import json
import ssl
import urllib.request
from pathlib import Path

import numpy as np
import pandas as pd
import matplotlib as mpl
import matplotlib.pyplot as plt

pd.set_option("display.width", 200)
pd.set_option("display.max_rows", 60)

HERE    = Path.cwd()
CACHE   = HERE / "data"
OUT_JS  = HERE.parent / "assets" / "data.js"
YEAR    = "2026"          # the vintage the essay is written against

# --- the site's palette, so the notebook and the published charts agree -------
PAPER  = "#FFFBF0"
SAND   = "#C0AE86"   # every other country: context, not an identity
MARK   = "#1E4A85"
BRICK  = "#BE3C2C"   # Israel, the one highlighted series
AMBER  = "#C0810C"
MARI   = "#F2C230"
INK    = "#12305C"
MUTED  = "#5C6779"

mpl.rcParams.update({
    "figure.dpi": 130, "savefig.dpi": 130,
    "font.family": "sans-serif", "font.style": "normal",
    # Not "Roboto" on purpose -- see analysis/README.md. An italic-only local
    # install makes matplotlib render every figure in italics.
    "font.sans-serif": ["Helvetica Neue", "Helvetica", "Arial", "DejaVu Sans"],
    "font.size": 9,
    "axes.edgecolor": "#E5DAC0", "axes.labelcolor": MUTED, "axes.titlesize": 11,
    "axes.titleweight": "bold", "axes.titlecolor": INK, "axes.titlelocation": "left",
    "axes.spines.top": False, "axes.spines.right": False,
    "xtick.color": MUTED, "ytick.color": MUTED,
    "xtick.labelcolor": MUTED, "ytick.labelcolor": MUTED, "text.color": INK,
    "figure.facecolor": PAPER, "axes.facecolor": PAPER, "savefig.facecolor": PAPER,
    "legend.frameon": False, "legend.fontsize": 8.5,
})
def fetch(url, timeout=90):
    # python.org macOS builds often ship without a usable CA bundle, which
    # turns every https call into CERTIFICATE_VERIFY_FAILED. Use certifi's
    # bundle when it is installed and fall back to the system default.
    try:
        import certifi
        ctx = ssl.create_default_context(cafile=certifi.where())
    except ImportError:
        ctx = None
    with urllib.request.urlopen(url, timeout=timeout, context=ctx) as r:
        return r.read()


print("cache:", CACHE, "|", "exists" if CACHE.is_dir() else "will be created")

cache: /Users/davidzeff/Desktop/blog/analysis/data | exists

OECD38 = {
    "AUS": "Australia",   "AUT": "Austria",    "BEL": "Belgium",       "CAN": "Canada",
    "CHL": "Chile",       "COL": "Colombia",   "CRI": "Costa Rica",    "CZE": "Czechia",
    "DNK": "Denmark",     "EST": "Estonia",    "FIN": "Finland",       "FRA": "France",
    "DEU": "Germany",     "GRC": "Greece",     "HUN": "Hungary",       "ISL": "Iceland",
    "IRL": "Ireland",     "ISR": "Israel",     "ITA": "Italy",         "JPN": "Japan",
    "KOR": "Korea",       "LVA": "Latvia",     "LTU": "Lithuania",     "LUX": "Luxembourg",
    "MEX": "Mexico",      "NLD": "Netherlands","NZL": "New Zealand",   "NOR": "Norway",
    "POL": "Poland",      "PRT": "Portugal",   "SVK": "Slovakia",      "SVN": "Slovenia",
    "ESP": "Spain",       "SWE": "Sweden",     "CHE": "Switzerland",   "TUR": "Türkiye",
    "GBR": "United Kingdom", "USA": "United States",
}
assert len(OECD38) == 38, len(OECD38)
print(f"{len(OECD38)} member countries")

