lab 9 min read
LGU Health Spending Lens
What an LGU spends on health, what primary care costs, and whether it can afford the gap, on one page
- status
- Early stage: nine Central Luzon briefs live, joining PIDS health spending with DBM COMPASS NTA; open data published
- role
- Concept, data spike, design and frontend
- stack
Philippine local governments spend about PHP 850 per person on health, less than half of what basic primary care costs. The figures that show this sit on separate platforms. This project joins them into one brief per LGU, for the people who decide local budgets.
Status: early stage. The data matching has been tested on public data, and a first version of the LGU brief is available for one region. Figures and plans may change as the work develops.
At a glance
- Problem
- health spending, the cost of primary care and what LGUs receive sit on separate platforms, with no benchmark beside them.
- Idea
- an LGU brief, one page per province or city, readable on a phone and printable for budget hearings.
- Users
- local health boards and local chief executives first; residents, CSOs and journalists second; the data owners throughout.
- Data
- PIDS Health Systems Data Explorer, PIDS Discussion Paper 2026-07, DBM COMPASS, PSA PSGC codes.
- Result so far
- all 117 provinces and HUCs in PIDS's data matched to official codes, with no manual fixes.
- My role
- concept, data spike, interface design and frontend.
The problem
Under devolution, LGUs run frontline health services. Research by the Philippine Institute for Development Studies (PIDS) shows they spend far less than those services need:
- The gap is large. LGUs spend about PHP 850 per person on health. Basic primary care costs about PHP 1,828 per person.
- It is everywhere. In nearly all provinces and highly urbanized cities (HUCs), the cost of primary care exceeds current spending; most would need two to three times their current budget.
- More money did not fix it. Larger National Tax Allotment (NTA) shares after the Mandanas-Garcia ruling brought only modest gains in health spending.
- It is a choice, not only a constraint. LGUs with similar fiscal space spend very differently on health.
The data to see this now exists, but it does not meet. A council member who wants to compare their province’s health spending with what it receives must open separate platforms, find the same LGU in each, and work it out with no benchmark. A local health board cannot hold anyone accountable for a gap it cannot see.
The BetterGovPH Fellowship’s Health challenge, set by DOH, PhilHealth and NNC with PIDS as research partner, names the need: “Open local health data to improve local health board accountability.”
The idea: one brief per LGU
Each LGU brief puts four numbers side by side, each with its source and year:
- Health spending per person (PIDS explorer, 2024).
- The cost of basic primary care per person: PHP 1,828, PIDS’s conservative estimate that excludes hospital care (PIDS DP 2026-07, Table 4).
- How affordable closing the gap is: the share of the LGU’s operating income that full primary care would take (PIDS DP 2026-07, Table 5). For many provinces it is 15% to 30%, which makes the gap a question of priorities rather than blame.
- NTA (DBM COMPASS), for the same year, as context for what the LGU receives.
Each LGU is compared only with LGUs of the same type: the national median for provinces or for HUCs, and its neighbours of the same type in its region. There is no national ranking. Every brief says: For discussion, not a finding.
Why Tarlac as the example. It is the most typical case, not an extreme one. Its spending of PHP 480 per person is close to the national median for provinces (PHP 518). Its gap, 3.8 times current spending, is the exact median for Central Luzon and close to the national median for provinces (3.6 times). Picking the region’s highest or lowest spender would make the example look like praise or blame; a typical province shows what the brief does for most LGUs. One figure is above typical: full primary care would take 28% of Tarlac’s operating income, against a national median of 20% for provinces. That is shown as published, not smoothed. Every number is PIDS’s own, cited on the brief; no figure is invented for illustration.
Does this kind of tool work?
No study has tested this exact tool, but two decades of research give a clear answer: information changes outcomes when it reaches people who can act on it, and rarely when it is only published.
When information worked:
- Uganda, school grants. Only about 20% of grants reached schools in the mid-1990s. After the government published monthly transfers in newspapers and at schools, about 80% did by 2001 (Reinikka and Svensson).
