Source & method

1800–1912: BLS historical compilation (Vermont family-living prices, Hoover and Rees estimates), with coverage different from modern national CPI. Whole-point historical indices have limited precision. Linked at the 1913 annual mean (29.7 → 9.9); BLS values from 1913 are preserved. Monthly coverage still starts January 1913. U.S. annual results use BLS CPI-U published annual-average indices (1913–2025), not World Bank rates. October 2025 monthly data are unavailable; the published BLS average for 2025 uses 11 months.

BLS · Census E135 (1800–1912)

Canada's annual coverage is 1913–2025. The 1913 value comes from Statistics Canada's historical annual CPI series K8, linked to the existing scale at the common year 1914. Values for 1914–2025 remain the means of 12 official monthly all-items CPI observations. Historical tables may differ in rounding and vintage. No prior-year rate is assigned to 1913. Monthly coverage begins in January 1914; incomplete 2026 is excluded from annual mode.

Statistics Canada · K8 · 1913–1975 — Adapted from Statistics Canada, Historical Statistics of Canada, K8 (1913–1975), 1983. This does not constitute an endorsement by Statistics Canada of this product.

The 1963 annual price level is available as the starting year; the first year-on-year rate is for 1964. Saudi annual rates for 1964–2006 use the World Bank/IMF series. Rates for 2007–2013 come from twelve-month means of the linked all-items CPI, checked against BIS's published annual rates; 2014–2025 use GASTAT's monthly all-items CPI means. 2013 bridges BIS and GASTAT; 2026 is not yet a full year.

Annual coverage is 1810–2025, excluding 1922, 1923 and 1948. 1810–1880 uses rounded Bundesbank purchasing-power equivalents based on Nuremberg prices; 1881–1913 covers food, 1914–1921 selected living costs, and 1924–1947 historical cost of living, not a homogeneous modern national CPI. Annual levels link at 1949. Monthly 1924-01–1944-09 uses the official five-person worker-family index; 1944-10–1948-06 is missing, without interpolation. Old monthly levels link through the published January 1957 index. July 1948 onward values and the January 1962/1991 links are preserved; annual values since 1949 are twelve-month means. Price controls, rationing and unrecorded prices limit 1938–1948 interpretation. The 1948 RM/DM 1:1 convention concerns current-payment purchasing power, not bank balances. Bundesbank uses a different geographical transition in 1994. Monthly 1920-02–1923-12 uses the BLS-published food, housing, fuel, light and clothing basket. It is approximately linked within the application at January 1924; 1924 overlap drift reaches 6.12% because baskets differ, not an official homogeneous series. M applies through 1923-11; the source uses rentenmarks from 1923-12, shown as the stable Rentenmark/Reichsmark unit RM. Denomination is applied once. From 1939-10 food quantities follow ration allocations; shortages and black-market costs are not measured. 1 fl. = 12/7 M; 1 RM = 10¹² M; 1 € = 1.95583 DM.

1800–1899: Van Riel worker-household cost-of-living estimate, linked to CBS at 1900. Dutch annual price levels for 1900–2001 use CBS's published historical all-items CPI; CBS describes this linked series as an approximate historical estimate. World Bank rates for 2002–2025 are unchanged. The CBS annual series can differ slightly from averages of its separately linked monthly bases. CBS also notes that rounding can make an annual rate calculated from its historical table differ from the officially published inflation rate. Guilder applies before 2002 and euro thereafter; the fixed €1 = ƒ2.20371 conversion is separate from inflation.

1830–1913 is a historical reconstruction. SCB calculated the 1915–1917 annual values by linear interpolation. 1830–1854: riksdaler riksgälds (rdr rg); 1855–1872: riksdaler riksmynt (rdr rm); 1873+: kr. 1 rdr rg = 1 rdr rm = 1 kr. SCB's published historical annual averages for 1830–1979 are linked to the official fixed CPI at the shared year 1980. The cost-of-living index applies through June 1954 and all-items CPI from July 1954; 1954 is a transition year. Published fixed annual averages are used for 1980–2025. Annual values are not reconstructed from rounded months.

1819–1864: Norges Bank reconstructed general price index, application-linked to SSB at 1865; not an officially published combined series. Amounts in 1819–1874 use spesidaler (spd), from 1875 kroner (kr); 1 spd = 4 kr. Both units coexisted in 1874. Norway uses SSB's published annual all-items index for 1865–2025 (table 14711). Before 1959 this is not the same coverage as today's national CPI: private-consumption price estimates, the Oslo cost-of-living index, and SSB's cost-of-living index precede it. Source indices are rounded to one decimal, making very early results approximate. Annual values are not derived from rounded monthly data.

Denmark's annual calculation uses Statistics Denmark's published PRIS8 annual-average all-items CPI for 1900–2025. Before 1914 the figures are historical price estimates. Early indices are rounded to integers, so short-interval results may be approximate. Annual values are not derived from the separate monthly series.