38 member countries

API = "https://www.imf.org/external/datamapper/api/v1/{}"

def imf(indicator):
    # -> {iso3: {year: value}}, cached on disk so the notebook is reproducible
    CACHE.mkdir(exist_ok=True)
    path = CACHE / f"imf-weo-{indicator}.json"
    if not path.exists():
        print(f"downloading {indicator} …")
        path.write_bytes(fetch(API.format(indicator)))
    else:
        print(f"{indicator}: cached")
    return json.loads(path.read_text())["values"][indicator]


nominal_all = imf("NGDPDPC")   # GDP per capita, current US$ (market exchange rate)
ppp_all     = imf("PPPPC")     # GDP per capita, current international $ (PPP)
gdp_all     = imf("NGDPD")     # GDP, current US$ billions

missing = [c for c in OECD38 if c not in nominal_all or c not in ppp_all]
assert not missing, f"IMF data missing for {missing}"
no_year = [c for c in OECD38 if YEAR not in nominal_all[c] or YEAR not in ppp_all[c]]
assert not no_year, f"no {YEAR} projection for {no_year}"
print(f"\nall 38 members present, all have a {YEAR} projection")

NGDPDPC: cached
PPPPC: cached
NGDPD: cached

all 38 members present, all have a 2026 projection

# One tidy frame, every member, every year the IMF publishes.
records = []
for iso, name in OECD38.items():
    years = set(nominal_all[iso]) & set(ppp_all[iso])
    for y in sorted(years):
        records.append({"iso3": iso, "country": name, "year": int(y),
                        "gdp_pc_usd": nominal_all[iso][y], "gdp_pc_ppp": ppp_all[iso][y]})

weo = pd.DataFrame(records)
weo["price_level"] = weo["gdp_pc_usd"] / weo["gdp_pc_ppp"] * 100
weo.to_csv(CACHE / "imf-weo-oecd.csv", index=False)
print(f"{len(weo):,} country-years, {weo.year.min()}–{weo.year.max()}")
weo.head()

1,896 country-years, 1980–2031

year = weo[weo.year == int(YEAR)].set_index("country")
tbl = pd.DataFrame({
    "nominal $": year["gdp_pc_usd"].round(0).astype(int),
    "PPP $":     year["gdp_pc_ppp"].round(0).astype(int),
    "price level": year["price_level"].round(1),
})
tbl["rank nominal"] = tbl["nominal $"].rank(ascending=False).astype(int)
tbl["rank PPP"]     = tbl["PPP $"].rank(ascending=False).astype(int)
tbl["rank change"]  = tbl["rank nominal"] - tbl["rank PPP"]     # negative = falls
tbl = tbl.sort_values("nominal $", ascending=False)

assert tbl.loc["United States", "price level"] == 100.0, "US must be 100 by construction"
display(tbl)

pl   = tbl["price level"]
isr  = tbl.loc["Israel"]
avg  = round(pl.mean(), 1)

checks = {
    "Israel nominal GDP per person = $69,804":   (int(isr["nominal $"]), 69804),
    "Israel PPP GDP per person = $59,095":       (int(isr["PPP $"]), 59095),
    "Israel ranks 11th by nominal":              (int(isr["rank nominal"]), 11),
    "Israel ranks 26th by PPP":                  (int(isr["rank PPP"]), 26),
    "…a fall of 15 places":                      (int(isr["rank nominal"] - isr["rank PPP"]), -15),
    "Israel price level = 118.1":                (round(float(isr["price level"]), 1), 118.1),
    "Israel is 3rd most expensive":              (int(pl.rank(ascending=False)["Israel"]), 3),
    "OECD simple average price level = 77.7":    (float(avg), 77.7),
    "Israel is +52% above that average":         (round((isr["price level"] / avg - 1) * 100), 52),
    "Israel GDP ≈ $720bn":                       (round(gdp_all["ISR"][YEAR]), 720),
    # rank change is nominal minus PPP, so a fall is negative and the
    # biggest non-Israel fall is the minimum.
    "next biggest fall is New Zealand's 8":      (int(-tbl["rank change"].drop("Israel").min()), 8),
}
width = max(len(k) for k in checks)
ok = True
for label, (got, want) in checks.items():
    good = got == want
    ok &= good
    print(f"{'PASS' if good else 'FAIL'}  {label:<{width}}  computed {got!r:>10}  essay says {want!r}")
assert ok, "the essay and the source no longer agree"
print("\nEvery published figure reproduces from the IMF series.")

