"""New York's subway against London's Underground, the Paris Métro and Tokyo Metro since the pandemic: the analysis behind a SparkyData article. py src/build_article.py four-metros-since-2019 How much of its last normal year before the pandemic each of four big metros carries now, and how each grew over the thirty years before (card ``world_metro_trips``; New York's daily counts, card ``mta_ridership``). Each system is compared only with itself, because each counts a trip its own way: - New York: subway entries by calendar year (the MTA's published totals; 2025 the sum of its daily estimates), against 2019; this year from the daily file, and weekdays and weekends against the MTA's own comparable pre-pandemic days; - London: Underground journeys by financial year (April to March), against 2018/19, the last year March 2020 did not cut; this year from TfL's four-weekly periods, per day against the same periods of 2019/20; and the Elizabeth line beside it, since it opened through central London in May 2022; - Paris: RATP métro trips by calendar year, against 2019 (its 2019 ran 4% below 2018, so 2018 is shown as a check); - Tokyo: Tokyo Metro's boardings by fiscal year (April to March), against fiscal 2018, the last year the pandemic did not touch; fiscal 2025 is not yet published. A share of a base year is a ratio of two totals, with no sampling error; it is reported as it is. The long view indexes each system to its own base year (= 100) from its first year in the files. Writes results.json, viz.json, charts.json, data.csv and build.json into output/articles/four_metros_since_2019/ and site/public/articles/four-metros-since-2019/, and copies this script there as analysis.py. Deterministic: no sampling, and SEED is kept for any resampling. """ from __future__ import annotations import re import pandas as pd from article_kit import Article, lines, r, strip from nycdata import mta, world_population as P, world_transit as W from sparkylib import yoy SLUG = "four-metros-since-2019" SEED = 20261106 PARIS_RA = ("paris_rapport_annuel_2025_transports_en_commun", "paris_ra2025_transports_en_commun.pdf") MONTH_NAMES = ["January", "February", "March", "April", "May", "June", "July", "August", "September", "October", "November", "December"] #: system -> (label, base year, latest year, year kind) SHORT = {"new_york": "New York", "london": "London", "paris": "Paris", "tokyo": "Tokyo"} # end labels on the long view SYSTEMS = {"new_york": ("New York subway", 2019, 2025, "calendar"), "london": ("London Underground", 2018, 2025, "financial"), "paris": ("Paris Métro", 2019, 2025, "calendar"), "tokyo": ("Tokyo Metro", 2018, 2024, "fiscal")} def system_trips() -> dict[str, pd.Series]: """Each system's trips by year, in millions (London and Tokyo by the year their fiscal year starts in).""" t = W.tokyo() return {"new_york": W.new_york().set_index("year")["trips"] / 1e6, "london": W.london().set_index("fy")["journeys"], "paris": W.paris().set_index("year")["trips"], "tokyo": t[t["operator"].eq("Tokyo Metro")].set_index("fy")["boardings"] / 1e3} def label(system: str, year: int) -> str: kind = SYSTEMS[system][3] return f"{year}/{(year + 1) % 100:02d}" if kind == "financial" else (f"fiscal {year}" if kind == "fiscal" else str(year)) def main(argv: list[str] | None = None) -> int: """Run the analysis and write every file (no options).""" A = Article(SLUG, __file__, seed=SEED) y = system_trips() # ---- recovery: the latest year against each system's last normal year rec = {} for k, (name, base, latest, _) in SYSTEMS.items(): rec[k] = {"base_year": base, "latest_year": latest, "base_label": label(k, base), "latest_label": label(k, latest), "base": r(y[k][base], 1), "latest": r(y[k][latest], 1), "share_pct": r(100 * y[k][latest] / y[k][base], 1)} A.results["recovery"] = rec assert min(rec, key=lambda k: rec[k]["share_pct"]) == "new_york", rec # the page: New York has won back the least lon = W.london().set_index("fy") both = lon["journeys"] + lon["elizabeth"] A.results["london_with_elizabeth"] = {"base": r(both[2018], 1), "latest": r(both[2025], 1), "share_pct": r(100 * both[2025] / both[2018], 1), "elizabeth_latest": r(lon.loc[2025, "elizabeth"], 1), "elizabeth_base": r(lon.loc[2018, "elizabeth"], 1)} # the measures table: each system's base-year and latest totals, in millions A.results["table"] = {k: {"base_m": r(y[k][SYSTEMS[k][1]], 0), "latest_m": r(y[k][SYSTEMS[k][2]], 0)} for k in SYSTEMS} A.results["paris_vs_2019"] = r(100 * (y["paris"][2025] / y["paris"][2019] - 1), 1) A.results["paris_vs_2018"] = r(100 * y["paris"][2025] / y["paris"][2018], 0) A.results["paris_2019_vs_2018"] = r(100 * (1 - y["paris"][2019] / y["paris"][2018]), 0) lp = W.london_periods() p13 = lp[lp["period"] == 13].set_index("fy") A.results["london_march_2020"] = {"pct_below_2019": r(100 * (1 - (p13.loc[2019, "journeys"] / p13.loc[2019, "days"]) / (p13.loc[2018, "journeys"] / p13.loc[2018, "days"])), 0)} A.results["tokyo_fy2019_vs_fy2018"] = r(100 * (y["tokyo"][2019] / y["tokyo"][2018] - 1), 2) A.results["london_vs_2019_20"] = r(100 * y["london"][2025] / y["london"][2019], 1) A.results["tokyo_vs_fy2019"] = r(100 * y["tokyo"][2024] / y["tokyo"][2019], 1) tt = W.tokyo() toei = tt[tt["operator"].eq("Toei")].set_index("fy")["boardings"] A.results["toei"] = {"share_fy2018_pct": r(100 * toei[2024] / toei[2018], 0)} # ---- the thirty years before: each system's growth to its base year from its first year (or 1991) grow = {} for k, (name, base, _, _) in SYSTEMS.items(): first = max(int(y[k].index.min()), 1991) grow[k] = {"first": first, "first_label": label(k, first), "pct": r(100 * (y[k][base] / y[k][first] - 1), 0), "first_m": r(y[k][first], 0), "y2000_m": r(y[k][2000], 0), "y2000_label": label(k, 2000), "base_m": r(y[k][base], 0), "base_label": label(k, base)} A.results["growth"] = grow g2000 = {k: r(100 * (y[k][SYSTEMS[k][1]] / y[k][2000] - 1), 0) for k in SYSTEMS} A.results["growth_2000"] = g2000 assert max(g2000, key=g2000.get) == "london" and min(g2000, key=g2000.get) == "paris", g2000 # the page: fastest, slowest near = min((t for t in y["new_york"].index if t < 2019), key=lambda t: abs(y["new_york"][t] - y["new_york"][2025])) A.results["nyc_2025_like"] = {"year": int(near), "gap_pct": r(100 * (y["new_york"][2025] / y["new_york"][near] - 1), 2)} assert near == 1999 and abs(A.results["nyc_2025_like"]["gap_pct"]) < 1, A.results["nyc_2025_like"] # the page: "where it stood in 1999" A.results["low_2020"] = {k: r(100 * y[k][2020] / y[k][SYSTEMS[k][1]], 0) for k in SYSTEMS} A.results["y2022"] = {k: r(100 * y[k][2022] / y[k][SYSTEMS[k][1]], 0) for k in SYSTEMS} ny = y["new_york"] peak = int(ny.loc[:2019].idxmax()) # over the whole series, from 1975 A.results["nyc_peak"] = {"year": peak, "trips": r(ny[peak], 1), "fall_to_2019_pct": r(100 * (1 - ny[2019] / ny[peak]), 1)} # each other system's highest year before the pandemic, over its whole series (London from 1991/92, Paris and # Tokyo Metro from 2000): all three in their 2018 years others_peak = {k: int(y[k].loc[:2019].idxmax()) for k in ("london", "paris", "tokyo")} A.results["others_peak_year"] = others_peak assert peak == 2015 and set(others_peak.values()) == {2018}, (peak, others_peak) # the page's However ev = mta.fare_evasion() e19, e25 = ev[ev["year"] == 2019]["share"].mean(), ev[ev["year"] == 2025]["share"].mean() evq = ev.set_index(["year", "quarter"])["share"] daily_sub = mta.long() daily_sub = daily_sub[daily_sub["Mode"] == "Subway"].set_index("Date")["Count"] q25 = daily_sub.loc["2025-01-01":"2025-12-31"].groupby(lambda t: (t.month - 1) // 3 + 1).sum() imp19 = mta.implied_2019_by_month("Subway", base_year=2024, skip_leap_day=True) q19 = imp19.groupby(lambda m: (m - 1) // 3 + 1).sum() f25 = sum(q25[q] / (1 - evq[(2025, q)]) for q in range(1, 5)) / q25.sum() f19 = sum(q19[q] / (1 - evq[(2019, q)]) for q in range(1, 5)) / q19.sum() A.results["evasion"] = {"share_2019_pct": r(100 * e19, 0), "share_2025_pct": r(100 * e25, 0), "entries_share_pct": r(100 * (ny[2025] * f25) / (ny[2019] * f19), 0), "entries_share_simple_pct": r(100 * (ny[2025] / (1 - e25)) / (ny[2019] / (1 - e19)), 1)} # New York on London's April-to-March year, for the latest year too fy25 = daily_sub.loc["2025-04-01":"2026-03-31"] assert len(fy25) == 365 A.results["nyc_fy2025"] = {"share_pct": r(100 * fy25.sum() / 1e6 / ny[2019], 0)} daily = mta.long() sub = daily[daily["Mode"] == "Subway"].set_index("Date")["Count"] fy20 = sub.loc["2020-04-01":"2021-03-31"] assert len(fy20) == 365, len(fy20) A.results["nyc_fy2020"] = {"share_pct": r(100 * fy20.sum() / 1e6 / ny[2019], 0)} # per resident: a count set against a year decades away carries its reading per resident (style guide, section 4) pn = P.population("new_york").set_index("year")["pop"] pl = P.population("london").set_index("year")["pop"] A.results["per_resident"] = { "nyc": {str(t): r(1e6 * ny[t] / pn[t], 0) for t in (1999, 2019, 2025)}, "london": {str(t): r(1e6 * y["london"][t] / pl[t], 0) for t in (1995, 2018, 2025)}, } # ---- this year d = mta.subway_daily_years(through=(10, 28)) d = d.set_index("year") summ = mta.summary() last = pd.Timestamp(summ["last_date"]) imp = mta.implied_2019_by_month("Subway", base_year=2024, skip_leap_day=True) # whole months only, January to September, against the 2019 months the comparable days imply d9 = mta.subway_daily_years(through=(9, 30)).set_index("year") A.results["nyc_2026_vs_2019"] = {"share_pct": r(100 * d9.loc[2026, "trips"] / imp.loc[1:9].sum(), 0), "months": 9} dsub = mta.long() dsub = dsub[dsub["Mode"] == "Subway"].set_index("Date")["Count"] ch = yoy.compare_years(dsub, base_year=2025, current_year=2026, last=last, rng=A.rng) A.results["now_nyc_range"] = {"change_pct": r(ch.change_pct, 