Unhealthy Food and Beverage Marketing on Public Transport Infrastructure in Western Melbourne

ABSTRACT

Background

Unhealthy and ultra‐processed food (UPF) consumption leads to dietary pattern changes, a deterioration of diet quality, and increases the risk of chronic diseases. Marketing of UPF on government‐owned assets, such as public transport infrastructure, contributes to their repeated exposure to the public, increasing consumption.

Objective

To quantify food and beverage marketing on public transport infrastructure in Western Melbourne, assess the proportion of advertising for unhealthy and ultra‐processed food, and explore the relationship between unhealthy advertising and socioeconomic disadvantage.

Methods

We performed a cross‐sectional study at all bus stops, tram stops and train stations in three local government areas (LGAs) in Western Melbourne. We grouped food advertisements using a modified classification from existing protocols and performed descriptive analysis by LGA and transport type. We evaluated the relationship between the proportion of unhealthy advertisements in each area and the index of relative socioeconomic disadvantage using Spearman's correlation.

Results

Of 693 total advertisements, 229 promoted food or drinks. Three‐quarters (174/229) advertised unhealthy products, most commonly sugar‐sweetened beverages (49%) and unhealthy meals (28%). Overall, 73% of food advertisements promoted ultra‐processed foods. Vending machines on train platforms accounted for 40% of all unhealthy advertisements. We found no statistically significant relationship between the proportion of unhealthy advertisements and socioeconomic disadvantage.

Conclusions

Unhealthy and ultra‐processed food marketing dominates public transport food advertising in Western Melbourne. Advertising on public transport has a meaningful impact on communities' exposure to unhealthy foods and the risk of obesity and chronic disease.

1. Introduction

Overconsumption of unhealthy, discretionary, and ultra‐processed food (UPF) is a key driver of the global obesity epidemic [1, 2]. UPFs are manufactured with low‐cost ingredients, contain little to no whole foods, are convenient (ready‐to‐consume, almost imperishable), highly palatable, and designed to compete with freshly prepared meals and dishes, maximising industry profit [3]. UPFs are usually nutrient‐poor and energy‐dense, contributing to both obesity and malnutrition [4].

Ultra‐processed and other discretionary foods with little nutritional value make up a large proportion of Australian diets [1, 5]; for example, UPF accounts for up to 54% of children's and adolescents' daily energy intake [6]. UPF consumption is higher among Australian adults with lower education and those living in areas of high socioeconomic disadvantage [7].

There is strong evidence that UPF consumption leads to dietary pattern changes, a deterioration of diet quality and increased risk of chronic diseases [8]. An umbrella review of 14 meta‐analyses found direct and consistent associations between greater exposure to UPF and higher risks of morbidity and mortality from cardiovascular diseases, type 2 diabetes, several cancers, obesity, and mental health disorders [9].

Advertising on and around public transport represents a significant source of potential exposure to unhealthy and ultra‐processed food and drink promotion for a large and diverse population. Outdoor advertisements (ads) such as billboards, buses, and digital displays are particularly prevalent in public spaces such as transport hubs, which receive high volumes of foot traffic, achieving repeated brand exposure [10]. Marketing of unhealthy foods increases preference for and consumption of those foods, particularly among pre‐adolescent children [11, 12, 13].

There is some evidence, although not consistent across studies, that unhealthy advertising is more prominent in areas of socioeconomic disadvantage [14, 15, 16]. In Australia, an audit of the Sydney train network showed that most food and drink ads (84%) were for unhealthy items, with a greater proportion occurring in areas with socioeconomic disadvantage [17]. A study of public transport stops across Eastern Melbourne revealed that ads for fast food chains and unhealthy drinks are more common in disadvantaged areas [18]. International studies from Sweden [19], the United States [20], and New Zealand [21] also found more unhealthy food advertising on public transport in areas of ethnic diversity or lower income [17].

The marketing of UPFs contributes to obesity and chronic diseases [11, 22], therefore regulating unhealthy food advertising represents a crucial public health strategy to mitigate diet‐related chronic diseases, with the potential to preferentially benefit people at greater socioeconomic disadvantage [23]. The experience of restricting unhealthy food advertising across London's public transport network in 2019 demonstrated a relative reduction in the salt, fat, sugar, and total energy content of unhealthy food purchases by households each week [24]. A similar regulatory approach has recently been adopted in South Australia [25].