1800–1913: Studer–Schuppli historical reconstruction linked to FSO at 1914. The 1800–1890 Zurich basket contains 12 goods and excludes rent; coverage differs from modern national CPI. Before 1852, amounts use federal-franc-equivalent accounting units, CHF (1852), with no conversion of cantonal coins. Switzerland's annual calculation uses FSO's published annual-average linked all-items CPI for 1914–2025. The 1914–1920 values are historical estimates. Annual rates use consecutive published annual indices, not averages reconstructed from rounded monthly values.

Annual values for 1800–1947 come from the historical price reconstruction by Hubmann–Jobst–Maier (2020); 1948–2025 uses Statistik Austria LHKI 1938. The app links the series at the 1948 annual average and retains all published values from 1948. Early coverage is predominantly Vienna, with a different basket from modern national CPI. Amounts use the dated unit below. Annual selections use the year-start convention: Bankozettel in 1811, Convention florins in 1858, old schillings in 1938. Vienna paper florins are used during 1816–1819, when silver florins circulated alongside them at variable rates; the fixed 1:2.5 relation is used from 1820. The person-specific 1947 cash-redemption haircut is not a general purchasing-power conversion. Nominal unit conversions are separate from price change. 1800–1811: fl B.Z.; 1812–1819: fl W.W.; 1820–1858: fl C.M.; 1859–1899: fl ö.W.; 1900–1924: K; 1925–1938-03: S (1925); 1938-04–1945: RM; 1946–2001: ATS; 2002+: €. 5 fl B.Z. = 1 fl W.W.; 2.5 fl W.W. = 1 fl C.M. (1820); 1 fl C.M. = 1.05 fl ö.W.; 1 fl ö.W. = 2 K; 10,000 K = 1 S (1925); 1.5 S (1925) = 1 RM; 1 RM = 1 ATS; 13.7603 ATS = 1 €.

Finland's 1860–2025 annual calculations compare the two selected published points of Statistics Finland's long cost-of-living index. The official value-of-money converter excludes incomplete 1914, so it is not selectable; the published years on either side can be compared directly without inventing a 1914 value. The historical basket differs from today's CPI. Currency conversions are separate from inflation: 100 old markkaa = 1 new markka from 1963 and 5.94573 markkaa = 1 euro from 2002.

Pakistan uses published IMF/IFS annual-average CPI for 1956–1959 and preserves World Bank/IMF annual indices for 1960–2025. The common 1960 level is identical. Annual values are not derived from monthly data or fiscal-year tables.

1953–1959: IMF/IFS via UNdata, linked at 1960; earlier urban/income-group coverage is narrower. Before 1959: West African pound (W£); 1959–1972: Nigerian pound (N£). Nigeria's 1960–2025 annual results compare the published World Bank/IMF annual-average CPI index. Before 1973, Nigerian pounds convert at the official rate of 1 pound = 2 naira separately from inflation; 2026 is incomplete.

1935–1941: Manila wage-earner COLI linked at the shared 1949 observation. 1942–1945 unavailable; no annual rates are invented for the gap. 1946–1948 use Manila wage-earner cost of living; 1949–1956 use Manila CPI as historical proxies, not national CPI. The application links annual-average levels at 1949 and 1957. Geography, baskets and weights differ; overlapping inflation rates are not forced to agree. 1945 is excluded because months are missing. Monthly coverage remains at 1957; missing historical months are not derived from annual data. Philippine annual results for 1957–2025 compare PSA's published national all-items CPI annual averages (2018=100). They use tables ACP13 (1957–1993), ACP15 (1994–2017), and ACP22 (2018 onward); 2026 is not a complete year.

Malaysia annual coverage: 1960–2025, published DOSM overall annual-average CPI (2010=100). Historical geography differs from current national coverage. 1960–1966 amounts refer to Malayan dollars; from 1967 the annual reference is the new Malaysian dollar, introduced at unchanged parity in June 1967 (1:1). This purchasing-power calculation does not value retained old banknotes after their November 1967 devaluation. M$/RM is a 1:1 symbol change. Existing annual DOSM CSV values are preserved and need not equal historical monthly means. Incomplete 2026 is excluded.

1953–1959 uses the IMF/IFS historical annual index distributed by DBnomics, linked through the shared 1960 level. Existing World Bank/IMF annual levels for 1960–1975 and TPSO annual averages for 1976–2025 remain unchanged, linked at 1976 on the 2023=100 scale. The pre-1962 cost-of-living predecessor covered low-income workers and officials in Bangkok, unlike the modern national basket. App years are Gregorian; Thai Buddhist years are 543 years ahead. Incomplete 2026 is excluded.

1902–1958 uses Wu Tsong-min's linked historical CPI column B, scaled to DGBAS through 1959: rural predecessor before 1937, academic reconstruction for 1937–1957, and the author's official-linked 1958 value. This differs from modern national CPI coverage. DGBAS annual averages for 1959–2025 remain unchanged (2021=100); through 1970 coverage is Taiwan Province and before 1981 urban. Amounts use year-end currency: Taiwan yen through 1945, old Taiwan dollars in 1946–1948, new dollars from 1949. The nominal 40000:1 conversion is separate from inflation and applied once. Incomplete 2026 is excluded.

