Political Advertising on Facebook During the 2022 Australian Federal Election: A Social Identity Perspective

AI summary

Overview

Research area: Computational social science / digital political advertising, combining large-scale analysis of Meta's Ad Library data with political psychology (Social Identity Theory).

Technical level: Intermediate. The paper is readable without deep technical background, but it relies on statistical testing (Mann-Kendall, KPSS, Pearson/Spearman correlations) and a multi-stage data curation pipeline.

Scope (1 sentence): The paper analyzes 56,224 Meta ads collected from March 1 to June 1, 2022 around the 2022 Australian federal election to describe spending, timing, targeting and message themes of the major parties, and interprets those patterns through Social Identity Theory.

What This Paper Is About

Most research on social-media political advertising comes from countries with voluntary voting, where mobilizing supporters to turn out is a central campaign goal. Australia has compulsory voting, which shifts campaign strategy away from "get out the vote" and toward persuasion of an electorate that will vote regardless. The authors build and analyze a Meta Ad Library dataset for the 2022 Australian federal election to document what parties actually did — how much they spent, when, whom they targeted and what they said — and to explain those choices using three psychological accounts of persuasion: material self-interest, automatic (non-deliberative) responses, and social identity.

Key Contributions

1. Curated a dataset of political advertisements from the 2022 Australian federal election campaign, collected at six-hour intervals from March 1 to June 1, 2022, yielding 56,224 ads from 1,549 sponsors, and released three derived datasets (PA, PAK, PAP) along with code.
2. Had a political science expert annotate all potential sponsors and page names with party and electorate affiliation, with a second expert independently annotating 10% of the data and a resulting Cohen's κ of 0.906.
3. Produced a general campaign-level analysis comparable to studies conducted in other countries, covering spending, reach, temporal dynamics, demographics, geography, keywords and targeting behavior.
4. Framed the observed advertising patterns through Social Identity Theory in the distinctive setting of compulsory voting, arguing that major and smaller parties pursued different identity-based persuasion strategies.

Main Findings

• Dominance of the two major parties: The Australian Labor Party (ALP) and the Liberal Coalition collectively accounted for over 70% of total ad impressions and spending. The ALP invested A$5M (40.2% of total political ad spending) across 13,047 ads; the Liberal Coalition spent A$3.9M (31.3%) across 9,364 ads. These two parties secured 68.27% of electors' first preference votes on election day.
• Overall scale: Party-sponsored ads generated over 714M impressions. The broader keyword-filtered dataset (47,503 ads from 1,438 sponsors) produced over 1.11B impressions Australia-wide with a total ad spend of A$18.4M.
• Highly skewed ad-level performance: Mean impressions per ad were 23,452 versus a median of 4,500 (max 1,000,000); mean spend was A$387 versus a median of A$49.5 (max A$137,500); mean duration 8.1 days versus a median of 6 (max 89); mean efficiency 67.5 impressions per A$ versus a median of 50 (max 4,319). A single ad exceeded A$137,500 in spend.
• Spending predicted reach: At the ad level, spend–impressions correlation was R = 0.72 (R² = 0.52), ρ = 0.82 (strong), while spend–duration and impressions–duration were weak (R = 0.10 and 0.17). Over time (after differencing), spend–impressions correlation reached R = 0.94 (R² = 0.89), ρ = 0.93 (very strong), with ad count–impressions at 0.83 and spend–ad count at 0.76.
• Ad activity surged toward election day: Activity increased notably from April 10, the day the election was called, and intensified toward a sharp peak on May 18, just before the mandated media blackout. Mann-Kendall confirmed a significant upward trend in spend and impressions (τ = 0.925, p = 8.86 × 10⁻³⁵).
• Party-level reach and efficiency: The ALP accounted for 40.9% of party ad impressions and the Liberal Coalition just over 36%. The UAP spent A$1.578M on 416 ads for 37.3M impressions, giving the highest cost per impression among major parties (A$42.2) and a 4.12% vote; the Greens spent A$0.6M on 2,088 ads for 31.8M impressions at A$19.1 CPI and a 12.25% vote.
• Demographic targeting: Most parties targeted the 18–24 and 25–34 age groups. The Greens focused particularly on 18–24 (and on young women aged 18–24); the UAP spread impressions more evenly, emphasized older users, showed minimal engagement with the youngest group, and skewed male in the 18–54 range. The ALP reached more women, especially those over 65. The Liberal Coalition and Independents maintained balanced gender reach.
• Geographic concentration mirrored population: All parties focused strongly on Queensland, New South Wales and Victoria, home to around two-thirds of the population. The top-spend state was Queensland for Labor, the Liberal Coalition and the UAP; New South Wales for Independents; Victoria for the Greens. The UAP gave Queensland 39.4% of its impressions — the highest regional share for any party — followed by New South Wales (28.3%), Western Australia (11.7%) and Victoria (11.5%).
• Distinct messaging themes: Major parties spent heavily promoting their own names and targeted rival party names (for example ALP targeting "scott_morrison"; UAP targeting "liberal" and "labor"), consistent with negative campaigning. Smaller parties emphasized issues: the Greens devoted 50.03% of their ad spend to anti-mining messaging, peaking at 82% in Western Australia and 81% in Tasmania, while the UAP was the party most focused on positive mining-related advertising at 4.36% of its spend. Independents also used climate-related keywords.
• Targeting was relatively simple: 84.7% of ads were aimed at a single Australian region (roughly 80% state-specific, 20% broadly distributed), with regional and national ads showing similar cost per thousand impressions (A$21.4); the UAP stood out at 1.9x for targeted and 2.9x for broad ads. 85.6% of ads targeted all six adult age intervals, generating 87.8% of impressions. Only 4.8% of ads were shown exclusively to one gender, most of those to females (62.7%); under a looser 60% threshold, 14.2% targeted females and 12.8% males.