Western Melbourne is an area of rapid population growth, demographic diversity, and nascent infrastructure development, with both new and well‐established transport hubs. Food and beverage advertising in the region has not been previously studied. We aimed to quantify the volume and content of food marketing at public transport sites in selected local government areas (LGAs) of Western Melbourne. A secondary aim was to investigate the relationship between the proportion of marketing that is unhealthy at public transit stops and the relative socioeconomic disadvantage of the surrounding area.

2. Methods

2.1. Study Design

This is a cross‐sectional study of food and beverage marketing on public transport and public transport infrastructure across three LGAs in Western Melbourne, Australia. We chose Wyndham, Hobsons Bay, and Merri‐bek based on their varied geographic, demographic, and transport infrastructure characteristics (Table S1). We selected LGAs without major commercial or commuting hubs to capture exposure for residents of the area rather than for commuters from elsewhere. We included all public transport stops and transport types (train, bus, and tram where present) in each LGA.

2.2. Definitions

We adopted a broad definition of marketing based on the World Health Organisation's (WHO) framework on the marketing of food and non‐alcoholic beverages to children: “any form of commercial communication or message that is designed to, or has the effect of, increasing the recognition, appeal and/or consumption of products and services. It comprises anything that acts to advertise or otherwise promote a product or service” [26]. This includes traditional paid advertising such as printed or digital posters, flyers, billboards, and signs, as well as less direct marketing including kiosk or shop‐front signage, store merchandising (physical products on display at the shop or within fridge windows), and displays of products within vending machine windows. In our enumeration we counted each instance of marketing as an ad.

We defined food or drink ads as either those featuring a food or drink product (as well as branding for a company that sells food or drink products), or those featuring branding for a company that sells food or drink products without a product pictured (‘brand‐only ads’). We classified ads that pictured or referred to food and drink products but did not contain food or drink branding as non‐food ads but counted these separately as potential indirect exposure to food and drink marketing.

2.3. Inclusion and Exclusion Criteria

We recorded the marketing that a typical commuter is exposed to at government‐owned public transport infrastructure including train stations, bus stops, and tram stops. This included commercially owned buildings and shops, billboards, mobile carts, and vending machines within the immediate area of public transport infrastructure. At train stations, this included the internal area (concourse, platforms, walkways, kiosks, and vending machines) and external areas (entry and exit points, walkways to and from bus interchanges, and external ads on moving buses, trams, and trains).

We excluded advertising on the inside of buses, trams, and trains. Signage that did not meet our definition of marketing was not counted, for example, government notices, public service announcements, public transport ticketing information, and ghost signage (retained historical advertising). We also excluded marketing inside kiosks (e.g., items or branding behind the counter not directly on display), or on privately owned vehicles (e.g., passing trucks and cars) and private infrastructure (e.g., phone booths and adjacent shop fronts outside the vicinity of public transport infrastructure).

2.4. Data Collection

We used mobile phone‐based REDCap surveys to record marketing information at all public transport sites. We documented: setting (e.g., outside of bus/tram, at a bus/tram shelter, along a train station platform), location (e.g., train station, bus stop number, external kiosk), format (e.g., printed poster or banner, digital alternating screen, product display, vending machine), and product category (food vs. non‐food). Food or drink marketing had further information collected including brand, product, food group classification, and whether the brand or product was categorised as ultra‐processed.

Trained personnel collected data between November 2023 and May 2024. To capture exposure to mobile ads for a typical commuter at peak hour, we observed moving buses immediately outside of train stations for 20 min between 4 pm and 6 pm on weekdays. Similarly, to capture ads on moving trams, we selected a single tram stop on each line and observed trams driving past for 10 min between 4 pm and 6 pm on weekdays. These 10 and 20‐min timeframes represent the maximum wait times for tram and train commuters in our study LGAs. We collected data on static ads at any time of the day.