Australia annual calculations use ABS Year Book Australia 2003 Table 28.4 Sydney retail prices for 1850–1900 (regional, not a national CPI), linked at common 1901 (47); all 101 published overlap years 1901–2001 agree exactly. Then the ABS Table 3A.1 linked retail/consumer-price predecessor for 1901–1948 and the arithmetic mean of four published original All Groups Australia quarterly CPI indices from Table 17 for 1949–2025. The historical table is rescaled through calendar 1949 (128 to 2.7825). Coverage is six state capitals in 1901–1980 and eight capital cities from 1981; the earlier Sydney series has regional coverage. Incomplete 2026 is excluded. Annual amounts before 1966 are pounds (£A); 1966 onwards are AUD. The actual decimalisation date is 14 February 1966; the annual selection treats all of 1966 as AUD. £A1=A$2 is applied separately from price change. There are no separate shilling/pence inputs.

Annual calculations use published World Bank/IMF annual-average all-items CPI levels for 1992–2025. Rosstat monthly data reach back to 1991, but the comparability of a 1991 annual mean derived from month-end changes with the later annual series is unproven. Historical IMF annual-average inflation for 1992 is 1,353%, while the former derivation implied about 1,526%; close 1992 levels do not validate that bridge. Therefore 1991 is unavailable in annual calculations and remains only in the separately qualified monthly view. The 1998 ruble 1,000:1 conversion is separate from inflation. The 1960–1990 retail-price indicator is kept separate.

Annual levels for 1910–2025 are calculated by chaining Stats SA B2 official annual inflation rates for 1911–2025 to 1910=100. Levels are calculated, not published index observations; rates retain one-decimal rounding. These are annual-average changes, not December/December rates. The former rounded B1 ratio of 17.65% for 1982 is replaced by the published 14.7%. Primary urban areas through 2008, all urban areas from 2009. The 14 February 1961 conversion of one South African pound to two rand is separate from inflation. Annual 1961 uses rand; monthly February keeps pounds, March uses rand.

GUS annual-average ratios cover 1949–2025. Levels for 1949–1950 are back-calculated using 1950=107.5 and 1951=109.6; all 1951+ levels remain unchanged. The 30 October 1950 reform uses the price/wage conversion 100 old zł=3 new zł for purchasing power, not cash exchange. Annual 1950 amounts use the end-year unit. The 1995 conversion of 10,000:1 is separate. Annual data are not derived from months.

Vietnam annual coverage is 1980–2025. IMF WEO April 2025 annual-average price changes link 1980–1994 to the unchanged WB/IMF 1995 index; 1995–2025 values are preserved. Published growth rates avoid heavily rounded early index levels. This is a historical link between different publication vintages. Annual amounts use the year-end currency: 10 old dong = 1 new dong on 14 September 1985. Nominal conversion is separate from inflation. Incomplete 2026 is excluded.

Historical extension: published Buenos Aires family cost-of-living annual indices for 1913–1933 are linked at the shared 1936 annual level. Actual October 1933–December 1942 monthly observations are rescaled to the existing series using the shared 1943 annual benchmark; no monthly values are generated. This is an approximate application link across different baskets, not an official monthly splice. Some archive months are printed in whole index points. Annual values from 1934 are twelve-month means; 1933 is not calculated from just its three available months. Existing values from 1943 are retained. Argentina's annual results are arithmetic averages of twelve linked monthly indices (1943–2025). The source and geography change in 2007–2016; this is not an uninterrupted official national CPI. The July 2012 transition uses the published rounded 1.9% monthly rate. Currency changed during 1983 and 1985; annual labels use the January unit.

Istat · 1861–2025 · BIS · 1947–1954 · Istat · FOI / NIC

Annual coverage 1861–2025; monthly coverage January 1947–August 2026. 1861–1946 uses Istat historical FOI, reconstructed before 1928. Monthly 1947–1953 uses workers/employees household FOI, 1954 NIC, supplied by BIS. Annual 1947–1954 uses full monthly averages. The early monthly segment is application-linked at January 1955; later monthly values and annual rates from 1956 remain unchanged. BIS/OECD historical numerical precision differences are recorded; this combined dataset is not a single official institutional series. Before 2002 amounts are Italian lire, then euros; €1 = 1,936.27 lire is separate from inflation.

1830–1935: Maluquer de Motes / Banco de España. 1936–1939: —. 1940–1960: Banco de España / INE. Historical annual estimate; urban coverage and published rounding. Linked at common year 1961. War years are not selectable; cross-period results use an approximate historical linked level. Annual currency is the beginning-of-year unit. 1830–1864: real de vellón (Rl); 1865–1868: escudo (Esc); 1869–2001: ₧. 4 Rl = 1 ₧; 1 Esc = 2,5 ₧. Spain's 1961–2001 annual price levels are twelve-month averages of INE's national all-items CPI; 1962–2001 rates derive from those averages. The 2002 rate is INE's published 3.5% annual-average change under its new method. This official rate is rounded to one decimal; the different-base 2001 and 2002 indices are not divided directly. World Bank rates for 2003–2025 are unchanged. Pesetas apply in 1869–2001 and euros from 2002; the fixed €1 = ₧166.386 conversion is separate from inflation.

