Taiwan campaign case study: how 12 policy messages were prioritised
SPI Political Knowledge Lake | Campaign message-testing case
Study note: This article uses 1,003 completed responses received by 21 August 2026. Fieldwork was still open at export, so the results are an interim snapshot. The analysis compares the relative performance of 12 tested messages within this sample. It is not voter support, candidate vote share or an election forecast.
Campaigns rarely suffer from having no message. More often, they try to say too much—or frame an issue in the wrong way. The audience then fails to understand, remember, identify with or believe it.
This case shows how a team with 12 different policy messages can use one comparative design to identify a priority development pool, messages that need strengthening, and messages that should not lead for now.
How this case differs from the commercial offer: SPI’s proprietary study compared 12 messages to demonstrate the full ranking method. A client engagement normally concentrates one issue into the 8–10 frames and proof points with the highest decision value, rather than spreading sample across lower-value options.
The answer first: three groups, but no single winner
Develop now
Childcare for shift workers | Night care + staffing
Healthcare for young children | Fee relief for ages 0–6
Senior care | Denture subsidy up to NT$50,000
Youth development | Integrated employment + entrepreneurship support
Strengthen next
Premature-baby care | NT$30,000 family allowance
Senior citizen card | Flexible use of 300 points
Do not lead with yet
Indigenous housing | Social-housing support
Major transport | Rail + express road
Parenting convenience | Pregnancy transport + temporary childcare
Industry and jobs | AI + drone corridor
Cross-party governance | Taipei–New Taipei coordination
Rural transport | Reservation + point-to-point service
The highest-scoring message was “Childcare for shift workers | Night care + staffing.” Yet it ranked first in only 64.15% of 2,000 bootstrap resamples—above SPI’s directional threshold but below the 80% threshold required to call the lead established. The correct decision is therefore not to declare that one frame “won.” It is to develop the four leading messages as the next-round priority pool, while strengthening the middle group for the next stage of competition.
Results for all 12 messages
The score is the number of times a message was selected “most likely to attract attention/support” minus the number of times it was selected “least likely,” divided by its designed exposure count. It is a relative priority measure within this set—not support, vote share or persuasion rate.
| Message frame | Decision | Summary shown to respondents | Score |
|---|
| Childcare for shift workers | Night care + staffing | Develop now | Increase caregiver support and offer night-time childcare, helping shift-working families combine work and parenting. | +0.1698 |
| Healthcare for young children | Fee relief for ages 0–6 | Develop now | Waive co-payments for outpatient and inpatient care for children aged 0–6, reducing healthcare pressure on families. | +0.1579 |
| Senior care | Denture subsidy up to NT$50,000 | Develop now | Subsidise dentures for people aged 65 and older by up to NT$50,000, reducing household costs. | +0.1429 |
| Youth development | Employment + entrepreneurship | Develop now | Provide accelerators, career mentors, young-farmer loans and digital tools to support youth employment and entrepreneurship. | +0.1392 |
| Premature-baby care | NT$30,000 family allowance | Strengthen | Provide a NT$30,000 care allowance to families of premature babies for additional costs after birth. | +0.1130 |
| Senior citizen card | Flexible use of 300 points | Strengthen | Let 300 senior-card points be used more flexibly for healthcare, medicine and other daily needs. | +0.1040 |
| Rural transport | Reservation + point-to-point service | Do not lead with yet | Expand community buses, digital reservation and point-to-point service for school and healthcare trips in remote areas. | +0.0768 |
| Industry and jobs | AI + drone corridor | Do not lead with yet | Create an AI and drone industry corridor and more industrial space to generate jobs for young people. | −0.0123 |
| Parenting convenience | Pregnancy transport + temporary childcare | Do not lead with yet | Extend pregnancy transport, pilot free temporary childcare at sports centres and expand public childcare. | −0.0273 |
| Major transport | Rail + express road | Do not lead with yet | Extend underground rail from Shulin to Yingge and advance Dahan River MRT and the Tamsui–Taipei express corridor. | −0.1941 |
| Cross-party governance | Taipei–New Taipei coordination | Do not lead with yet | Promote KMT–TPP joint governance and coordination between Taipei and New Taipei on transport, education, culture and the environment. | −0.3084 |
| Indigenous housing | Social-housing support | Do not lead with yet | Accelerate Indigenous social housing in Shulin’s Nanyuan area to support housing needs in a defined community. | −0.3616 |
The strategic value is the next decision—not the ranking itself
Three decisions a team can make
- Turn the four leading items into a Message House. Identify the common theme, claim, proof and credible messenger.
- Strengthen the middle items. Clarify who benefits, whether delivery is feasible and what evidence supports the promise before advancing them.
- Pause large-scale use of low-priority items. Reframe or add proof before committing scarce channel resources to copy that was relatively rejected in this comparison.
Message House example: turning housing justice into four testable pillars
A ranking is only the beginning. The four leading messages named a clear beneficiary, an immediate and tangible benefit, and concrete proof. The example below applies that structure to SPI Knowledge Lake’s first policy demonstration—housing justice—and produces a next-round Message House. These four housing pillars are strategic proposals; they were not compared with each other in this study.
Roof | Common housing-justice claim
Make renting affordable and secure; make suitable homes accessible; give younger families a realistic path to ownership.
Housing justice cannot remain a slogan. Every promise must identify who benefits, how much help is offered, where and when it starts, and who is accountable.