2.5. Advertisement Coding

All ads and marketing displays were coded at the point of data collection, with any ambiguity in coding reviewed by the principal investigators. We broadly classified food and drink marketing as health or unhealthy and then into one of 21 food categories using a modified guide based on the Council of Australian Governments (COAG) Health Council national interim guide for food promotion [27], the INFORMAS protocol for measuring outdoor advertising near schools [28] and the Australian Dietary Guidelines [29] (Table S2). We used the NOVA classification scheme to further classify food and drink products as ultra‐processed or not ultra‐processed [30].

We based the food category on the food or drink products in the ad, even if the brand did not align with the product (e.g., fruit and vegetables on an ad for a take‐away delivery service were classified based on the fruit and vegetables rather than the take away meals the service is traditionally known for). Where ads for food or drink brands did not contain specific products, we classified according to the most prominent product on the brand's Australian website, as recommended by WHO coding guidelines [31]. The representative product was captured in a screenshot of the website and the time and date recorded. We classified brand‐only ads by food group and ultra‐processing based on their representative product in the same way as ads containing food and drink products.

Products without branding (e.g., a display of baked pastries in a kiosk window), were classified into food groups based on (Table S2). Displays of these products were classified as ‘not appliable’ (N/A) for ultra‐processing, given the lack of ingredient list.

To ensure consistency, we agreed upon methods for counting ads and marketing prior to data collection. We counted each physical panel of advertising as a single ad, even if it was for the same product (e.g., all three sides of a vending machine), and counted each ad separately on alternating or scrolling digital marketing displays. At kiosks or vending machines with several products on display, we counted each display as a single form of marketing and classified accordingly (e.g., a display of sandwiches, donuts, pastries and fried food was classified as one “unhealthy meals” ad). Printed ads displayed at kiosks (e.g., signage for the kiosk's products), and signs that promoted the kiosk itself were counted individually. Finally, we counted ads on external buses or trams individually. Ads covering an entire vehicle (bus or tram ‘wrap’) were counted and classified as 1 ad.

2.6. Quality Assurance

Three rounds of quality assurance took place. First, a subset of data collectors blind‐coded a random selection of food and drink marketing and compared it with the original classification. Discrepancies highlighted a small subset of ads with complex classifications, which were reviewed by a decision panel of senior staff who also tested and refined the classification protocols. Finally, two study investigators (MZ and FA) audited the full dataset. Investigators tracked and logged all changes to classifications.

2.7. Data Analysis

Raw data were managed and cleaned in Microsoft Excel by the study investigators. We described findings at individual LGA level and for all LGAs combined.

We assessed the relationship between unhealthy advertising and socioeconomic disadvantage at bus stops in Merri‐bek and Hobsons Bay. Train stations were excluded because assumptions about the residential area of train patrons could lead to misclassification bias. Tram stops were excluded as there were too few unhealthy ads observed to include them in this analysis. We classified bus stops by index of relative socio‐economic disadvantage (IRSD) [32] quintile according to the Statistical Area 2 (SA2) [33] in which they were located, with Quintile 1 representing the most disadvantaged SA2s and Quintile 5 the least disadvantaged. The relationship between the proportion of unhealthy ads at bus stops and the IRSD quintile of the surrounding SA2 was tested with Spearman rank correlation coefficient, using R statistical software (v. 4.4.0) [34].

3. Results

We identified a total of 693 ads across 1993 bus stops, 121 tram stops, and 29 train stations in the three LGAs. Table 1 shows results stratified by food group and LGA. One‐third (229/693, 33%) of ads marketed food and drinks, three‐quarters of which (174/229, 76%) marketed unhealthy food or drinks, including alcohol. The proportion of food ads marketing unhealthy foods was similar across all three LGAs (73%–78%).

TABLE 1.

Characteristics of ads by LGA.