1907–1954 annual price changes use INSEE archived purchasing-power coefficients (2007 edition, euros of a common year), application-linked at 1955. 1956–1990 OECD annual averages, 1991–2001 INSEE and 2002–2025 World Bank rates are unchanged. Before 1960, nominal conversion already included in coefficients cancels in adjacent-year ratios; currency conversion is applied separately by the calculator. Early scope is Paris, then urban households through 1992. Current rounded 1901 levels and unverified original 1901–1906 pages are excluded.

1923–1937 uses Şevket Pamuk's historical Istanbul reconstruction linked at 1938 to the Treasury index published by TÜİK; 1938–1954 is scaled at the 1955 annual mean. Annual 1955–2005 levels average twelve linked months; 2006–2025 World Bank rates and existing 1982+ calculations are preserved. This application link combines changing city, urban and national coverage, not one homogeneous official series. Before 2005: 1,000,000 old TL = 1 TL; denomination conversion is separate from inflation.

Brazil: annual 1912–2025; monthly January 1939–August 2026. Rio de Janeiro cost of living (IBGE 1990, Table 5.1) supplies 1912–1938; the São Paulo IPC-FIPE predecessor supplies 1939–1947. These regional baskets are not the current national IPCA. Application links use the 1939 annual mean, January 1948 to IGPC-MTb Brasil, and December 1979 to IPCA. All existing 1948+ indices and 1995+ World Bank rates remain unchanged. Before November 1942 amounts are mil-réis (1000 réis); 1 mil-réis = 1 Cr$. Annual amounts use year-end currency. Nominal conversion is separate from inflation. First index 1912; first annual rate 1913. The 1925 publication discrepancy is unresolved: the complete IBGE 1990 edition is selected without mixing cells from different publications. 1912 → 1913. 1980–1994: IPCA; 1995–2025: World Bank. 1925: IBGE 1990 = 259; Ipeadata = 258; MTIC 1946 = 257. 1967-02 → 1967-03 (1000:1); 1970-05 → 1970-06 (1:1). 1970-11: 221.6 (IBGE), 221.5 (Ipeadata).

Published CDKO levels for 1750–1799 are from ONS Economic Trends 604, Table 1 p.43, not compounded annual rates. 1750 is the initial index year; rates start in 1751. Source rounding is retained. The 1750–1850 section uses historical consumables prices, not the modern national CPI basket. Annual coverage is 1750–2025; monthly coverage is July 1914–August 2026. Before 1950 these are historical price predecessors, not modern CPI: published ONS CDKO annual indices; ONS cost-of-living observations (BoE archive M6/M), then CDKO retail prices from June 1947. Cost of living links to CDKO in June 1947; old monthly levels link to existing CPI in January 1950. Annual historical indices are scaled at the common calendar 1950 average, not derived from monthly values. Existing 1950–2025 annual values remain twelve-month means of linked JFDN/D7BT. JFDN 1950–1987 links to D7BT at January 1988; 1950–1996 is retrospectively modelled and 1950–1987 is not a National Statistic. All existing observations from 1950 onward are unchanged. Cross-series comparisons are approximate historical purchasing power, with basket and method changes. No missing months are estimated; the 1914 annual value is the published CDKO full-year index, not a mean of the partial monthly coverage. Amounts are pounds (£).

Mexico annual coverage: 1939–2025. 1939–1969: INEGI 2009, table 17.20, Mexico City worker cost of living (23 categories; 1978=100). Application-linked at the common year 1969; not an official combined national CPI. World Bank rates for 1970–2025 remain unchanged. The 2009 value 13.9 is used for 1952, without mixing the older 13.0 vintage. Historical values have limited rounding precision. Monthly coverage still starts 1969-01; the 1957 candidate is not fully verified. Before 1993: 1 new peso = 1,000 old pesos (MXP), separately from inflation. MXP old peso (before 1993) · 1 new peso (MX$) = 1,000 MXP

Published CDKO levels for 1750–1799 are from ONS Economic Trends 604, Table 1 p.43, not compounded annual rates. 1750 is the initial index year; rates start in 1751. Source rounding is retained. The 1750–1850 section uses historical consumables prices, not the modern national CPI basket. Annual coverage is 1750–2025; monthly coverage is July 1914–August 2026. Before 1950 these are historical price predecessors, not modern CPI: published ONS CDKO annual indices; ONS cost-of-living observations (BoE archive M6/M), then CDKO retail prices from June 1947. Cost of living links to CDKO in June 1947; old monthly levels link to existing CPI in January 1950. Annual historical indices are scaled at the common calendar 1950 average, not derived from monthly values. Existing 1950–2025 annual values remain twelve-month means of linked JFDN/D7BT. JFDN 1950–1987 links to D7BT at January 1988; 1950–1996 is retrospectively modelled and 1950–1987 is not a National Statistic. All existing observations from 1950 onward are unchanged. Cross-series comparisons are approximate historical purchasing power, with basket and method changes. No missing months are estimated; the 1914 annual value is the published CDKO full-year index, not a mean of the partial monthly coverage. Amounts are pounds (£).