PASS  Israel nominal GDP per person = $69,804  computed      69804  essay says 69804
PASS  Israel PPP GDP per person = $59,095      computed      59095  essay says 59095
PASS  Israel ranks 11th by nominal             computed         11  essay says 11
PASS  Israel ranks 26th by PPP                 computed         26  essay says 26
PASS  …a fall of 15 places                     computed        -15  essay says -15
PASS  Israel price level = 118.1               computed      118.1  essay says 118.1
PASS  Israel is 3rd most expensive             computed          3  essay says 3
PASS  OECD simple average price level = 77.7   computed       77.7  essay says 77.7
PASS  Israel is +52% above that average        computed         52  essay says 52
PASS  Israel GDP ≈ $720bn                      computed        720  essay says 720
PASS  next biggest fall is New Zealand's 8     computed          8  essay says 8

Every published figure reproduces from the IMF series.

moves = tbl[["rank nominal", "rank PPP", "rank change", "price level"]] \
            .sort_values("rank change")
print("Biggest falls (expensive relative to output):")
display(moves.head(6))
print("Biggest climbs (cheap relative to output):")
display(moves.tail(6).iloc[::-1])

second = moves["rank change"].drop("Israel").min()
print(f"\nIsrael falls {-int(moves.loc['Israel','rank change'])} places; "
      f"next biggest fall is {-int(second)}. "
      f"Ratio {moves.loc['Israel','rank change'] / second:.2f}×.")

Biggest falls (expensive relative to output):

Biggest climbs (cheap relative to output):

Israel falls 15 places; next biggest fall is 8. Ratio 1.88×.

order = pl.sort_values(ascending=False)
fig, ax = plt.subplots(figsize=(7.6, 8.6))
colors = [BRICK if n == "Israel" else SAND for n in order.index]
ax.barh(range(len(order)), order.values, color=colors, height=.72)
ax.set_yticks(range(len(order)), order.index)
ax.tick_params(axis="y", length=0)
ax.invert_yaxis()
for i, (n, v) in enumerate(order.items()):
    ax.text(v + 1.4, i, f"{v:.1f}", va="center", fontsize=8,
            color=INK if n == "Israel" else MUTED,
            fontweight="bold" if n == "Israel" else "normal")
for lbl in ax.get_yticklabels():
    if lbl.get_text() == "Israel":
        lbl.set_color(BRICK); lbl.set_fontweight("bold")

ax.axvline(100, color=MUTED, lw=1, ls="-", zorder=0)
ax.axvline(avg, color=AMBER, lw=1.2, ls="--", zorder=0)
ax.text(100, -1.4, "US = 100", ha="center", fontsize=8, color=MUTED)
ax.text(avg, -1.4, f"OECD avg {avg}", ha="center", fontsize=8, color=AMBER)
ax.set_xlim(0, 145)
ax.set_xlabel("comparative price level, United States = 100")
ax.grid(axis="x", color="#EFE5CD", lw=.8)
ax.set_axisbelow(True)
ax.set_title("Third most expensive country in the OECD", pad=22)
fig.text(.02, .04, f"IMF World Economic Outlook, {YEAR} projections · "
                   "price level = nominal ÷ PPP GDP per person", fontsize=7.5, color=MUTED)
plt.show()

def ols(frame, log_x):
    x = np.log(frame["gdp_pc_ppp"].values) if log_x else frame["gdp_pc_ppp"].values
    y = frame["price_level"].values
    b = ((x - x.mean()) * (y - y.mean())).sum() / ((x - x.mean()) ** 2).sum()
    a = y.mean() - b * x.mean()
    pred = a + b * x
    resid = y - pred
    r2 = 1 - (resid ** 2).sum() / ((y - y.mean()) ** 2).sum()
    return a, b, pred, resid, r2


