Skip to content

public-finance

Category: data-handling
Field: economics
License: private (curator-owned)
Updated: 2026-05-20
Stages: data-acquisition

Curator-private skill — copy text from 100xOS/shared/skills/data/public-finance.md.

Public Finance & Macro Data Skill

FRED API Patterns

The Federal Reserve Economic Data (FRED) API is the standard source for U.S. macro time series. Key series for finance research:

Interest Rates & Monetary Policy

  • DFF — Federal funds effective rate (daily)
  • DGS10 — 10-year Treasury constant maturity (daily)
  • DGS2 — 2-year Treasury constant maturity (daily)
  • T10Y2Y — 10-Year minus 2-Year spread (yield curve)
  • DFEDTARU — Fed funds target upper limit

Prices & Inflation

  • CPIAUCSL — CPI for all urban consumers (monthly, seasonally adjusted)
  • CPILFESL — Core CPI (ex food & energy)
  • PCEPI — PCE price index (Fed's preferred measure)
  • BREAKEVEN10 — 10-year breakeven inflation rate

Output & Activity

  • GDP — Gross domestic product (quarterly, billions)
  • GDPC1 — Real GDP (chained 2017 dollars)
  • UNRATE — Unemployment rate (monthly)
  • PAYEMS — Total nonfarm payrolls (monthly)
  • INDPRO — Industrial production index

Financial Conditions

  • VIXCLS — CBOE VIX (daily)
  • BAMLH0A0HYM2 — ICE BofA high-yield OAS (daily)
  • DTWEXBGS — Trade-weighted U.S. dollar index (daily)
  • SP500 — S&P 500 (daily, use Yahoo Finance for more granular)

Money & Credit

  • M2SL — M2 money supply
  • TOTCI — Total commercial & industrial loans
  • BUSLOANS — Commercial & industrial loans

Yahoo Finance Patterns

Use yfinance for market price data: - Equity indices: ^GSPC (S&P 500), ^DJI (Dow), ^IXIC (Nasdaq) - Volatility: ^VIX - Commodities: GC=F (gold), CL=F (crude oil), BTC-USD - ETFs as proxies: SPY, QQQ, TLT (long bonds), HYG (high yield) - Individual equities by ticker

Merge Strategies

When combining macro data with micro-level panel data:

  1. Frequency alignment: Macro series are daily/monthly/quarterly. Panel data is often at irregular frequencies. Merge on the appropriate time unit.
  2. For daily panel data: merge on exact date using FRED daily series
  3. For monthly panels: use end-of-month or average macro values

  4. For event studies: use the event date to look up the macro variable level

  5. Lags and leads: In many specifications, you want the macro variable before the micro outcome:

    Text Only
    macro_lag1 = macro_value.shift(1)  # Previous period

  6. Transformations:

  7. Log differences for growth rates: np.log(series).diff()
  8. Standardize for cross-variable comparisons: (x - x.mean()) / x.std()
  9. HP filter for trend/cycle decomposition (use statsmodels.tsa.filters.hp_filter)

  10. First differences for non-stationary series

  11. Controls vs. instruments: Macro variables are typically controls in micro regressions (absorb time-varying aggregate shocks). When used as instruments, the exclusion restriction must be argued carefully — macro shocks affect many channels simultaneously.

Common Pitfalls

  • Vintage effects: FRED revises macro data. For real-time analysis, use the ALFRED (Archival FRED) real-time dataset.
  • Seasonality: Prefer seasonally adjusted (SA) series unless studying seasonal patterns. FRED SA series end in SL or SA.
  • Unit roots: Most macro series are non-stationary. Test with ADF before running regressions in levels. Use differences or cointegration as appropriate.
  • Generated regressors: If you construct a macro variable (e.g., residual from a time-series model) and use it as a regressor, standard errors need correction (Murphy-Topel or bootstrap).