Ad type		Wyndham		Hobsons bay		Merri‐bek		All LGAs
		n	%	n	%	n	%	n	%
Unhealthy	Sugar‐sweetened drinks	37	18%	18	9%	30	10%	85	12%
	Artificially sweetened drinks	2	1%	2	1%	7	2%	11	2%
	Confectionary	1	0%	0	0%	0	0%	1	0%
	Savoury snacks	2	1%	0	0%	0	0%	2	0%
	Sweet snacks	0	0%	0	0%	0	0%	0	0%
	Desserts, ice creams and confections	6	3%	3	2%	0	0%	9	1%
	Unhealthy Meals—packaged or fast foods	19	9%	23	12%	6	2%	48	7%
	Processed meat	0	0%	1	1%	0	0%	1	0%
	Alcohol	2	1%	1	1%	14	5%	17	2%
	Total (unhealthy)	69	33%	48	26%	57	19%	174	25%
Healthy	Breads and cereals	4	2%	0	0%	0	0%	4	1%
	Fruits	0	0%	1	1%	0	0%	1	0%
	Vegetables	0	0%	0	0%	0	0%	0	0%
	Dairy products & alternatives	6	3%	0	0%	6	2%	12	2%
	Meat & alternatives	1	0%	0	0%	0	0%	1	0%
	Bottled water	1	0%	0	0%	5	2%	6	1%
	Healthy meals	8	4%	10	5%	2	1%	20	3%
	Total (healthy)	20	10%	11	6%	13	4%	44	6%
Miscellaneous*		5	2%	3	2%	3	1%	11	2%
Total food and drink		94	45%	62	33%	73	25%	229	33%
Non‐food ads		114	55%	126	67%	224	75%	464	67%
Total ads		208	100%	188	100%	297	100%	693	100%

Location		Wyndham			Hobsons bay			Merri‐bek			All LGAs
		Unhealthy (n)	Total food (n)	Total (n)	Unhealthy (n)	Total food (n)	Total (n)	Unhealthy (n)	Total food (n)	Total (n)	Unhealthy (n)	Total food (n)	Total (n)
Bus	Bus exterior	16	38	135	2	5	48	0	3	32	18	46	215
	Bus stop	0	0	0	26	33	94	21	30	116	47	63	210
	Bus total	16	38	135	28	38	142	21	33	148	65	109	425
Train station	Vending machines	21	22	22	19	19	19	30	33	33	70	74	74
	Kiosks	29	30	30	1	1	1	0	0	0	31	31	31
	Printed signage	1	1	11	0	6	26	2	2	52	3	9	89
	Other‡	2	3	10	0	0	0	0	0	3	2	3	13
	Train station total	53	56	73	20	26	46	32	35	88	107	117	207
Tram	Tram exterior	N/A	N/A	N/A	N/A	N/A	N/A	3	3	39	3	3	39
	tram stop	N/A	N/A	N/A	N/A	N/A	N/A	1	2	22	1	2	22
	Tram total	N/A	N/A	N/A	N/A	N/A	N/A	4	5	61	4	5	61
Total ads		69	94	208	48	64	188	57	73	297	174	231	693

^*Miscellaneous includes tea and coffee, recipe additions, supplements, and baby food.

^‡ Other includes external train advertising or alternating digital screens.

Nearly half (85/174, 49%) of unhealthy food advertisements were for sugar‐sweetened beverages, more than any other category, whereas few (11/174, 6%) advertised artificially sweetened drinks. The most common unhealthy food group advertised in Wyndham and Merri‐bek was sugar‐sweetened beverages (37/94, 39% and 30/73, 41% respectively); in Hobsons Bay, the most common unhealthy food group advertised was unhealthy meals (23/62, 37%). Alcohol ads constituted a tenth (17/174, 10%) of all unhealthy food marketing; nearly all of these were in Merri‐bek (14/17, 82%). A small number of ads (19/229, 8%) featured an unhealthy food or drink brand but no specific product (‘brand only’ ads); most of these advertised sugar‐sweetened beverage companies (17/19, 89%).

Almost three quarters (167/229, 73%) of food and drink ads advertised UPFs, most of which we classified as unhealthy (158/167, 95%). Half (84/167, 50%) of UPF ads were for sugar‐sweetened beverages; over a quarter advertised unhealthy meals (43/167, 26%). The small proportion of ads for UPFs that were classified as a healthy food group (7/167, 4%) advertised items such as fortified breakfast cereal or plant‐based milks.

Vending machines located on train platforms accounted for the highest proportion of unhealthy food ads (70/174, 40%), followed by bus stops (47/174, 27%) (Table 1). In Merri‐bek, vending machine marketing accounted for more than half (30/57, 53%) of all unhealthy marketing.