1), "lo_pct": r(ch.lo_pct, 1), "hi_pct": r(ch.hi_pct, 1)} if ch.direction == "up": A.claim("now_nyc_range", "up") win = yoy.compare_years(dsub, base_year=2025, current_year=2026, last="2026-08-22", start=(4, 1), rng=A.rng) A.results["nyc_london_window"] = {"change_pct": r(win.change_pct, 1), "lo_pct": r(win.lo_pct, 1), "hi_pct": r(win.hi_pct, 1), "per_day_m": r(win.mean_current / 1e6, 2), "per_day_2025_m": r(win.mean_base / 1e6, 2), "level": win.level} A.results["now_nyc"] = {"now": int(d.loc[2026, "trips"]), "before": int(d.loc[2025, "trips"]), "days": int(d.loc[2026, "days"]), "change_pct": r(100 * (d.loc[2026, "trips"] / d.loc[2025, "trips"] - 1), 1), "last_day": last.date().isoformat()} assert int(d.loc[2026, "days"]) == int(d.loc[2025, "days"]) == 301, d split = {(row["year"], row["day_type"]): row["Subway_vs_2019"] for row in summ["years"]} # whole years only: the baseline averages every day of 2024, so a part year would set January to September # against a whole year's weekends (the weekend baseline is seasonal; the weekday one is not) A.results["nyc_daytype"] = {str(yy): {"weekday": r(100 * split[(yy, "weekday")], 0), "weekend": r(100 * split[(yy, "weekend")], 0)} for yy in range(2021, 2026)} rows25 = {row["day_type"]: row for row in summ["years"] if row["year"] == 2025} base = summ["baseline"]["Subway"] miss = {t: rows25[t]["days"] * (base[t] - rows25[t]["Subway"]) for t in ("weekday", "weekend")} A.results["method"] = { "level": f"{ch.level:.0f}%", "this_year": ("the same days of 2025 and 2026, a matched-week bootstrap of the ratio (sparkylib.yoy.compare_years) " f"with seed {SEED}"), "new_york_years": "the MTA's published totals to 2024 (NYU Rudin Center's tabulation of NYCT counts before 2014); 2025 the sum of the daily file", "london_years": "the GLA's table to 2009/10, TfL's four-weekly periods from 2010/11", "paris_years": "SDES to 2024, the City of Paris for 2025", "tokyo_years": "each fiscal year's TMG statistical yearbook", "weekdays": "whole years against the 2019 baseline rebuilt from the MTA's comparable-day percentages in its daily file"} A.results["nyc_missing_2025"] = {"weekday_share_pct": r(100 * miss["weekday"] / sum(miss.values()), 0), "missing_millions": r(sum(miss.values()) / 1e6, 0), "annual_gap_millions": r(ny[2019] - ny[2025], 0)} p26, p19, p25 = W.london_part_year(2026, 5), W.london_part_year(2019, 5), W.london_part_year(2025, 5) per_period = lp[lp["fy"].isin([2025, 2026]) & (lp["period"] <= 5)].assign(pd_=lambda x: x["journeys"] / x["days"]) pp = per_period.pivot_table(index="period", columns="fy", values="pd_") A.results["london_periods_below"] = int((pp[2026] < pp[2025]).sum()) A.results["now_london"] = {"periods": 5, "first_day": p26["first_day"], "last_day": p26["last_day"], "per_day": r(p26["per_day"], 2), "per_day_2019": r(p19["per_day"], 2), "per_day_2025": r(p25["per_day"], 2), "share_2019_pct": r(100 * p26["per_day"] / p19["per_day"], 0), "vs_2025_pct": r(100 * (p26["per_day"] / p25["per_day"] - 1), 1), "with_elizabeth_share_pct": r(100 * (p26["per_day"] + p26["elizabeth_per_day"]) / (p19["per_day"] + p19["elizabeth_per_day"]), 0)} A.latest(p26["last_day"], "TfL Underground journeys, periods 1-5 of 2026/27; MTA daily subway