def spec(yr, drop_outliers, log_x):
    f = weo[weo.year == yr].set_index("country")
    if drop_outliers:
        f = f.drop(["Luxembourg", "Ireland"])
    a, b, pred, resid, r2 = ols(f, log_x)
    f = f.assign(predicted=pred, residual=resid, gap_pct=resid / pred * 100)
    ranked = f["gap_pct"].rank(ascending=False).astype(int)
    return f, {
        "year": yr,
        "x": "log" if log_x else "linear",
        "sample": "36 (drop LUX, IRL)" if drop_outliers else "all 38",
        "R²": round(r2, 3),
        "Israel actual": round(f.loc["Israel", "price_level"], 1),
        "Israel predicted": round(f.loc["Israel", "predicted"], 1),
        "Israel gap %": round(f.loc["Israel", "gap_pct"]),
        "Israel rank": int(ranked["Israel"]),
        "runner-up": f["gap_pct"].drop("Israel").idxmax(),
    }


rows, frames = [], {}
for yr in (2024, int(YEAR)):
    for log_x in (False, True):
        for drop in (True, False):
            f, row = spec(yr, drop, log_x)
            frames[(yr, log_x, drop)] = f
            rows.append(row)

robust = pd.DataFrame(rows)
display(robust)
print(f"Israel is the largest positive residual in "
      f"{(robust['Israel rank'] == 1).sum()} of {len(robust)} specifications.")

Israel is the largest positive residual in 8 of 8 specifications.

f = frames[(int(YEAR), False, True)]           # the essay's own specification
a, b, pred, resid, r2 = ols(f, log_x=False)

fig, ax = plt.subplots(figsize=(8, 5))
x, y = f["gdp_pc_ppp"], f["price_level"]
grid = np.linspace(x.min() * .95, x.max() * 1.03, 50)
ax.plot(grid, a + b * grid, color=MUTED, lw=1.4, ls="--", zorder=2,
        label=f"OLS fit (R² = {r2:.2f})")

others = f.drop("Israel")
ax.scatter(others["gdp_pc_ppp"], others["price_level"], s=34, color=SAND, zorder=3,
           edgecolor=PAPER, linewidth=1.2, label="OECD member")
ix, iy = f.loc["Israel", "gdp_pc_ppp"], f.loc["Israel", "price_level"]
ax.plot([ix, ix], [a + b * ix, iy], color=BRICK, lw=1.2, ls=":", zorder=4)
ax.scatter([ix], [iy], s=110, color=BRICK, zorder=5, edgecolor=PAPER, linewidth=1.5,
           label="Israel")
ax.annotate(f"ISRAEL\n{f.loc['Israel','gap_pct']:+.0f}% above the line",
            (ix, iy), textcoords="offset points", xytext=(12, 4),
            fontsize=9, fontweight="bold", color=BRICK, linespacing=1.4)

for n in ("Switzerland", "Iceland", "Norway", "United States", "Korea", "Türkiye"):
    ax.annotate(n, (f.loc[n, "gdp_pc_ppp"], f.loc[n, "price_level"]),
                textcoords="offset points", xytext=(7, -3), fontsize=7.5, color=MUTED)

ax.xaxis.set_major_formatter(lambda v, _: f"${v/1000:,.0f}k")
ax.set_xlabel("output per person, PPP international $")
ax.set_ylabel("comparative price level (US = 100)")
ax.grid(color="#EFE5CD", lw=.8)
ax.set_axisbelow(True)
ax.legend(loc="upper left")
ax.set_title("Israel sits furthest above the line of any OECD country", pad=10)
fig.text(.02, -.02, f"IMF WEO {YEAR} projections · Luxembourg and Ireland excluded: "
                    "their GDP is inflated by corporate accounting", fontsize=7.5, color=MUTED)
plt.show()

print(f"Israel: predicted {f.loc['Israel','predicted']:.0f}, "
      f"actual {f.loc['Israel','price_level']:.1f}, "
      f"gap {f.loc['Israel','gap_pct']:+.0f}%")
print("\nfurthest above the line:")
display(f.nlargest(5, "gap_pct")[["price_level", "predicted", "gap_pct"]].round(1))