Annual coverage is 1810–2025, excluding 1922, 1923 and 1948. 1810–1880 uses rounded Bundesbank purchasing-power equivalents based on Nuremberg prices; 1881–1913 covers food, 1914–1921 selected living costs, and 1924–1947 historical cost of living, not a homogeneous modern national CPI. Annual levels link at 1949. Monthly 1924-01–1944-09 uses the official five-person worker-family index; 1944-10–1948-06 is missing, without interpolation. Old monthly levels link through the published January 1957 index. July 1948 onward values and the January 1962/1991 links are preserved; annual values since 1949 are twelve-month means. Price controls, rationing and unrecorded prices limit 1938–1948 interpretation. The 1948 RM/DM 1:1 convention concerns current-payment purchasing power, not bank balances. Bundesbank uses a different geographical transition in 1994. Monthly 1920-02–1923-12 uses the BLS-published food, housing, fuel, light and clothing basket. It is approximately linked within the application at January 1924; 1924 overlap drift reaches 6.12% because baskets differ, not an official homogeneous series. M applies through 1923-11; the source uses rentenmarks from 1923-12, shown as the stable Rentenmark/Reichsmark unit RM. Denomination is applied once. From 1939-10 food quantities follow ration allocations; shortages and black-market costs are not measured. 1 fl. = 12/7 M; 1 RM = 10¹² M; 1 € = 1.95583 DM.

1965-01–1969-12: IMF/IFS archive, linked using the 1970 calendar-year mean. 1957–1964 excluded because the 1961 break remains unresolved. All 1970+ values retained. ƒ (guilder) before 2002 · € from 2002 · €1 = ƒ2.20371

1931–June 1954: application-linked historical estimate from the Riksbank/Socialstyrelsen all-items consumption index. Revised December 1949 is 175.3. The 1955 link uses the published approximate 1.76 coefficient; this is not an official homogeneous CPI back-series. Riksbank / Socialstyrelsen → SCB. 1931-01–2026-08. July 1954–December 1979 uses SCB's historical all-items CPI, published as whole-number monthly indices. It is linked at January 1980 to SCB's rebased fixed CPI (2020=100); the current fixed index applies from 2026. Early monthly results are approximate because the source is rounded.

1914–1919: only 23 published dated cost-of-living observations from SSB yearbooks; missing months are not filled. The tax-inclusive 1919 a variant is checked against the 1925 yearbook. March 1920 is an application level link (288 → 3.7), not an official combined series. SSB's published monthly all-items index for March 1920–September 2026 (table 14710) is used. Pre-1959 observations come from a historical cost-of-living predecessor, not the current national CPI coverage. Very early results are approximate because source indices are published to one decimal.

The all-items series uses OECD's FRED archive for January 1967–December 1977, DST PRIS1 for January 1978–December 1979, DST PRIS12 for January 1980–November 2000, and DST PRIS01 for December 2000–September 2026. Different bases are scaled at common months in January 1978, January 1980, and December 2000. Current PRIS01 values remain unchanged; no month is inferred from annual data.

Genuine monthly observations in FSO's linked all-items CPI cover January 1921–September 2026. Annual estimates repeated in monthly slots for 1914–1920 are not treated as observations. Published monthly indices are rounded to one decimal, so very short comparisons can be approximate.

Monthly values for January 1920–December 1947 use Hubmann–Jobst–Maier (2020), linked at the common January 1948 observation to existing LHKI 1938. January 1948–September 2026 values are retained. Sparse observations from 1914–1919 are not used to invent missing months. September 2026 is provisional. Amounts use the dated unit below. Annual selections use the year-start convention: Bankozettel in 1811, Convention florins in 1858, old schillings in 1938. Vienna paper florins are used during 1816–1819, when silver florins circulated alongside them at variable rates; the fixed 1:2.5 relation is used from 1820. The person-specific 1947 cash-redemption haircut is not a general purchasing-power conversion. Nominal unit conversions are separate from price change. 1800–1811: fl B.Z.; 1812–1819: fl W.W.; 1820–1858: fl C.M.; 1859–1899: fl ö.W.; 1900–1924: K; 1925–1938-03: S (1925); 1938-04–1945: RM; 1946–2001: ATS; 2002+: €. 5 fl B.Z. = 1 fl W.W.; 2.5 fl W.W. = 1 fl C.M. (1820); 1 fl C.M. = 1.05 fl ö.W.; 1 fl ö.W. = 2 K; 10,000 K = 1 S (1925); 1.5 S (1925) = 1 RM; 1 RM = 1 ATS; 13.7603 ATS = 1 €.