ridership to " f"{A.results['now_nyc']['last_day']}") A.long_view(grow["new_york"]["first"], 2025, "Subway entries in New York, 1991-2025, and Underground journeys in London, " "1991/92-2025/26, each indexed to its last normal year; Paris from 2000, Tokyo Metro from fiscal 2000") # ---- the documents Q = A.quotes() quotes = {} for key, (src, needle) in {"paris_precovid": (PARIS_RA, "Par rapport à la situation pré-COVID, le réseau Métro est en hausse de 1,1 %"), "paris_line14": (PARIS_RA, "ligne 14 en avril 2024 jusqu")}.items(): q = Q.quote(src[0], src[1], needle) quotes[key] = {"quote": q["quote"], "page": q.get("page"), "source": q.get("source")} A.results["quotes"] = quotes A.results["history"] = {"paris_precovid_pct": float(re.search(r"hausse de ([\d,]+) %", quotes["paris_precovid"]["quote"]) .group(1).replace(",", "."))} # ---- the reproduction A.reproduce("London Underground journeys, 2024/25, millions (TfL Q4 2024/25 report)", published=1216, published_moe90=None, reproduced=r(y["london"][2024], 0), kind="count", tolerance=4) A.reproduce("Paris métro trips, 2025 against 2019 (the City of Paris's report), percent", published=A.results["history"]["paris_precovid_pct"], published_moe90=None, reproduced=A.results["paris_vs_2019"], kind="pct", tolerance=0.05) A.reproduce("New York subway trips, 2024, millions (the MTA's yearly page against its daily file)", published=r(y["new_york"][2024], 1), published_moe90=None, reproduced=r(mta.subway_daily_years().set_index("year").loc[2024, "trips"] / 1e6, 1), kind="count", tolerance=1.0) A.results["reproduction_source"] = "TfL's Q4 2024/25 Quarterly Performance Report; the City of Paris's 2025 annual report; the MTA" A.results["reproduction_note"] = ("TfL's report preceded later revisions (within 0.3%); the MTA's daily estimates come within " "0.06% of its published year.") # ---- printed figures R = A.results["recovery"] for k in SYSTEMS: A.printed(f"recovery.{k}.share_pct", f"{R[k]['share_pct']:.0f}%") A.printed("recovery.new_york.base", f"{R['new_york']['base'] / 1000:.2f} billion") A.printed("recovery.new_york.latest", f"{R['new_york']['latest'] / 1000:.2f} billion") A.printed("london_with_elizabeth.share_pct", f"{A.results['london_with_elizabeth']['share_pct']:.0f}%") A.printed("london_with_elizabeth.elizabeth_base", f"{A.results['london_with_elizabeth']['elizabeth_base']:.0f} million") for k in SYSTEMS: A.printed(f"table.{k}.base_m", f"{A.results['table'][k]['base_m']:,.0f}") A.printed(f"table.{k}.latest_m", f"{A.results['table'][k]['latest_m']:,.0f}") A.printed("now_nyc_range.lo_pct", f"{A.results['now_nyc_range']['lo_pct']:.1f}%") A.printed("now_nyc_range.hi_pct", f"{A.results['now_nyc_range']['hi_pct']:.1f}%") A.printed("london_with_elizabeth.elizabeth_latest", f"{A.results['london_with_elizabeth']['elizabeth_latest']:.0f} million") A.printed("paris_vs_2019", f"{A.results['paris_vs_2019']:.1f}%") A.printed("london_vs_2019_20", f"{A.results['london_vs_2019_20']:.0f}%") G = A.results["growth"] for k in SYSTEMS: A.printed(f"growth.{k}.pct", f"{G[k]['pct']:.0f}%") for k in SYSTEMS: A.printed(f"growth_2000.