Israel: predicted 73, actual 118.1, gap +62%

furthest above the line:

isr_ts = weo[(weo.iso3 == "ISR") & (weo.year >= 1995)].set_index("year")
actual = isr_ts[isr_ts.index <= 2024]
proj   = isr_ts[isr_ts.index >= 2024]

fig, ax = plt.subplots(figsize=(8.4, 4.2))
ax.axvspan(2024.5, isr_ts.index.max(), color=MARI, alpha=.16, zorder=0)
ax.plot(actual.index, actual["price_level"], color=MARK, lw=2.2, zorder=3, label="outturn")
ax.plot(proj.index, proj["price_level"], color=MARK, lw=2.2, ls=":", zorder=3,
        label="IMF projection")
ax.axhline(100, color=MUTED, lw=1, zorder=1)
ax.text(1996, 101.5, "United States = 100", fontsize=7.5, color=MUTED)
ax.text(isr_ts.index.max() - .2, ax.get_ylim()[0] + 3, "projected", ha="right",
        fontsize=8, color=AMBER, fontweight="bold")

# 2024 is labelled below and left so it does not collide with the climb.
for yr, dx, dy, ha in ((2024, -8, -26, "right"), (int(YEAR), 6, 10, "left")):
    v = isr_ts.loc[yr, "price_level"]
    ax.scatter([yr], [v], s=52, color=BRICK, zorder=4, edgecolor=PAPER, linewidth=1.4)
    ax.annotate(f"{yr}\n{v:.1f}", (yr, v), textcoords="offset points", xytext=(dx, dy),
                ha=ha, fontsize=8.5, fontweight="bold", color=BRICK, linespacing=1.3)

ax.set_ylabel("Israel's comparative price level")
ax.set_xlabel("year")
ax.grid(color="#EFE5CD", lw=.8)
ax.set_axisbelow(True)
ax.legend(loc="upper left")
ax.set_title("The 118 is a projection, and a record", pad=10)
plt.show()

hist = isr_ts.loc[1995:2024, "price_level"]
print(f"1995–2024 outturn: mean {hist.mean():.1f}, min {hist.min():.1f} "
      f"({hist.idxmin()}), max {hist.max():.1f} ({hist.idxmax()})")
print(f"2024 outturn        : {isr_ts.loc[2024,'price_level']:.1f}")
print(f"{YEAR} projection      : {isr_ts.loc[int(YEAR),'price_level']:.1f}")
print(f"\nnominal GDP per person {isr_ts.loc[2024,'gdp_pc_usd']:,.0f} -> "
      f"{isr_ts.loc[int(YEAR),'gdp_pc_usd']:,.0f}  "
      f"({isr_ts.loc[int(YEAR),'gdp_pc_usd']/isr_ts.loc[2024,'gdp_pc_usd']-1:+.1%})")
print(f"PPP GDP per person     {isr_ts.loc[2024,'gdp_pc_ppp']:,.0f} -> "
      f"{isr_ts.loc[int(YEAR),'gdp_pc_ppp']:,.0f}  "
      f"({isr_ts.loc[int(YEAR),'gdp_pc_ppp']/isr_ts.loc[2024,'gdp_pc_ppp']-1:+.1%})")

1995–2024 outturn: mean 99.4, min 84.0 (2006), max 113.0 (2021)
2024 outturn        : 100.3
2026 projection      : 118.1

nominal GDP per person 54,294 -> 69,804  (+28.6%)
PPP GDP per person     54,126 -> 59,095  (+9.2%)

y24 = weo[weo.year == 2024].set_index("country")
print(f"2024 outturn — OECD simple average price level: {y24['price_level'].mean():.1f}")
print(f"2024 outturn — Israel {y24.loc['Israel','price_level']:.1f}, "
      f"rank {int(y24['price_level'].rank(ascending=False)['Israel'])} of 38, "
      f"{y24.loc['Israel','price_level']/y24['price_level'].mean()-1:+.0%} above the average")
display(y24["price_level"].round(1).nlargest(6))

2024 outturn — OECD simple average price level: 71.8
2024 outturn — Israel 100.3, rank 3 of 38, +40% above the average

country
Iceland          110.0
Switzerland      109.7
Israel           100.3
United States    100.0
Australia         94.3
Luxembourg        92.5
Name: price_level, dtype: float64