Monthly coverage is January 1921–September 2026. Before August 1939, Statistics Finland 12rg total cost-of-living levels are used. Original Tilastokatsauksia tables supply 36 missing cells in 1925, 1931 and 1932 and replace 13 malformed or duplicated archive cells in November 1927 and 1928. The application links the historical block to unchanged 11xn observations at August 1939 using 100/1089.6; this is not an official combined series. Historical baskets and direct-tax coverage differ from modern CPI; annual 11xy is separately published, not derived from monthly averages. No monthly interpolation or extrapolation. Currency conversions in 1963 and 2002 remain separate from inflation.

January 1957–June 2017 uses the IMF/IFS historical consumer-price predecessor; July 2017–September 2026 uses PBS national CPI. Earlier urban coverage changed over time. The July 2017 common-month level link is calculated by the application, not an official combined series. Seven source break flags are retained and boundary ratios are checked against published changes. Existing PBS values and 2025 revisions are preserved; no missing months are invented.

1988–1994: IMF/IFS via DBnomics, application-linked to NBS using the mean of all 12 shared 1995 months. Earlier comparability breaks remain unresolved. Monthly results from January 1995 to September 2026 compare published months of NBS national all-items CPI (2024=100). The current workbook includes earlier months on the same base; older CBN vintages are not appended with an unverified factor.

For 1997–1998, the World Bank annual rates agree with CBN's older annual table, while NBS's later historical monthly data give different year averages. The 1997 monthly values differ between the two official publications; the series remain separate.

Philippine monthly results for January 1957–September 2026 compare two published PSA national all-items CPI months (2018=100). They use official tables ACP13, ACP15, and ACP22; missing months are not estimated.

Malaysia monthly coverage: 1968-01–2026-09. 1968–1979 Peninsular Malaysia CPI (national sources/BIS), then existing DOSM CPI from 1980. The application links levels at the common month January 1980; this is not an official combined series. All 36 overlap months in 1980–1982 were checked. Modern DOSM values are preserved; no months are interpolated. December 1992 M$/RM symbol change is 1:1. Sources: National sources; BIS long consumer prices database, www.bis.org/statistics/cp.htm.

Thailand's monthly calculations use TPSO's published, unadjusted national all-items CPI (January 1976–September 2026, 2023=100). TPSO supplies the entire historical series on that base; no extra rebasing or interpolated months are used.

Taiwan monthly calculations use the DGBAS Table 1-8 published, unadjusted all-items CPI index (January 1959–September 2026, 2021=100). Hidden workbook rows for 1959–1961 are included. Through 1970, observations cover Taiwan Province; before 1981, the linked predecessor is an urban series. No missing months are estimated; recent three months may be revised.

Australia quarterly calculations use RBA G1 original GCPIAG historical retail-price observations for June quarter 1922–June quarter 1948, then the published ABS Table 17 original All Groups CPI (A2325846C), September quarter 1948–September quarter 2026, without extra rebasing or interpolation. All 313 common RBA/ABS quarters agree exactly, with no additional splice factor. Source: Reserve Bank of Australia 2026 and ABS, under RBA Financial Data terms section 5 and ABS CC BY 4.0. Periods are January–March, April–June, July–September and October–December, not individual months. Coverage is six state capitals before 1981, then eight capital cities. From December quarter 2025 ABS averages the complete monthly CPI and splices it to the historical quarterly series at September quarter 2025; the application retains the published quarterly levels. Both endpoints are quarters. Decimalisation occurred on 14 February 1966. The quarter-ending currency applies: £A through December quarter 1965, A$ from March quarter 1966; £A1=A$2 is applied separately from inflation. The March quarter 1966 straddles the reform and does not represent an exact daily currency date. Source: Australian Bureau of Statistics, CC BY 4.0; no ABS endorsement.

Australia monthly calculations link two official ABS products in the application: the Monthly CPI Indicator for September 2017–March 2024 and the complete Monthly CPI for April 2024–September 2026. Earlier indicator levels are multiplied by 96.4/123.7 at the common April 2024 observation; published complete Monthly CPI levels are unchanged. This is not a single ABS-published combined series. Aligning levels does not remove differences in collection methods. Both sources are unadjusted All Groups indices representing the weighted average of eight capital cities. No quarterly interpolation or missing-month estimates are used. The 2025 annual index averages the historical quarterly series and need not equal the twelve-month average of the new monthly series.

Monthly coverage is January 1981–September 2026. Twelve 1981 levels are calculated from GUS 1982/1981 same-month ratios and anchored at December 1981 using January 1982 MoM. These are approximate calculated levels. Rounding gives a 1982 annual-average comparison difference of about −0.07815 percentage point; annual data are not forced to match. All 1982+ points are preserved; published linked indices are used from 1999. No missing months are filled. The 1995 currency conversion is separate from inflation.

Vietnam's monthly calculations chain NSO E 11.01 month-on-month changes from January 1995 to December 2025 and NSO releases for January–September 2026. The 1995–1997 figures are the retail-price predecessor of the official CPI first published in 1998. The transition uses NSO's published January 1998 monthly change. Long comparisons crossing this methodological change are historical approximations. Scaling the 2020 annual mean to 100 does not change ratios. Rounded monthly changes can differ slightly from the separate annual CPI series. Missing months are not estimated.

