Statistically significant trends and policy recommendations
Time series from all three editions are merged and tested with linear regression and Mann–Kendall, with a one-step forecast; district-level spatial analysis by year on a map; and cross-edition data-quality checks.
01Summary of tests
20Series analysed
4Significant increase (p < 0.05)
7Significant decrease (p < 0.05)
9No significant trend
Headline findings
Births↓ Significant decrease
From 2557 to 2567 (11 points) the value moved from 99,618 to 52,237 persons (-47.6%) · linear slope -5,223 persons/year (p < 0.001, R² = 0.97) · confirmed by Mann–Kendall
Natural increase (births − deaths)↓ Significant decrease
From 2557 to 2567 (11 points) the value moved from 57,542 to 2,317 persons (-96.0%) · linear slope -6,311 persons/year (p < 0.001, R² = 0.93) · confirmed by Mann–Kendall
Registered population↓ Significant decrease
From 2562 to 2568 (7 points) the value moved from 5,666,264 to 5,319,580 persons (-6.1%) · linear slope -48,674 persons/year (p < 0.001, R² = 0.93) · confirmed by Mann–Kendall
Cumulative HIV cases in Bangkok↑ Significant increase
From 2562 to 2566 (5 points) the value moved from 88,183 to 103,852 cases (+17.8%) · linear slope 3,883 cases/year (p = 0.001, R² = 0.98) · confirmed by Mann–Kendall
Working older persons↑ Significant increase
From 2562 to 2566 (5 points) the value moved from 251,615 to 350,638 persons (+39.4%) · linear slope 25,544 persons/year (p = 0.001, R² = 0.98) · confirmed by Mann–Kendall
Primary students↓ Significant decrease
From 2562 to 2567 (6 points) the value moved from 197,459 to 174,770 students (-11.5%) · linear slope -4,799 students/year (p = 0.003, R² = 0.91) · confirmed by Mann–Kendall
Chao Phraya Express Boat passengers↓ Significant decrease
From 2558 to 2567 (10 points) the value moved from 13,205,340 to 7,264,336 passengers (-45.0%) · linear slope -1,244,983 passengers/year (p = 0.011, R² = 0.58)
Deaths↑ Significant increase
From 2557 to 2567 (11 points) the value moved from 42,076 to 49,920 persons (+18.6%) · linear slope 1,089 persons/year (p = 0.013, R² = 0.51) · confirmed by Mann–Kendall
Age structure (population pyramid):
Population aged 60+ is 22.9% (2566) → 23.9% (2567);
ageing index 171 → 195 (per 100 children aged 0–14);
children 0–14 fell from 13.4% to 12.3%.
(the 2025 edition repeats the 2024 pyramid, so it is not compared)
02Trends by topic with statistical tests
Solid line = observed · dashed = linear trend · gold point and band = next-period forecast with 95% prediction interval (series with ≥ 5 points only)
👥 Population & society
Registered population
↓ Significant decrease
Latest
5,319,580 persons
vs first
-6.1%
CAGR
-1.05%/yr
Slope
-48,674/yr
p (OLS)
< 0.001
p (MK)
0.003
R²
0.93
n
7
From 2562 to 2568 (7 points) the value moved from 5,666,264 to 5,319,580 persons (-6.1%) · linear slope -48,674 persons/year (p < 0.001, R² = 0.93)
✔ Both methods agree (OLS and Mann–Kendall) — more robust
From 2558 to 2567 (10 points) the value moved from 13,205,340 to 7,264,336 passengers (-45.0%) · linear slope -1,244,983 passengers/year (p = 0.011, R² = 0.58)
⚠ The last 5 points move opposite to the overall trend — interpret with care
⚠ The series has a COVID-19 break (2020–21) — a straight line may not fit
Choose a topic, indicator and year to colour the 50-district map (Leaflet + OpenStreetMap) — compare years, normalise per population/area, and test spatial clustering with Moran’s I.