WB = ("https://api.worldbank.org/v2/country/{iso}/indicator/{ind}"
      "?format=json&per_page=2000&date=2015:2024")

def worldbank(indicator):
    CACHE.mkdir(exist_ok=True)
    path = CACHE / f"worldbank-{indicator}.json"
    if not path.exists():
        print(f"downloading {indicator} …")
        path.write_bytes(fetch(WB.format(iso=";".join(OECD38), ind=indicator)))
    else:
        print(f"{indicator}: cached")
    payload = json.loads(path.read_text())
    return pd.DataFrame([
        {"iso3": r["countryiso3code"], "year": int(r["date"]), indicator: r["value"]}
        for r in payload[1] if r["value"] is not None
    ])


ppp_factor = worldbank("PA.NUS.PPP")      # PPP conversion factor, GDP (LCU per int'l $)
fx_rate    = worldbank("PA.NUS.FCRF")     # official exchange rate (LCU per US$)

wb = ppp_factor.merge(fx_rate, on=["iso3", "year"])
wb["price_level_wb"] = wb["PA.NUS.PPP"] / wb["PA.NUS.FCRF"] * 100
wb = wb.sort_values("year").groupby("iso3").tail(1)          # latest year per country
wb["country"] = wb["iso3"].map(OECD38)
wb.to_csv(CACHE / "worldbank-price-level-oecd.csv", index=False)

wb_pl = wb.set_index("country")["price_level_wb"]
print(f"\nWorld Bank, latest available year ({wb.year.mode()[0]}), {len(wb)} members")
print(f"  Israel: {wb_pl['Israel']:.1f}   rank {int(wb_pl.rank(ascending=False)['Israel'])} of {len(wb_pl)}")
print(f"  OECD simple average: {wb_pl.mean():.1f}")
print(f"  Israel vs average:  {wb_pl['Israel']/wb_pl.mean()-1:+.1%}")

PA.NUS.PPP: cached
PA.NUS.FCRF: cached

World Bank, latest available year (2024), 38 members
  Israel: 94.7   rank 4 of 38
  OECD simple average: 70.2
  Israel vs average:  +34.9%

compare = pd.DataFrame({
    f"IMF {YEAR} (projection)": pl,
    "IMF 2024 (outturn)":       y24["price_level"].round(1),
    "World Bank / ICP 2024":    wb_pl.round(1),
}).sort_values(f"IMF {YEAR} (projection)", ascending=False)

fig, ax = plt.subplots(figsize=(8, 8.4))
yy = np.arange(len(compare))
ax.hlines(yy, compare["World Bank / ICP 2024"], compare[f"IMF {YEAR} (projection)"],
          color=SAND, lw=1.6, zorder=1)
ax.scatter(compare["World Bank / ICP 2024"], yy, s=32, color=AMBER, zorder=3,
           label="World Bank / ICP, 2024")
ax.scatter(compare[f"IMF {YEAR} (projection)"], yy, s=32, color=MARK, zorder=3,
           label=f"IMF WEO, {YEAR} projection")
i = list(compare.index).index("Israel")
ax.scatter(compare[f"IMF {YEAR} (projection)"].iloc[i], i, s=90, color=BRICK, zorder=4)
ax.scatter(compare["World Bank / ICP 2024"].iloc[i], i, s=90, color=BRICK, zorder=4)

ax.set_yticks(yy, compare.index)
ax.tick_params(axis="y", length=0)
ax.invert_yaxis()
for lbl in ax.get_yticklabels():
    if lbl.get_text() == "Israel":
        lbl.set_color(BRICK); lbl.set_fontweight("bold")
ax.axvline(100, color=MUTED, lw=1, zorder=0)
ax.set_xlabel("comparative price level, United States = 100")
ax.grid(axis="x", color="#EFE5CD", lw=.8)
ax.set_axisbelow(True)
ax.legend(loc="lower right")
ax.set_title("Two institutions, two answers, one ordering", pad=10)
plt.show()