The type of public transport infrastructure and associated marketing varied across the LGAs. Unhealthy ads at bus stops were most common in Hobsons Bay (26/48, 54%). In contrast, Wyndham had no ads at bus stops. In Merri‐bek, we recorded ads on the exterior of trams (39/297, 13%) and at tram stops (22/297, 7%), however tram infrastructure rarely promoted food or drink (4/61, 7%). In Wyndham, we found 42% of all unhealthy marketing (29/69) on kiosks located across three train stations, demonstrating a high density of unhealthy ads in these areas. Hobsons Bay had only one train station kiosk, while Merri‐bek had none.

Healthy products comprised less than a fifth (44/229, 19%) of all food and drink ads, with healthy meals the most common group advertised (20/44, 45%), followed by dairy products (12/44, 27%) and fruit (1/44, 2%). We saw no vegetable ads. Food and drinks incidentally featured in 14 ads not marketing food or drink products or brands (2% of all ads).

The proportion of unhealthy ads by IRSD quintile is shown in Table 2. Bus stops in Quintile 2 (areas with more relative socioeconomic disadvantage than the Australian average) displayed the most unhealthy ads. We found the fewest ads overall, and only one unhealthy ad, in the most disadvantaged areas (IRSD quintile 1).

TABLE 2.

Unhealthy ads by IRSD quintile.

IRSD quintile	SA2s	Total population	Total ads (n)	Unhealthy ads (n)	Unhealthy (%)
1	2	28 785	24	1	4
2	2	21 172	65	18	28
3	3	32 419	44	7	16
4	5	72 009	67	6	9
5	5	70 776	70	5	7
Total	17	225 161	270	37	14

Given the small number of unhealthy ads in Quintile 1, we excluded this quintile from statistical analysis. We found no statistically significant relationship between socioeconomic disadvantage and the proportion of ads that advertised unhealthy food (Spearman's correlation = −1, p = 0.083). A valid confidence interval could not be calculated because of the perfect ordinal association between the two variables.

4. Discussion

Three in four food and drink ads observed in this study of public transport infrastructure in Western Melbourne advertise unhealthy products or brands, and 73% advertise UPFs. These results align with previous Australian and New Zealand studies, which have generally focused on children's exposure to unhealthy food marketing near schools [17, 23, 24, 35], rather than the whole transport network. The dominance of sugar‐sweetened beverage advertising on vending machines is consistent with findings elsewhere [17, 18, 19, 21, 36]. Given the prevalence of vending machines at train stations, restrictions on their placement or changes to the products they advertise and sell would represent a sizable reduction in unhealthy food and drink exposure.

While the proportion of food marketing that was unhealthy was consistent across the LGAs, public transport infrastructure and location of ads varied. The prominence of kiosks and associated advertising at major train stations in Wyndham may reflect the lack of retail food outlets nearby. In contrast, major stations in Hobsons Bay and Merri‐bek are in established commercial areas with more retail food options. This suggests that people in some areas are limited to a narrow range of food and drink options within the station itself and are exposed to more intensive advertising in that context. These variations in marketing relative to the local commercial environment highlight the challenges of reducing unhealthy marketing at a local level.

Food advertising at bus stops is common in Hobsons Bay and Merri‐bek, but non‐existent in Wyndham. Discussions with Wyndham City Council highlighted possible reasons for this, including a local government ordinance restricting visual clutter and administrative barriers to obtaining bus shelter advertising permits. Further examination revealed few options available to local government to deliberately restrict unhealthy food advertising. Council can regulate advertising on council‐controlled assets or via permits for use of facilities, but this would not capture public transport infrastructure. An ordinance in the Victorian Planning Provisions requires that councils consider the impact of advertising signs on the amenity of the area, but not their content [37]. While one could argue that unhealthy food advertising detrimentally impacts the amenity of an area by undermining local health promotion policies, there is no precedent for this interpretation. Other research on the capacity of local governments to create healthier food environments has similarly found that they lack the policy levers and authorising environment to reduce exposure to unhealthy products for their residents, either via limiting advertising or by restricting unhealthy retail outlets [38, 39].