Methodology in Plain English

The researchers queried the Meta Ad Library API every six hours between March 1 and June 1, 2022 to capture ads running around the election. Each ad record includes creation time, target demographic distributions, text, page name, sponsor, URL, and impression and spending ranges; the authors used the mean of the upper and lower bounds for spend and impressions, and estimated impressions by age, region and gender by multiplying mean impressions by the share of reach attributed to each category. Because Meta relies on advertisers self-declaring political ads, they added two orthogonal filters. First, a keyword filter: starting from the seed terms "government" and "election", they used snowball sampling and two-author agreement to build a final list of 43 keywords, producing the PAK dataset of 47,503 ads and excluding 111 sponsors. Second, a party-affiliation filter: a political science expert annotated all potential sponsors and page names with party and electorate, with a second expert checking 10% of the data (Cohen's κ = 0.906), producing the PAP dataset of 30,767 ads from candidates contesting the election. The 13–17 age group and users with unknown gender were excluded because they accounted for only 0.1% and 1.1% of impressions respectively. Analysis then covered correlations between ad-level metrics, time-series analysis with a 3-day moving average plus stationarity checks and differencing, and breakdowns by age, gender, region, keyword and targeting type.

Why This Matters

The study provides one of the few systematic empirical accounts of digital political advertising in a compulsory-voting democracy, a setting where the usual turnout-mobilization logic does not apply and persuasion becomes the central strategic problem. It also shows how combining platform ad data with political psychology can generate testable explanations for observed targeting choices, rather than only descriptive counts.

Real-world applications:

• Regulatory oversight: The paper's dataset and party-affiliation list give electoral commissions and watchdogs a reproducible baseline for auditing spending, timing and geographic targeting.
• Platform transparency policy: The findings feed directly into debates about Meta's reliance on voluntary self-declaration, citing earlier work estimating that 92.3% of undeclared political ads go undetected.
• Campaign practice: The documented relationship between spend and impressions (R = 0.94 on differenced series) shows how reliably audience scale can be bought and calibrated on the platform.
• Media and civil-society monitoring: The keyword and thematic breakdowns let journalists trace which issues (for example mining or climate) individual parties pushed and where.

Industry relevance: the analysis bears on ad-platform transparency tooling, political consulting and ad buying, and the design of disclosure regimes, since the entire measurement depends on what the API exposes and on self-declared "paid by" flags required under Australian law.

Future Directions

• The paper's own discussion, limitations and conclusion sections are not included in the supplied content, so its explicitly stated future work is not reported here; the following are open questions the presented material raises.
• Quantifying the size of the undeclared-ad gap in Australia specifically, given the cited estimate that 92.3% of undeclared political ads are not detected by Meta's systems.
• Building tools to detect advanced targeting tactics that the Meta API may obscure, since the authors note that the observed behavior may reflect API limitations rather than true absence of sophisticated micro-targeting.
• Testing whether the identity-based persuasion strategies the authors infer actually changed voter attitudes or behavior, which the described analyses do not measure.
• Checking whether the pattern of major parties reinforcing partisan identity and smaller parties cultivating issue-based identities holds in other compulsory-voting systems or in later Australian elections, including the 2025 contest (not examined in this paper).

Target Audience

Political communication and computational social science researchers; election regulators and platform policy staff working on ad transparency; political journalists and fact-checking organizations covering Australian elections; and campaign strategists or analysts interested in how spending, timing, targeting and message themes interact in a compulsory-voting environment.

Authors’ abstract

The spread of targeted advertising on social media platforms has revolutionized political marketing strategies. Monitoring these digital campaigns is essential for maintaining transparency and accountability in democratic processes. Leveraging Meta's Ad Library, we analyze political advertising on Facebook and Instagram during the 2022 Australian federal election campaign. We investigate temporal, demographic, and geographical patterns in the advertising strategies of major Australian political actors to establish an empirical evidence base, and interpret these findings through the lens of Social Identity Theory (SIT). Our findings not only reveal significant disparities in spending and reach among parties, but also in persuasion strategies being deployed in targeted online campaigns. We observe a marked increase in advertising activity as the election approached, peaking just before the mandated media blackout period. Demographic analysis shows distinct targeting strategies, with parties focusing more on younger demographics and exhibiting gender-based differences in ad impressions. Regional distribution of ads largely mirrored population densities, with some parties employing more targeted approaches in specific states. Moreover, we found that parties emphasized different themes aligned with their ideologies-major parties focused on party names and opponents, while smaller parties emphasized issue-specific messages. Drawing on SIT, we interpret these findings within Australia's compulsory voting context, suggesting that parties employed distinct persuasion strategies. With turnout guaranteed, major parties focused on reinforcing partisan identities to prevent voter defection, while smaller parties cultivated issue-based identities to capture the support of disaffected voters who are obligated to participate.