display(compare.loc[["Iceland", "Switzerland", "Israel", "Australia",
                     "Norway", "Korea", "Türkiye"]])
print(f"rank correlation between the two columns: "
      f"{compare[f'IMF {YEAR} (projection)'].corr(compare['World Bank / ICP 2024'], method='spearman'):.3f}")

rank correlation between the two columns: 0.955

lines = [
    "/* ------------------------------------------------------------------",
    "   data.js - every OECD number used on this site lives here.",
    "",
    "   GENERATED FILE. Do not hand-edit.",
    "   Produced by analysis/israel-cost-of-living.ipynb, which pulls the",
    "   series from the IMF DataMapper API and asserts that every figure",
    "   quoted in the essay still reproduces. Re-run that notebook to",
    "   regenerate.",
    "",
    f"   Source: IMF World Economic Outlook, {YEAR} projections, indicators",
    "   NGDPDPC and PPPPC, OECD member countries only (38).",
    "",
    "   nominal = GDP per capita, current US$",
    "   ppp     = GDP per capita, current international $ (PPP)",
    "",
    "   Price level is DERIVED below: nominal / ppp * 100. That ratio is the",
    "   implied PPP conversion factor over the market exchange rate -- i.e.",
    "   the comparative price level, US = 100.",
    "   ------------------------------------------------------------------ */",
    "",
    "const OECD = {",
]
w = max(len(n) for n in tbl.index) + 3
for name, row in tbl.iterrows():
    key = '"' + name + '":'
    nom_v = int(row["nominal $"])
    ppp_v = int(row["PPP $"])
    lines.append(f"  {key:<{w}} [{nom_v:>6}, {ppp_v:>6}],")
lines[-1] = lines[-1].rstrip(",")
lines += ["};", ""]

js = "\n".join(lines)
tail = OUT_JS.read_text().split("/* ---- derived series")[1]
OUT_JS.write_text(js + "/* ---- derived series" + tail)
print(f"wrote {OUT_JS.relative_to(HERE.parent)}  ({OUT_JS.stat().st_size:,} bytes)\n")
print(js[js.index("const OECD"):][:520] + " …")

wrote assets/data.js  (3,427 bytes)

const OECD = {
  "Luxembourg":     [158733, 156719],
  "Ireland":        [140186, 159129],
  "Switzerland":    [126177, 105680],
  "Iceland":        [110048,  82730],
  "Norway":         [105877, 115548],
  "United States":  [ 94430,  94430],
  "Denmark":        [ 83445,  89667],
  "Netherlands":    [ 79918,  87773],
  "Australia":      [ 75648,  74755],
  "Sweden":         [ 70676,  77094],
  "Israel":         [ 69804,  59095],
  "Austria":        [ 67761,  78334],
  "Germany":        [ 65303,  76747],
  "Belgium" …

# Final self-check: the file just written must still carry the headline figures.
txt = OUT_JS.read_text()
assert '"Israel":' in txt and "69804" in txt and "59095" in txt
head = txt[txt.index("const OECD"):txt.index("/* ---- derived series")]
assert head.count("],") + 1 == 38, "expected 38 country rows"
print(f"38 countries written · Israel {int(isr['nominal $']):,} nominal / "
      f"{int(isr['PPP $']):,} PPP · price level {isr['price level']}")
print(f"cached inputs: {sorted(p.name for p in CACHE.glob('*'))}")

38 countries written · Israel 69,804 nominal / 59,095 PPP · price level 118.1
cached inputs: ['imf-weo-NGDPD.json', 'imf-weo-NGDPDPC.json', 'imf-weo-PPPPC.json', 'imf-weo-oecd.csv', 'pse_admpop_adm1_2023.csv', 'worldbank-PA.NUS.FCRF.json', 'worldbank-PA.NUS.PPP.json', 'worldbank-price-level-oecd.csv']