January 1991–September 2026 monthly results chain Rosstat's national, unadjusted month-end changes. The 1991 observations predate the current CPI methodology, so long comparisons touching 1991 are indicative historical measures. No pre-1991 or unpublished months are invented. Amounts from January 1998 are in new rubles; the 1,000:1 denomination change is separate from inflation.

BIS/Stats SA monthly history for January 1922–December 2008 is application-linked at January 2009 to the existing Stats SA all-urban series. January 2009–September 2026 values are unchanged. The older one-decimal 1980–2008 B1 levels are replaced by the more detailed historical chain; vintage and scope differences are quantified in the report. This is not a single official combined series. No missing months are invented. Currency conversion is separate; historical urban coverage is narrower than the whole modern population.

BIS WS_LONG_CPI · Stats SA ↗

The World Bank/IMF annual rates for 2019 and 2020 differ from rates calculated from calendar-year averages of the PBS national monthly index (by about 1.18 and 0.27 percentage points). PBS introduced a new base and national coverage in 2019; this alone does not establish the full cause of the gap. The published series remain separate and are not forced to match.

Historical extension: published Buenos Aires family cost-of-living annual indices for 1913–1933 are linked at the shared 1936 annual level. Actual October 1933–December 1942 monthly observations are rescaled to the existing series using the shared 1943 annual benchmark; no monthly values are generated. This is an approximate application link across different baskets, not an official monthly splice. Some archive months are printed in whole index points. Annual values from 1934 are twelve-month means; 1933 is not calculated from just its three available months. Existing values from 1943 are retained. The chain uses INDEC Greater Buenos Aires CPI through 2006, San Luis through June 2012, Buenos Aires City through March 2016, INDEC Greater Buenos Aires through November 2016, and national INDEC CPI thereafter. The July 2012 link uses the published rounded 1.9% monthly increase and is approximate. Geographic and methodological changes make this a comparison estimate, not one continuous official national CPI.

Official redenominations in 1970, 1983, 1985 and 1992 are applied across dates. The June 1985 reform occurred mid-month; June is labeled in peso argentino and July in austral.

Annual coverage 1861–2025; monthly coverage January 1947–August 2026. 1861–1946 uses Istat historical FOI, reconstructed before 1928. Monthly 1947–1953 uses workers/employees household FOI, 1954 NIC, supplied by BIS. Annual 1947–1954 uses full monthly averages. The early monthly segment is application-linked at January 1955; later monthly values and annual rates from 1956 remain unchanged. BIS/OECD historical numerical precision differences are recorded; this combined dataset is not a single official institutional series. Before 2002 amounts are Italian lire, then euros; €1 = 1,936.27 lire is separate from inflation.

March 1954–1960: BIS historical series scaled to INE using the 1961 common-year mean. Early coverage differs from modern national CPI. The 1939–1954 archival link is incomplete. Spain's official national all-items CPI uses INE Base 1992 through 2001 and INE Base 2025 from 2002. Because sale prices entered the CPI in 2002, these two indices must not be divided directly. Cross-boundary calculations follow INE's three-stage method using its published monthly 2002 year-on-year rates. Those rates are rounded to one decimal, so results are approximate. Post-2002 indices are unchanged; peseta/euro conversion is applied separately.

January 1951–December 1954 BIS/INSEE historical retail-price predecessor is application-linked at January 1955 to the existing OECD chain. January 1955–December 1995 OECD/FRED and January 1996 onward INSEE values are unchanged. BIS/OECD overlap in 1955–1959 differs by at most 0.5732%; these are different historical published chains. No 1949/1950 months are inferred. Paris coverage through 1962, urban households through 1992, all metropolitan households in 1993–1998 and all France from 1999. Currency conversion is separate.

Annual 1907–1962: Park and Kim's historical Seoul CPI research published by the Bank of Korea, including early scholarly estimates. World Bank/IMF rates for 1963–2025 are unchanged. Monthly January 1960–December 1964 OECD/FRED Seoul history is linked to KOSIS by the 12 common months' mean in 1965; later KOSIS values are unchanged. This application-linked series has changing geographic coverage, not national CPI since 1907. Monthly 1955–1959 cells were not verified and are excluded. 1953:100 old won=1 hwan;1962:10 hwan=1 won. Annual reform years use the year-end unit; monthly June 1962 uses hwan and July won. War and price-control periods have historical measurement limitations.

BIS · Japan · 1946–1969 · e-Stat · 1948–1959

Japan annual coverage 1947–2025: published annual rates excluding imputed rent for 1948–1959; existing World Bank rates from 1960. Monthly coverage August 1946–August 2026: BIS CPI excluding imputed rent through 1969, existing all-items CPI from 1970. BIS is scaled at January 1970; all later 680 values are preserved. Actual paid rents remain included. Early coverage is urban; rural coverage expands in 1962/1965. The incomplete five-month 1946 is excluded from annual comparisons. Annual/monthly publication precision and vintages differ. No new yen conversion applies.

