 63

Impact of Understanding on Attitudinal Outcomes
Seven studies measured the interaction between understanding and attitudinal outcomes and reported interactions.Five studies found no interaction, showing that greater understanding of advertising did not limit favorable attitudes toward the advertised product 44,47 ,48,58,65 and 2 found some evidence of an interaction. 62,67Six of these studies were found to have some concern of bias, and the other was assessed to have low risk of bias. 58his study found, for children aged 13-18 years, recognition of commercial intent had no effect on brand attitude for either an unfamiliar or familiar brand.The age range for children from studies that found no interaction was broader than those that found interactions (5-18 vs 7-14 years).Of the 2 studies that found an interaction between lack of persuasion knowledge and greater attitudinal outcomes, the first had online pop-up adverts, which are heavily embedded, as the advertising exposure, 72 whereas the second only found an interaction among children that understood the snack was unhealthy (the interaction was not observed if children thought the advertised snack was healthy). 67

Quality Assessment
For nonrandomized studies, 2 were rated as low and 11 as moderate risk of bias (Supplemental Fig 11).Moderate risk of bias was mostly caused by the domain "bias caused by confounding," as not enough information was provided or confounding variables were not included in analyses.Of the randomized studies, 10 were rated as low risk of bias, 13 as some concerns, and 3 with high risk of bias (Supplemental Fig 12).The studies with some concerns were mostly because of lack of detail about the randomization process or unreported information about the selection of the reported results.Results were consistent between studies rated as low to high risk of bias.Sensitivity analyses were run excluding studies rated as high risk from the meta-analysis (Supplemental Fig 14).
The overall impact of advertising on attitudes remained but product attitude subgroup was no longer significant.

DISCUSSION
In this systematic review, data suggested that children's understanding of advertising intent was limited and not nuanced, ie, children could recognize that adverts intended to sell a product but not that these were intended to change their attitudes and behavior.Our findings indicate that children and young people of all ages have some difficulties in understanding advertising.This fits with the developmental perspective that young people's critical reasoning abilities continue developing into late adolescence. 29We found that greater understanding does not necessarily protect against advertising, consistent with the Food Marketing Defense Model that challenges the focus on understanding to counteract the effects of advertising.The model instead proposes that advertising influences young people without conscious processing and that motivation to resist is also required, which may be lower among young people. 25We did not include disclosure or media literacy intervention exposures in this review, but our findings suggest that the inclusion of disclosures (eg, declarations stating "this is an advert") or media literacy training designed to increase understanding or advertising literacy would not necessarily protect children and adolescents from the influence of advertising. 76This is supported in the literature as 1 experiment found that children who viewed food marketing with a disclosure actually consumed significantly more of a marketed snack than a control group. 76A study in adolescents found that disclosures did not mitigate persuasion and increased brand memory, despite increasing understanding of persuasive intent. 77Media literacy programs are a strategy often suggested by the food and beverage industry to increase persuasion knowledge in children, in lieu of improved regulations, such as industry-funded Media Smart (see https://mediasmart.uk.com/). 78,79r findings that advertising had a positive impact on attitudes are consistent with previous research on food advertising. 12,14,80,81Further supporting these findings, adverts (TV and advergames) for "unhealthy" unfamiliar food products have been found to elicit positive attitudes in children (aged 7-12 years) to a greater extent with advergames compared with TV advertising. 82We found effects on attitudes regardless of age, consistent with other studies in different age groups.
There is evidence that preschool children exposed to adverts for a range of child-directed foods had positive attitudes about these foods, 83 and that adolescents reported positive attitudes after viewing online adverts for fast food and confectionery. 84mparing digital and nondigital advertising formats, we found no difference in impact on attitudes in subgroup meta-analysis, but narrative synthesis indicated that understanding was lower for digital formats.This is unsurprising since digital advertising is more integrated and, therefore, may be less explicit and more difficult to identify and understand, in addition to greater personalization and targeting. 21,25This is important given the ubiquity of these formats, especially for adolescents, who because of their extensive engagement with digital media with less supervision, may be more susceptible to digital advertising. 85For adolescents, media plays an important role in their social identity development, as they place more value on the opinions and actions of peers and figure out their perception of how they fit with others. 17,18,27igital marketing, especially on social media, is designed to target these unique developmental vulnerabilities. 86