04Policy recommendations
1. Plan for a stable-to-ageing population
Births are falling significantly, the elderly share is rising and natural increase is almost nil. Shift from expanding services with headcount to improving quality, long-term and community care, and retaining working-age residents with affordable housing, transport and good jobs.
2. Right-size schools and teachers to fewer children
Primary enrolment is falling significantly while kindergarten rebounded slightly in 2024. Use the district map to find fast-shrinking areas, repurpose spare classrooms as learning, early-childhood or elderly-care centres, and adjust teacher allocation without hurting quality.
3. Workforce planning for retirement and attrition
Attrition is volatile with no clear trend (2022 was unusually high), yet synchronised retirements remain a structural risk. Prepare succession plans, knowledge transfer and digital automation, and audit the permanent-employee data (2024 retirements look abnormally low).
4. Support working elders and chronic disease
Working older persons rose significantly (data to 2023) and cumulative HIV cases keep growing. Promote safe work and income security, expand proactive health services and monitor high-risk groups (both series have only 5 points; interpret with caution).
Solid waste fell significantly over a window that includes COVID-19, so watch for a rebound as the economy recovers and build on food-waste separation. River-boat ridership is still far below pre-COVID levels: strengthen feeder links, piers and waterway tourism.
6. Target resources with district maps
Compare districts on per-capita/per-area indicators (not raw counts) and read Moran’s I to see whether a problem clusters in neighbouring districts — if so, use zone-level measures; if it is scattered, use district-specific ones.
7. Improve statistical quality and continuity
State the reference year of every figure and validate year labels before release — the 2025 edition has migration and birth–death tables whose year labels are shifted by one year.
Separate tables that are updated to the latest year from those carried over (several 2025-edition tables are identical to 2024).
Keep a revision log: many series were revised between editions.
Publish multi-year district series as open data to enable district-level trend analysis and forecasting.
05Data quality and method
Data caveats
จำนวนเกิด — table 3-C edition 2568: values equal the previous year of the earlier edition (year label shifted), so they were excluded
จำนวนตาย — table 3-C edition 2568: values equal the previous year of the earlier edition (year label shifted), so they were excluded
การเพิ่มตามธรรมชาติ (เกิด − ตาย) — table 3-C edition 2568: values equal the previous year of the earlier edition (year label shifted), so they were excluded
การย้ายสุทธิ (ย้ายเข้า − ย้ายออก) — table 3-B edition 2568: values equal the previous year of the earlier edition (year label shifted), so they were excluded
Students and classrooms, BMA schools, academic year 2025 — table 4-1: the 2025 edition is identical to 2024 (not updated)
Need for caregivers and main caregivers — table 5-6: the 2025 edition is identical to 2024 (not updated)
Population by age and sex, 2025 (population pyramid) — table 3-2: the 2025 edition is identical to 2024 (not updated)
Deaths by age group and sex, 2025 — table 3-22: the 2025 edition is identical to 2024 (not updated)
In-migration and out-migration, 2020–2025 — table 3-B: the 2025 edition is identical to 2024 (not updated)
Births and deaths over 10 years (2015–2025) — table 3-C: the 2025 edition is identical to 2024 (not updated)
Communities by type, 2025 — table 3-5: the 2025 edition is identical to 2024 (not updated)
Series revised between editions (newer value used): ร้อยละประชาชนอายุ 6 ปีขึ้นไปที่ใช้อินเทอร์เน็ต (รายไตรมาส) (2)
Method
Series come from year/quarter tables in all three editions; the newer edition wins on conflicts.
Significance: linear regression (slope t-test, df = n − 2) and Mann–Kendall (distribution-free); “significant” when p < 0.05, “weak trend” when p < 0.10.
One-step forecast from the regression line with a 95% prediction interval — a guide, not a certainty.
Map: Moran’s I (6 nearest-neighbour weights from district centroids, 999-permutation test).
Limits: short series (5–11 points) mean low power, serial correlation and the COVID-19 break (2020–21) — “not significant” does not mean “no change”.