China annual coverage is 1950–2025. The application links historical urban workers’ cost-of-living/urban CPI for 1950–1984, NBS national CPI for 1985–1986, and the existing World Bank annual rates from 1987. The historical urban proxy does not have the same coverage as modern national CPI. Published annual rates are chained without intermediate rounding. 1950 is the initial index year; changes start in 1951. Annual amounts use year-end currency: 10,000 old yuan = 1 new yuan, with the legal reform on 1 March 1955; annual 1955 uses new yuan. Currency conversion is separate from price change. Monthly coverage is January 1986–August 2026. Published IMF/IFS monthly index levels for 1986–1992 come from the DBnomics archive and are aligned using the average of the 12 overlapping months in 1993. All values from 1993 remain unchanged: OECD in 1993–2010 and NBS from 2011. No months are inferred from annual data or reconstructed from year-on-year rates. City archives from 1926 were found but are not activated because the May/June 1949 price linkage remains unverified.

Saudi monthly all-items values for February 1980–December 1989 come from archived IMF/IFS data and are linked to BIS using the shared months of 1990. BIS data for 1990–2012 are linked to GASTAT using 2013; existing values from 1990 onward are unchanged. IMF/IFS flags January 1981 as a series break, and its archive has no 1979 monthly observations. Earlier city coverage was narrower than today's national index.

Annual values for 1950–1956 use BPS's Jakarta family-group I total cost-of-living index, linked to IMF at 1957. The source footnote on historical totals specifies the average of January, September and December; no twelve-month averages are reconstructed. Annual: 1957–1965 IMF/IFS annual-average price indices; existing World Bank rates for 1966–2025 are retained. Annual amounts before 1965 use old rupiah; from 1965 they use year-end currency. 1,000 old Rp = 1 Rp. Monthly: 1966–1967 Jakarta cost-of-living indices (CBS/BPS, Federal Reserve IFDP 3 Table 3) are linked at January 1968. Early coverage is not nationwide. Historical vintages differ; every existing month from 1968 is retained. The historical annual series and monthly Jakarta series differ; monthly averages are not forced to match annual rates. Indonesia's monthly series uses the OECD all-items index through March 1979 and BPS headline inflation thereafter, linked at April 1979. Seven monthly rates in 1996–2006 were corrected against an archived BPS table; conflicting archive values were not applied. The old and new 2008 index bases are linked through their December 2007 reference values; the May–June bridge therefore differs from BPS's published monthly rate for the new basket. Earlier city coverage was narrower. Annual mode uses separate World Bank/IMF data.

1954-04–1968-07: Labour Bureau A15 General CPI-IW (1949=100) is linked to the 1960 base using the published factor 1.2154. Industrial-worker coverage differs from later Combined CPI. Annual 1955–1959 uses Economic Survey calendar-year All Items means; existing World Bank/IMF rates from 1960, including the 1960 bridge, are retained. This splice is an application historical estimate. India's August 1968–December 1999 CPI-IW months come from the Labour Bureau table; January 2000–December 2010 uses the OECD archive republished by FRED to construct the historical estimate. Old bases are linked with the published 4.93 and 4.63 factors and anchored to the January 2011 MoSPI rural-and-urban combined CPI level. These are not official combined CPI observations; values from 2011 onward are unchanged. Annual mode uses separate data.

Brazil: annual 1912–2025; monthly January 1939–August 2026. Rio de Janeiro cost of living (IBGE 1990, Table 5.1) supplies 1912–1938; the São Paulo IPC-FIPE predecessor supplies 1939–1947. These regional baskets are not the current national IPCA. Application links use the 1939 annual mean, January 1948 to IGPC-MTb Brasil, and December 1979 to IPCA. All existing 1948+ indices and 1995+ World Bank rates remain unchanged. Before November 1942 amounts are mil-réis (1000 réis); 1 mil-réis = 1 Cr$. Annual amounts use year-end currency. Nominal conversion is separate from inflation. First index 1912; first annual rate 1913. The 1925 publication discrepancy is unresolved: the complete IBGE 1990 edition is selected without mixing cells from different publications. 1912 → 1913. 1980–1994: IPCA; 1995–2025: World Bank. 1925: IBGE 1990 = 259; Ipeadata = 258; MTIC 1946 = 257. 1967-02 → 1967-03 (1000:1); 1970-05 → 1970-06 (1:1). 1970-11: 221.6 (IBGE), 221.5 (Ipeadata).

January 1955–December 1981 uses the IMF IFS archive mirrored by DBnomics (August 2025 vintage), scaled at January 1982. All existing 1982+ TÜİK/CBRT values and the 1988/1994/2003 and 2004/2005 links are preserved. Historical coverage varies; annual data are not interpolated into months. Before 2005: 1,000,000 old TL = 1 TL, separate from inflation.

Mexico's 1969–1992 monthly indices are chained backward from the January 1993 INEGI index using Banco de México's official monthly rates. Published rates have limited precision; INEGI values from 1993 onward remain unchanged. Before 1993, MXP means old pesos: 1 new peso (MX$) = 1,000 MXP. Total change shows price-level change only.