Tram and tram stop ads were common in Merri‐bek, the only LGA that we included that had tram routes. However, few advertised food, drinks, or alcohol. This contrasts with frequent unhealthy food advertising at bus stops in the area. Each transport service provider has contracts with the state government, which include clauses to limit offensive or inappropriate ads and, in some cases, prohibit alcohol advertising [40]. However, commercial‐in‐confidence precludes public access to current contracts. The tram service provider is the only provider with an advertising content review guideline [41] which, while it does not explicitly mention food advertising, acts as an additional governance mechanism that could explain the discrepancy between trams and other modes of transport.

We did not observe a statistically significant relationship between socioeconomic disadvantage and unhealthy marketing in our study; nor did others in Adelaide and the UK [35, 42, 43]. In contrast, studies in Sweden [19], the USA [20], and Melbourne's East [18] demonstrated greater unhealthy food and drink advertising in the lowest compared to highest sociodemographic areas; all had larger sample sizes than our study.

It is likely that any true relationship is not linear across levels of disadvantage, and would no doubt be subject to other factors, such as cultural background and zoning. For example, the USA study found more unhealthy food and drink marketing in predominantly Latino areas across four cities [20]. In our study, several unmeasured factors could have contributed to the non‐significant trend towards more unhealthy food and drink advertising in more disadvantaged areas, including the presence of large retail districts with food outlets and associated local advertising, which we did observe in one Quintile 2 area.

Finally, food and drink incidentally featured in 14 non‐food ads (e.g., fruit displayed on a table for a furniture ad). While few of these were unhealthy, this is an important category to consider in the context of regulation, as it represents a potential opportunity for unhealthy food marketing if only ads that directly promote food products are restricted.

5. Strengths and Limitations

This is the first study to document public transport marketing in the western suburbs of Melbourne, Victoria. A strength of our study is that we collected data from every public transport site across the study area and examined all sources of food marketing (e.g., product displays at kiosks and vending machines) to reflect exposure for a typical commuter. A limitation is that data collection occurred over several months, so it is possible that advertising campaigns changed in one LGA while we surveyed others. However, anecdotally the researchers observed many of the same ads months after data collection. We speculated that seasonal variation could influence the type of products and brands being marketed, but this was not observed.

The classification system used did not involve strict nutritional profiling as has been used elsewhere [44, 45], but rather a practical food group‐based assessment, an approach that has also been adopted in other protocols [27, 28]. This has the benefit of use by the general community in future surveys without the need for nutritional expertise, and it reflects the understanding that nutritional benefits are realised from eating an appropriate mix of whole foods rather than individual nutrients alone. However, this approach could result in classifying some products with reasonable nutritional profiles into unhealthy categories, and vice versa.

Finally, data collection at public transport sites reflects marketing at government‐owned infrastructure but does not reflect all forms of advertising a commuter is typically exposed to, for example, advertising on roadside billboards and in association with retail precincts.

6. Conclusion

This study provides valuable local data regarding food marketing on public transport infrastructure in Western Melbourne. It shows that ads for unhealthy and UPFs dominate the food marketing landscape, creating regular exposure for daily commuters, school children, and other public transport users. This reflects experiences elsewhere [17, 18, 35, 36], showing that advertising on public transport is likely to have a meaningful impact on the consumption of unhealthy foods and the risk of obesity and chronic disease.

Action to control advertising could be taken at the local or state level; however, opportunities for local government to modify local planning laws are impractical and potentially subject to legal challenge. State‐level action, as undertaken in other jurisdictions [25], offers a potentially effective and efficient way to reduce unhealthy marketing on public transport.

Funding

The authors have nothing to report.

Ethics Statement

Ethics approval was obtained from the Western Health Office for Research (reference number QA.2023.96). No personal or identifying information was collected. All data was collected in public spaces.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Characteristics of selected local government areas.

Table S2: Classification of ads which are for food and drink brands and/or products.

Zsori M., Gardiner T., Boelsen‐Robinson T., et al., “Unhealthy Food and Beverage Marketing on Public Transport Infrastructure in Western Melbourne,” Health Promotion Journal of Australia 37, no. 3 (2026): e70205, 10.1002/hpja.70205.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.