The Data Space

Input data shared by every by-policy projection.

The seven input CSVs are read here, once per model, and referenced from Projection as data. Projection is parameterized by point_id, so each Projection[N] is a separate ItemSpace with its own cells cache; if the readers lived there, every model point would re-read every file. Holding them in an unparameterized Space reads each file once no matter how many policies are projected.

Inputs are external files: plain CSVs in the model folder’s parent directory, products/indemnity_medical/, rather than data stored inside the model. The model folder therefore holds nothing but formulas — no _data/, no IOSpec, no embedded values — so a diff of the model shows logic changes only. This follows annuallife.TradLife_A; contrast basiclife.BasicTerm_S, which keeps its inputs inside the model through modelx’s IOSpec machinery.

The consequence worth knowing: the model is not portable on its own. Copying the Medical_KR_S folder without its parent’s CSVs produces a model that reads and then fails on first evaluation.

input_dir() resolves the directory from _model.path.parent at run time, so the model works wherever the repository is checked out. Each table has a filename Reference and a reader Cells:

Reference	Cells	File
model_point_file	model_point_table()	model_point_table.csv
mort_table_file	mort_table()	mort_table.csv
lapse_table_file	lapse_table()	lapse_table.csv
utilisation_table_file	utilisation_table()	utilisation_table.csv
severity_table_file	severity_table()	severity_table.csv
claim_shape_file	claim_shape_table()	claim_shape_table.csv
oop_ceiling_file	oop_ceiling_table()	oop_ceiling_table.csv

Every one of these files is a [std] construction, and that is a finding

보험개발원 (Korea Insurance Development Institute) is the statutory 보험요율 산출기관 under 보험업법 제176조, and the 장기손해보험 참조순보험요율 it publishes covers 일반상해, 교통상해, 질병 사망률, 후유장해, 입원율, 암 발생률, 비용손해, 재물손해 and 배상책임 — 실손의료보험 is not among them. The 산출방법서, where an insurer’s 예정위험률 and 예정사업비율 actually live, is a 기초서류 filed under 보험업법 제5조제3호 and 제127조 and is not published. There is therefore no public Korean indemnity-medical morbidity or severity basis at all: this is a positive finding about what Korea publishes, not an unfetched document, and it is the exact boundary at which this model marks its basis [std].

What the supervisor does publish, annually and in quantity, is aggregate experience: in-force counts by generation, premium income, claims split 급여/비급여, 경과손해율 overall and by generation against a stated break-even of about 85%, claims by treatment category and by provider class, per-policy claim amounts by generation, a twelve-band claim-size distribution by generation, the 65% zero-claim mass and the top-decile concentration, the NHIS coverage ratios by provider class and age band, and the 본인부담상한제 threshold table. Every table below is anchored on those, and every row says which.

mort_table.csv

A [std] Makeham construction, q(x) = 1 - exp(-(A + B c^x)) above age 15 with a log-linear child schedule below it, fitted to the four summary statistics 국가데이터처 publishes — 기대수명 at birth, 기대여명 at 40 and at 65, and survival to age 80. It is not a transcription: 제10회 경험생명표, the industry table applied from 2024-04, is not published in full, and the single-year 완전생명표 qx tables live behind KOSIS and were not downloaded. The direction of prudence runs the wrong way here: on a one-year indemnity contract death releases the liability, so an over-statement of mortality is anti-conservative, the reverse of every protection product in this library.

lapse_table.csv

A [std] annual lapse schedule by policy year. No 실손-specific persistency table exists in public. The ultimate rate is anchored on the only 실손-specific figure there is — the 1-3세대 in-force block fell 3.3% in 2025 — which blends lapse, death and conversion, so the 2.0% ultimate is what is left of it once mortality and the renewal decline are taken out. The first-year rate is set against an [unverified] 장기손해보험 13회차 유지율 of about 86%, from a news summary rather than a retrieved disclosure. There is nothing to break the fall on this contract: with no surrender value there is no 보험료 자동대출납입, so a missed premium really does lapse the policy.

utilisation_table.csv

Annual claim frequencies per policy by sex and five-year age band — admissions, 급여 and 비급여 outpatient visits, and the three 3대비급여 act counts — plus the mean length of stay that the room-differential daily-average cap needs. These are frequencies of events giving rise to a paid claim, averaged over the whole population including the roughly 65% of insureds who claim nothing in a year, which is why they look low against national utilisation. The level is solved: the age-40 male row is set so that the anchor cell’s first policy year reproduces the published 4세대 2022 급여 and 비급여 loss ratios on the published 2021 new-business premium anchor. The age curves and the sex factors are [std] shapes following the NHIS coverage ratios by age band. Pregnancy and childbirth are excluded from cover, so there is deliberately no maternity bump in the female rows.