- Uganda, health report cards. In a randomized trial, report cards comparing facilities, discussed in community and staff meetings, increased use of services and reduced child mortality (Björkman and Svensson, 2009).
- Wisconsin, hospitals. Hospitals whose performance was reported publicly, rather than privately, did more to improve where they scored low (Hibbard, Stockard and Tusler, 2003).
When it did not:
- The Uganda scale-up. Across 376 health centres, report cards and meetings produced no change in child mortality or use of services (EGAP Brief 58).
- The reviews. Transparency works where the chain from information to decision is short (Kosack and Fung, 2014). Tools that only release information tend to fail; those that pair it with institutions able to act do better. As Jonathan Fox puts it, “voice needs teeth to have bite” (Fox, 2015).
In the Philippines, a 2024 review by Government Watch found “no mandatory fund allocation for health under the NTA,” health boards that were often not functioning, and health funds that were hard to trace below the regional level (G-Watch, 2024).
So the design:
- Aims at the people who decide: local health boards and mayors, in budget hearings.
- Compares fairly: against the benchmark and same-type LGUs, never a national ranking.
- Keeps data owners in the loop: PIDS and DBM get data-quality flags back.
- Measures use, not views: success is boards and owners using the brief.
The spike: does the matching work?
Before proposing anything, I tested the riskiest step: whether LGUs in one dataset can be reliably matched to a shared code, so they can be joined to any other dataset.
Method. I read PIDS’s 2024 health spending table (provinces and HUCs) from its public explorer, built a list of PSGC codes from an open-source copy of the Philippine Statistics Authority’s list (June 2026), and matched them on normalized name and LGU type, so the province of Cebu never collides with Cebu City. I then cross-checked every LGU against Table 5 of PIDS’s discussion paper.
What it taught me:
- The data can be viewed but not downloaded. I found no download option, so the pipeline needs PIDS’s source file.
- Some values are rounded. The explorer rounds values above PHP 1,000. PIDS’s paper (Table 5) gives exact totals to cross-check against.
- The years line up. PIDS spending ends at 2024, and COMPASS publishes NTA from FY 2022 to FY 2027, so each brief shows FY 2024 NTA beside 2024 spending. Every figure still carries its year.
- Geography moves. The current code list moves Sulu to Region IX and adds the Negros Island Region, so regional comparisons carry their year too.
One value is worth checking with its owner: the lowest figure shown is PHP 4 per person, for a province created in 2022. That is an outlier to send back to PIDS, not a fact to publish. The COMPASS side matched too: all nine Central Luzon LGUs found their NTA by name in DBM’s public export, which has no PSGC column.
Design decisions
- Connect, do not rebuild. PIDS and COMPASS stay the authorities for their numbers; the brief only links them.
- Use the benchmark as PIDS built it. PIDS costed primary care from national standards into one uniform cost per person (DP 2026-07, section 4.1.4). Differences between LGUs come from what they spend, not from different targets.
- No “health spending as a share of NTA.” LGUs fund health from all their income, so that ratio would accuse them of something the data cannot show. PIDS measures affordability against operating income instead, and so does the brief.
- Fair comparisons, kept simple. Same-type comparisons only. PIDS found spending is only weakly linked to fiscal capacity, and the brief shows affordability directly, so finer groups are not needed.
- Every figure is traceable. Source, period and date under every number, following COMPASS’s own rule.
- Neutral by design. No rankings, no good or bad labels, and a visible way to report an error.
- Built for phones and paper. Residents use phones; budget hearings still run on paper.
How it fits together
- Pipeline: Python. One command rebuilds everything from the source files; tests block any figure without its source and date.
- Outputs: an open crosswalk with Frictionless Data and DCAT metadata, a documented read-only API, and an MCP server so AI assistants can cite each figure’s source.
- Frontend: Next.js and Tailwind CSS, static, mobile-first, with a print layout for A4.