{k}", f"{g2000[k]:.0f}%") A.printed("low_2020.london", f"{A.results['low_2020']['london']:.0f}%") N = A.results["now_nyc"] A.printed("now_nyc.change_pct", f"{N['change_pct']:.1f}%") A.printed("now_nyc.last_day", f"{int(N['last_day'][8:])} {MONTH_NAMES[int(N['last_day'][5:7]) - 1]}") DT = A.results["nyc_daytype"] A.printed("nyc_daytype.2025.weekday", f"{DT['2025']['weekday']:.0f}%") A.printed("nyc_daytype.2025.weekend", f"{DT['2025']['weekend']:.0f}%") A.printed("nyc_2026_vs_2019.share_pct", f"{A.results['nyc_2026_vs_2019']['share_pct']:.0f}%") A.printed("nyc_peak.fall_to_2019_pct", f"{A.results['nyc_peak']['fall_to_2019_pct']:.0f}%") A.printed("evasion.share_2019_pct", f"{A.results['evasion']['share_2019_pct']:.0f}%") A.printed("evasion.share_2025_pct", f"{A.results['evasion']['share_2025_pct']:.0f}%") A.printed("evasion.entries_share_pct", f"{A.results['evasion']['entries_share_pct']:.0f}%") for t in ("1999", "2019", "2025"): A.printed(f"per_resident.nyc.{t}", f"{A.results['per_resident']['nyc'][t]:.0f}") for t in ("1995", "2018", "2025"): A.printed(f"per_resident.london.{t}", f"{A.results['per_resident']['london'][t]:.0f}") A.printed("nyc_fy2020.share_pct", f"{A.results['nyc_fy2020']['share_pct']:.0f}%") for k in SYSTEMS: A.printed(f"y2022.{k}", f"{A.results['y2022'][k]:.0f}%") L = A.results["now_london"] A.printed("now_london.vs_2025_pct", f"{abs(L['vs_2025_pct']):.1f}%") A.printed("now_london.last_day", f"{int(L['last_day'][8:])} {MONTH_NAMES[int(L['last_day'][5:7]) - 1]}") A.printed("history.paris_precovid_pct", "1.1%") A.printed("paris_vs_2018", f"{A.results['paris_vs_2018']:.0f}%") A.printed("paris_2019_vs_2018", f"{A.results['paris_2019_vs_2018']:.0f}%") A.printed("london_march_2020.pct_below_2019", f"{A.results['london_march_2020']['pct_below_2019']:.0f}%") A.printed("nyc_missing_2025.weekday_share_pct", f"{A.results['nyc_missing_2025']['weekday_share_pct']:.0f}%") A.printed("nyc_fy2025.share_pct", f"{A.results['nyc_fy2025']['share_pct']:.0f}%") A.printed("nyc_london_window.change_pct", f"{A.results['nyc_london_window']['change_pct']:.1f}%") A.printed("toei.share_fy2018_pct", f"{A.results['toei']['share_fy2018_pct']:.0f}%") A.printed("tokyo_vs_fy2019", f"{A.results['tokyo_vs_fy2019']:.0f}%") for k in SYSTEMS: if k in ("new_york", "london"): A.printed(f"growth.{k}.first_m", f"{G[k]['first_m']:,.0f}") A.printed(f"growth.{k}.y2000_m", f"{G[k]['y2000_m']:,.0f}") W_ = A.results["nyc_london_window"] A.printed("nyc_london_window.per_day_m", f"{W_['per_day_m']:.2f} million") A.printed("nyc_london_window.per_day_2025_m", f"{W_['per_day_2025_m']:.2f} million") A.printed("now_london.per_day", f"{L['per_day']:.2f} million") A.printed("now_london.per_day_2025", f"{L['per_day_2025']:.2f} million") A.printed("now_nyc.now", f"{N['now'] / 1e6:,.0f} million") A.printed("now_nyc.before", f"{N['before'] / 1e6:,.0f} million") A.printed("nyc_missing_2025.missing_millions", f"{A.results['nyc_missing_2025']['missing_millions']:,.0f} million") A.printed("nyc_missing_2025.annual_gap_millions", f"{A.results['nyc_missing_2025']['annual_gap_millions']:,.0f} million") # ---- charts A.chart("recovery", strip( title="New York's subway has won back the smallest share of its last normal year", subtitle="Trips in 2022 and in each system's latest year, as a share of its last normal year before the pandemic", x=[["2022", "2022 (2022/23 