Indicator	Meaning
`NGDPDPC`	GDP per capita, current prices, U.S. dollars — converted at the market exchange rate
`PPPPC`	GDP per capita, current prices, international dollars — converted at purchasing power parity

	iso3	country	year	gdp_pc_usd	gdp_pc_ppp	price_level
0	AUS	Australia	1980	11021.309	10298.991	107.013483
1	AUS	Australia	1981	12544.129	11553.350	108.575686
2	AUS	Australia	1982	12250.122	12074.479	101.454663
3	AUS	Australia	1983	11606.529	12330.927	94.125357
4	AUS	Australia	1984	12590.852	13418.795	93.829975

	price_level	predicted	gap_pct
country
Israel	118.1	72.7	62.4
Iceland	133.0	90.7	46.7
New Zealand	89.2	72.1	23.7
Mexico	59.2	48.1	23.2
Australia	101.2	84.6	19.6

Israel's price level against the OECD — checking the essay's arithmetic¶

1. The comparison group¶

2. Pull the IMF series¶

3. The table the essay is built on¶

4. Assert the published figures¶

5. The fifteen-place fall¶

6. How expensive, ranked¶

7. More expensive than output explains¶

8. Is 118 a normal number for Israel?¶

9. An independent estimate¶

10. What holds, and what doesn't¶

11. Export the dataset the site reads¶

	nominal $	PPP $	price level	rank nominal	rank PPP	rank change
country
Luxembourg	158733	156719	101.3	1	2	-1
Ireland	140186	159129	88.1	2	1	1
Switzerland	126177	105680	119.4	3	4	-1
Iceland	110048	82730	133.0	4	8	-4
Norway	105877	115548	91.6	5	3	2
United States	94430	94430	100.0	6	5	1
Denmark	83445	89667	93.1	7	6	1
Netherlands	79918	87773	91.1	8	7	1
Australia	75648	74755	101.2	9	13	-4
Sweden	70676	77094	91.7	10	11	-1
Israel	69804	59095	118.1	11	26	-15
Austria	67761	78334	86.5	12	10	2
Germany	65303	76747	85.1	13	12	1
Belgium	65112	78607	82.8	14	9	5
United Kingdom	61056	67585	90.3	15	18	-3
Canada	60305	70006	86.1	16	14	2
Finland	60130	68861	87.3	17	15	2
France	52083	68567	76.0	18	17	1
New Zealand	52023	58308	89.2	19	27	-8
Italy	46505	65761	70.7	20	19	1
Spain	41563	59187	70.2	21	25	-4
Slovenia	40630	60664	67.0	22	22	0
Czechia	39795	63550	62.6	23	20	3
Estonia	37718	51653	73.0	24	29	-5
Korea	37412	68624	54.5	25	16	9
Lithuania	36545	61052	59.9	26	21	5
Japan	35703	59207	60.3	27	24	3
Portugal	35434	52841	67.1	28	28	0
Poland	31336	59792	52.4	29	23	6
Slovakia	31242	49466	63.2	30	31	-1
Greece	29696	47175	62.9	31	32	-1
Latvia	28913	45840	63.1	32	34	-2
Hungary	28430	50570	56.2	33	30	3
Costa Rica	20299	34157	59.4	34	36	-2
Chile	20240	37336	54.2	35	35	0
Türkiye	19018	46672	40.7	36	33	3
Mexico	15779	26643	59.2	37	37	0
Colombia	10104	23576	42.9	38	38	0

	year	x	sample	R²	Israel actual	Israel predicted	Israel gap %	Israel rank	runner-up
0	2024	linear	36 (drop LUX, IRL)	0.572	100.3	66.9	50	1	New Zealand
1	2024	linear	all 38	0.448	100.3	67.2	49	1	Iceland
2	2024	log	36 (drop LUX, IRL)	0.548	100.3	69.4	45	1	Mexico
3	2024	log	all 38	0.527	100.3	68.8	46	1	Iceland
4	2026	linear	36 (drop LUX, IRL)	0.507	118.1	72.7	62	1	Iceland
5	2026	linear	all 38	0.376	118.1	73.0	62	1	Iceland
6	2026	log	36 (drop LUX, IRL)	0.480	118.1	75.3	57	1	Iceland
7	2026	log	all 38	0.453	118.1	74.7	58	1	Iceland