- Feedback loop: unmatched or doubtful records go back to DBM and PIDS.
The first version is the brief only; the pipeline, API and MCP server are the plan for the fellowship.
Roadmap and pilot
In Month 1, the comparison method is agreed with PIDS, DOH and the fellowship mentors, and the brief is tested with the people it is for: a local health board or health office, a mayor’s office, and a CSO or journalist.
Choosing the pilot
Every region matched cleanly, so the pilot is chosen for where the brief can be tested well and used. The final choice is made with the challenge partners in Month 1.
Central Luzon (proposed). 7 provinces and 2 HUCs, the best mix for same-type comparisons; only one rounded value; borders unchanged; close to Manila. Full primary care would cost a median 3.8 times current spending.
CALABARZON (alternative). The largest gaps of any region near Manila: full primary care would cost a median 7.1 times current spending. Only one city.
Metro Manila’s cities (alternative). A city-to-city test where local capacity to act is highest; several cities already spend above the benchmark. The richest LGUs in the country, so the least representative.
What this is not
- Not a scorecard. It shows a gap for discussion; it does not grade LGUs.
- Not a replacement for the PIDS explorer or COMPASS. It depends on both.
- Not finished. PIDS’s source files and testing with users come next.
Related work: I helped build the frontend of Naga City’s People’s Budget Portal (budget.naga.gov.ph), an open-source participatory budgeting platform.
Sources and credits
Data
- Medina, R., Hagag, S. R., Casas, L., Banzon, J. E., Rodriguez, H., Dollente, Y., Ferrer, F.V., Banaag, M. & Ulep, V. G. (2026). Health Systems Data Explorer. Health Economics and Finance Program, Philippine Institute for Development Studies. hefp.pids.gov.ph. Values read 13 November 2026.
- Ulep, V. G. T., Herrin, A. N., Sarne, S. R., Lorenzo, P. J. M. and Calubayan, J. D. (2026). Are Philippine Local Governments Spending Enough on Primary Health Care? PIDS Discussion Paper 2026-07. doi.org/10.62986/dp2026.07. Benchmark of PHP 1,827.79 per person: Table 4 and section 4.1.4; LGU comparison: Table 5.
- DBM COMPASS: compass.gov.ph, public NTA export (FY 2022 to FY 2027), downloaded 16 November 2026
- The nine briefs’ figures as open data, each with its source, period and pull date: CSV, JSON, Data Package
- Philippine Statistics Authority, PSGC: psa.gov.ph/classification/psgc, via @aivangogh/ph-address (list as of 30 June 2026).
Coverage of the findings
- PIDS press release: New data platform reveals wide gaps in LGU health spending across the Philippines
- Inquirer, Inquirer Plus, Manila Bulletin
Evidence
- Reinikka, R. and Svensson, J. The Power of Information: Evidence from a Newspaper Campaign to Reduce Capture. World Bank Policy Research Working Paper 3239. Summary
- Björkman, M. and Svensson, J. (2009). “Power to the People.” Quarterly Journal of Economics 124(2): 735 to 769. Record
- Hibbard, J. H., Stockard, J. and Tusler, M. (2003). “Does Publicizing Hospital Performance Stimulate Quality Improvement Efforts?” Health Affairs. Record
- Parkerson, D., Posner, D. and Raffler, P. (2019). “Does Bottom-Up Accountability Work?” EGAP Brief 58. Brief
- Kosack, S. and Fung, A. (2014). “Does Transparency Improve Governance?” Annual Review of Political Science 17: 65 to 87. Record
- Fox, J. (2015). “Social Accountability: What Does the Evidence Really Say?” World Development. Record
- Government Watch (2024). Gaps and Opportunities in the Access and Quality of RMNCA Health Services in the Philippines. PDF
Challenge statement: BetterGovPH Fellowship 2026, Health challenge, fellowship.bettergov.ph
All figures belong to their publishers. Any errors in matching or presentation are mine.