Implications
The findings from this review support understanding not being fully developed during childhood or adolescence.We also found that advertising influences the attitudes of young people of all ages, suggesting a need to protect older as well as younger children.Our results suggest that understanding does not protect children from the harmful impacts and influence of advertising, as per the Food Marketing Defense Model. 25educing exposure to advertising is therefore likely to be more effective than improving understanding through disclosures or media literacy training.Existing regulations typically only apply to children up to 12 years of age, as they have historically been regarded as more vulnerable to advertising, therefore needing greater protection. 87Our findings do not support lesser restrictions for advertising to teenagers, as there is no distinct evidence-based threshold for understanding that supports a cut-off of 12 years and suggest that appropriate protection from advertising exposure would benefit all young people. 17

Limitations
The limitations of this review include a lack of suitable data or studies to meta-analyze the impact of advertising on understanding or the influence of age.Meta-analysis limitations include the high heterogeneity of studies, despite using a random effects model and standardized mean difference outcome.The machine learning method has limitations, as a large number of articles were excluded without screening on title and abstract.The majority of the included studies were assessed as having some concerns of bias, which needs to be taken into consideration when interpreting the findings, although sensitivity analyses removing studies with high risk of bias and the largest effect size were conducted and not found to impact results.We may not have identified all product placement exposure studies, as this term was not included in our search strategy; however, studies with advert or marketing key words were included.Some of the studies may have been conducted in the same or similar group of participants (Tarabashkina [66][67][68] ; Duke and Farrelly 53,54 ; Uribe 69,74 ; van Berlo 57,58 ; Castonguay 40,50 ), but these do not interfere with the meta-analysis as only 1 was included.The time since the searches were completed is a limitation, with original searches completed in October 2018 and then updated in December 2020.This subject area is complex, so the review process is time intensive.Updating the searches would be low yield as the substantive findings of the work remained unchanged following the update searches, and we have no reason to believe the main findings of the paper would be subject to change.
The main strength of the paper is that it meets an evidence gap, specifically addressing if children over 12 years of age have critical reasoning capacity and can therefore resist the effects of advertising.We were also able to quantitatively assess the impact of advertising on attitudinal outcomes.The search was carefully planned and executed, with double screening and data extraction.Studies were contemporary, adding to the relevance for current policy.Because of the delay observed in research, we found fewer studies using digital advertisement exposures, which is an area where more primary research is needed.There is also a need for further primary research in teenagers in relation to critical reasoning and advertising, and especially digital formats.

CONCLUSIONS
This systematic review and metaanalysis provide evidence that advertising impacts upon the attitudes of children and young people of all ages, regardless of their level of understanding and critical reasoning abilities.These findings may be useful to inform the thinking of policy makers, particularly in terms of restrictions based on age and changing patterns of media consumption.

FIGURE 1 Preferred
FIGURE 1Preferred Reporting Items for Systemic Reviews and Meta-analysis screening flowchart.

FIGURE 2 Forest
FIGURE 2Forest plot showing SMD in brand and product attitudes between any advertising exposure and no advert or neutral advert controls; 95% CIs and study weights are indicated.Overall SMD was generated by a random effects model.(1) Data from cola product placement vs control with cola attitude question;(2) Data from juice product placement vs control with juice attitude question; (3) Data from milk product placement vs control with milk attitude question.

Forest
plot showing SMD in brand or product attitudes between digital and non-digital advertising exposure and no advert or neutral advert controls; 95% CIs and study weights are indicated.Overall SMD was generated by a random effects model.Matthes (1) brand attitude outcome; Matthes, (2) product attitude outcome; Royne (1) data from cola product placement versus control with cola attitude question; Royne (2) Data from juice product placement versus control with juice attitude question; Royne (3) data from milk product placement versus control with milk attitude question.

TABLE 1
Descriptive Summary of the Included Experimental Studies

TABLE 1 Continued
NS, not stated.a Half of the sample may be reported in both.b Same sample but reporting of different outcomes.c May be the same participants across all 3 studies.d May be the same participants across the 2 studies.e Three out of the 4 schools may be reported in both.