severity_table.csv

A discrete cost distribution per event for each of eight streams. A single mean is unusable on this product: the deductible is max(flat floor, percentage), which is a flat amount below a crossing point and a percentage above it, so the payment is a kinked function of cost and the shape of the cost distribution decides the claim. The 건강보험심사평가원 price survey found 도수치료 quoted anywhere between ₩5,000 and ₩600,000 across Seoul hospitals; that dispersion is the product’s structural problem and a mean would erase it.

claim_shape_table.csv

The distribution of a policy’s annual rated 비급여 claim across ten buckets, as a share and a representative amount. This is the input the experience-rating loop runs on, and it is the one table whose dispersion matters more than its mean: the 요율 상대도 band is decided by where a policy’s annual claim falls against fixed money thresholds, so the band mix is a property of the distribution and not of the average. Bucket 0 carries the 72.9% of contracts assessed with no rated claim, and the four positive groups are calibrated so that at the anchor cell’s first-year claim level the band mix reproduces the published commencement distribution 72.9 / 25.3 / 0.8 / 0.7 / 0.3 exactly — which is what makes the solved band-1 relativity come out at the specification’s 0.9575.

oop_ceiling_table.csv

The 본인부담상한제 annual ceiling by NHI-contribution decile, 2026 scale. This is the only table here that is a transcription of a published number rather than a construction, and it is the single most important interaction in the product: the NHIS refunds the excess of a member’s annual statutory co-payment over the ceiling, and the 표준약관 excludes the refundable amount from cover twice over, so the 급여 half of the claim is truncated — and truncated differently by income.

To swap in a company basis, replace the CSVs with same-schema files, or point the *_file References at different names, and clear the cache. No formula changes.

Cells Descriptions

input_dir()

The directory holding the input CSVs: the model folder’s parent.

Inputs are external files, not data stored inside the model, so the model folder is pure formulas. The path is resolved at run time from where the model was read, following annuallife.TradLife_A.

model_point_table()

The model point table, read from model_point_table.csv.

Indexed by point_id. point_id = 1 is the technical notes’ worked-example anchor cell — male, 만나이 40 at issue, all five 보장종목 held, ₩50,000,000 a year per 보장종목 and the ₩11,982 monthly office premium the 2021 launch release prints for that exact life. The others exercise the 보장종목 elections, the 보험가입금액 menu, the optional modules and the edge ages.

mort_table()

The [std] mortality construction by sex and age, from mort_table.csv.

Not a copy of 제10회 경험생명표, which 보험개발원 does not publish in full: a Makeham construction fitted to the four summary statistics 국가데이터처 does publish. Read as an annual rate of mortality at 만나이 age. On this product death is a liability release, so the usual direction of prudence is inverted — see the Space docstring.

lapse_table()

The [std] annual lapse rates by policy year, from lapse_table.csv.

Policy years beyond the last row take that row. Non-payment lapse only: the policyholder’s separate right to decline the annual renewal is a different decrement and carries its own rate, Projection.renewal_decline_rate.

utilisation_table()

The [std] annual claim frequencies by sex and age band, from utilisation_table.csv.

Indexed by sex and age_start, the lower edge of the five-year band. The columns are the frequencies of events giving rise to a paid claim, per policy per year, averaged over a population most of which claims nothing: adm_rate, visit_rate_ge, visit_rate_np, act_rate_physio, act_rate_inject and act_rate_mri, plus los_days, the mean length of stay the 상급병실료 daily-average cap is applied against.

severity_table()

The [std] discrete cost distribution per event, from severity_table.csv.

Indexed by stream and point. Eight streams: ge_in and ge_out are the 급여 본인부담금 per admission and per visit, np_in and np_room the 비급여 cost per admission excluding and comprising the 상급병실료 차액, np_out the 비급여 cost per visit, and physio, inject and mri the per-act cost of the three 3대비급여 classes. The deductible is a kinked function of cost, so the shape of each distribution, not its mean, decides the claim.

claim_shape_table()

The [std] distribution of a policy’s annual rated 비급여 claim, from claim_shape_table.csv.

Indexed by bucket. share is the proportion of contracts in the bucket and claim_amount the representative annual rated 비급여 claim of a contract in it, in KRW at the anchor cell’s first-year level; the model reads the amounts as multiples of the table’s own mean and rescales them to whatever claim level it is projecting. Bucket 0 is the no-claim mass that becomes 1단계. This is the table the 비급여 할인·할증 loop runs on.

oop_ceiling_table()

The 본인부담상한제 annual ceiling by NHI-contribution decile, from oop_ceiling_table.csv.

Indexed by decile, 1 to 10, on the 2026 scale. The NHIS refunds a member’s annual 본인일부부담금 above this amount, and the 표준약관 excludes anything so refundable from cover outright, so this is the level at which the 급여 half of the insured loss stops accruing.