in London, fiscal 2022 in Tokyo)"], ["latest", "Latest year"]], series=[[k, f"{SYSTEMS[k][0]} ({R[k]['latest_label']} against {R[k]['base_label']})", SYSTEMS[k][0]] for k in SYSTEMS], cells={"": {k: {"2022": A.results["y2022"][k], "latest": R[k]["share_pct"]} for k in SYSTEMS}}, format="pct", axis="Share of the last normal year", note="Each system against itself, because each counts a trip its own way: New York entries (2019), London's " "Underground journeys (2018/19, April to March), Paris's métro entries (2019), " "Tokyo Metro's boardings (fiscal 2018). New York's are paid entries; counting fare evaders it would stand at " "about four-fifths of 2019.", provenance="MTA; Transport for London; SDES and the City of Paris; Tokyo Metropolitan Government")) years = list(range(1991, 2026)) lowest = {} for t in range(2021, 2026): shares = {k: y[k][t] / y[k][SYSTEMS[k][1]] for k in SYSTEMS if t in y[k].index} lowest[str(t)] = min(shares, key=shares.get) A.results["lowest_by_year"] = lowest # the long-view caption: New York lowest of the four from 2021 assert set(lowest.values()) == {"new_york"}, lowest def idx(k): base = y[k][SYSTEMS[k][1]] return {str(t): r(100 * y[k][t] / base, 1) for t in years if t in y[k].index} A.chart("long_view", lines( title="New York's subway peaked in 2015; the others rose to records before the pandemic", subtitle="Trips each year as a share of each system's last normal year before the pandemic (= 100)", x=[[str(t), str(t)] for t in years], series=[[k, SYSTEMS[k][0], SHORT[k]] for k in SYSTEMS], cells={"": {k: idx(k) for k in SYSTEMS}}, format="dec1", axis="Index, last normal year = 100", x_ticks=[str(t) for t in years if t % 5 == 0 or t in (1991, 2025)], note="New York and Paris by calendar year; London and Tokyo by the year their fiscal year (April to March) starts " "in. London from 1991/92, Paris from 2000, Tokyo Metro from fiscal 2000; Tokyo's fiscal 2025 is not yet published.", provenance="MTA and NYU Rudin Center; Transport for London and the GLA; SDES and the City of Paris; Tokyo Metropolitan Government")) dt_years = sorted(DT) A.chart("weekdays", lines( title="New York's weekend riding has recovered more than its weekday riding", subtitle="Average subway trips a day, Monday to Friday and Saturday and Sunday, as a share of the MTA's comparable days of 2019", x=[[yy, yy] for yy in dt_years], series=[["weekend", "Saturdays and Sundays", "Weekends"], ["weekday", "Monday to Friday (holidays included)", "Weekdays"]], cells={"": {t: {yy: DT[yy][t] for yy in dt_years} for t in ("weekday", "weekend")}}, format="pct", axis="Share of 2019", note="Whole years. The MTA's own comparable pre-pandemic days, recovered from its daily file; paid entries only.", provenance="MTA daily ridership")) rows_out = [] for k in SYSTEMS: for t, v in y[k].items(): rows_out.append({"system": SYSTEMS[k][0], "year": int(t), "year_kind": SYSTEMS[k][3], "trips_millions": r(v, 3)}) for t, row in lon.iterrows(): rows_out.append({"system": "London Elizabeth line (TfL Rail)", "year": int(t), "year_kind": "financial", "trips_millions": r(row["elizabeth"], 3)}) A.write(data=pd.DataFrame(rows_out)) print(f"{SLUG}: wrote {A.out}") return 0 if __name__ == "__main__": raise SystemExit(main())