Pillar 1 | Affordability
Reduce the monthly burden first
- Priority group
- Younger renters, families and vulnerable households facing heavy rent burdens
- Policy action
- Provide accessible, sustained rent support or burden reduction
- Proof required
- Amount, eligibility, households covered and launch date
- Decision question
- Does “how much less each month” outperform an abstract fairness claim?
Pillar 2 | Security
Protect leases, deposits and repairs
- Priority group
- Renters concerned about arbitrary increases, deposit disputes or housing quality
- Policy action
- Set explicit protections for contracts, deposits, repairs and dispute resolution
- Proof required
- Service standard, resolution time, complaint channel and responsible agency
- Decision question
- Can rights protection strengthen a rent-subsidy-only proposition?
Pillar 3 | Access
Make social housing visible—and attainable
- Priority group
- Young families, families with children and people with demonstrated housing need
- Policy action
- Accelerate social housing, managed-rental or other affordable supply
- Proof required
- Additional units, locations, completion or move-in schedule and allocation rules
- Decision question
- Can units and timing turn a policy promise into a credible outcome?
Pillar 4 | Ownership
Help first-time buyers cross the first hurdle
- Priority group
- First-time buyers with stable work but insufficient deposits or unaffordable monthly payments
- Policy action
- Offer financially viable interest, down-payment or affordable-home support
- Proof required
- Support level, price cap, eligibility, funding source and duration
- Decision question
- Can first-home support broaden agreement among younger people and families?
Foundation | Evidence every pillar needs
Beneficiaries, amount or unit count, location, eligibility, funding, launch date and accountable agency
Without these elements, housing justice remains a slogan. With them, a team can decide which pillar deserves the candidate’s time and media resources.
How to test nextGive “affordability, security, access and ownership” equal exposure in a comparative test.
What the result decidesSelect one lead pillar, one supporting pillar and the proof each still needs.
How to activateTurn the winning pillar into candidate language, beneficiary examples, website content and a four-week tracker item.
Evidence boundary: This study tested only one “Indigenous housing | Social-housing support” statement, which scored −0.3616. That exact wording ranked low within this comparison; it does not show that housing justice as a whole is ineffective. The four pillars above are proposed from the “concrete benefit + explicit proof” pattern seen in leading messages. They are a next-round test design—not validated winning messages.
This is why an “issue” and a “frame” cannot be treated as the same thing. Housing justice is the issue; affordability, security, access and ownership are four frames for the next comparison. The data can support the exact tested wording—not a sweeping claim that “housing justice works.”
Study at a glance
Research organisation: ShareParty Insight (SPI); lead: Tatt Chen
Field dates: 17–21 August 2026 (interim snapshot while fieldwork remained open)
Method: SPI Taiwan online research panel, recruited through LINE, Google, Meta and other major platforms; mobile self-completion; 12-message MaxDiff design
Sample: 1,003 complete Taiwan responses; 3,009 designed exposures per message
Analysis: Unweighted; 2,000 respondent-level bootstrap resamples to assess stability
Precision reference: approximately ±3.1 percentage points; an online-sample precision reference, not a probability-sample confidence interval
Full methodology (panel scale, responses, stability thresholds, projectability and limitations)
Research disclosure and evidence boundaries
- Commissioning and responsible research organisation: ShareParty Insight (SPI).
- Research lead: Tatt Chen.
- Field period: 17 August 2026 19:01:53 to 21 August 2026 09:31:21 (Asia/Taipei), based on start and submission times for the 1,003 complete responses used here.
- Sampling: SPI’s Taiwan online research panel. Respondents were recruited through major internet platforms including LINE, Google and Meta and completed a mainly mobile questionnaire after consenting to research participation.
- Taiwan research panel: As of August 2026, SPI had accumulated more than 60,000 Taiwan online research participants across age, gender and region. The 60,000 figure is the pool available for research—not Taiwan’s electorate or internet population.
- Responses and valid sample: By 21 August 2026, Alchemer dedicated tracking links had recorded 1,034 responses; this article analyses 1,003 complete Taiwan responses.
- Funding: SPI-funded proprietary research.
- Design: Each respondent completed 9 MaxDiff tasks, saw 4 messages per task, and each message was designed to appear 3 times.
- Analysis: Unweighted; 3,009 designed exposures per message; 2,000 respondent-cluster bootstrap resamples. The highest-scoring message ranked first in 64.15% of resamples, below the 80% threshold required to call a winner established, so SPI does not claim a unique winner.
- 95% precision reference: A simple-random-sample formula gives approximately ±3.1 percentage points for 1,003 responses. Because this is an online panel and the results are unweighted, this is only a convenient sample-precision reference—not a formal probability-sampling confidence interval.
- How to read the result: The target research population is Taiwan internet users. Once pre-specified quotas and weighting are complete, research may estimate overall trends among Taiwan internet users. This interim message test is unweighted and compares only the relative performance of the 12 messages within this sample; it does not estimate support among all Taiwan voters.
- What cannot be claimed: This is not support, vote share, a causal persuasion effect or a candidate-win forecast. The unweighted MaxDiff score is not an absolute share of Taiwan voters. Findings apply to the exact tested set of wording and cannot establish that a policy issue is inherently effective.
What a buyer should take from this case
SPI does not stop at a ranking table. We translate comparative evidence into three decisions—develop now, strengthen next, or do not lead with yet—and deliver each conclusion with its evidence status and claim boundary. That is the difference between a decision tool and a poll report.
Explore SPI Taiwan Political Intelligence, or contact sales@sharepartyinsight.com.
