Analyze Earnings Quality for Free with Python and SEC EDGAR

Python Investor’s Toolkit: Part 1: Momentum · Part 2: Monte Carlo · Part 3: Mean Reversion · Part 4: Earnings Quality · Part 5: Macro Regime

Part 4 of 5 — The Python Investor’s Toolkit: Practical quant tools built with free data.

Price and momentum tell you what the market thinks about a stock. Earnings quality analysis tells you whether the market is right to believe it.

The Sloan Accrual Anomaly — published in 1996 and still generating alpha today — shows that companies with high accounting accruals relative to their assets systematically underperform. The reason: when earnings are driven by accruals (accounting adjustments) rather than cash flow, those earnings are fragile. They tend to reverse. The market, for years, consistently mispriced this. And to a meaningful extent, it still does.

This tool pulls data directly from SEC EDGAR’s free public API — no Bloomberg, no financial data subscription, no API key — and computes the core earnings quality metrics for any public company.

What You’ll Need

pip install requests pandas

The Key Metrics Explained

Finding a Company’s CIK Number

Every public company has a CIK (Central Index Key) — a unique identifier in the SEC system. Find it at EDGAR Full-Text Search or use this quick lookup function in the code below.

The Code

import requests
import pandas as pd

# Your info (required by SEC for User-Agent)
HEADERS = {'User-Agent': 'YourName your@email.com'}

def find_cik(company_name):
    """
    Look up a company's CIK by name using EDGAR's company search API.
    Returns the first match — refine the name if results are ambiguous.
    """
    # Direct CIK lookup (use known CIKs for reliability)
    known_ciks = {
        'AAPL': 320193,
        'MSFT': 789019,
        'AMZN': 1018724,
        'GOOGL': 1652044,
        'META': 1326801,
        'NVDA': 1045810,
        'TSLA': 1318605,
        'JPM': 19617,
        'JNJ': 200406,
        'XOM': 34088,
        'WMT': 104169,
        'BAC': 70858,
        'BRK-B': 1067983,
    }
    return known_ciks.get(company_name.upper())

def get_company_facts(cik):
    """Download all XBRL financial facts for a company from SEC EDGAR."""
    cik_padded = str(cik).zfill(10)
    url = f"https://data.sec.gov/api/xbrl/companyfacts/CIK{cik_padded}.json"
    response = requests.get(url, headers=HEADERS)
    if response.status_code != 200:
        raise ValueError(f"No data found for CIK {cik}. Check the CIK is correct.")
    return response.json()

def extract_annual(facts, concept, unit='USD'):
    """
    Extract annual (10-K) values for a specific financial concept.
    Common concepts: NetIncomeLoss, NetCashProvidedByUsedInOperatingActivities, Assets
    """
    try:
        data = facts['facts']['us-gaap'][concept]['units'][unit]
        df = pd.DataFrame(data)
        df = df[df['form'] == '10-K'].copy()
        df['end'] = pd.to_datetime(df['end'])
        # Keep one entry per fiscal year end
        df = df.sort_values('end').drop_duplicates('end', keep='last')
        return df[['end', 'val']].rename(columns={'val': concept}).reset_index(drop=True)
    except KeyError:
        return pd.DataFrame()

def analyze_earnings_quality(ticker_or_cik, label=None):
    """
    Full earnings quality report for a company using SEC EDGAR data.
    Pass either a ticker symbol (for known companies) or a CIK integer.
    """
    if isinstance(ticker_or_cik, str):
        cik = find_cik(ticker_or_cik)
        label = label or ticker_or_cik
        if not cik:
            print(f"CIK not found for {ticker_or_cik}. Add it to known_ciks dict.")
            return None
    else:
        cik = ticker_or_cik
        label = label or f"CIK {cik}"

    print(f"nFetching EDGAR data for {label} (CIK: {cik})...")
    facts = get_company_facts(cik)

    # Pull three core line items
    ni  = extract_annual(facts, 'NetIncomeLoss')
    cfo = extract_annual(facts, 'NetCashProvidedByUsedInOperatingActivities')
    ta  = extract_annual(facts, 'Assets')

    if ni.empty or cfo.empty or ta.empty:
        print(f"  Incomplete data — one or more line items missing for {label}")
        return None

    # Merge on fiscal year end date
    df = ni.merge(cfo, on='end').merge(ta, on='end')
    df.columns = ['end', 'net_income', 'cfo', 'total_assets']
    df['year'] = df['end'].dt.year

    # Compute earnings quality metrics
    df['avg_assets']     = (df['total_assets'] + df['total_assets'].shift(1)) / 2
    df['accruals']       = df['net_income'] - df['cfo']
    df['accruals_ratio'] = (df['accruals'] / df['avg_assets']).round(3)
    df['cash_conversion']= (df['cfo'] / df['net_income']).round(2)

    def signal(row):
        good = row['accruals_ratio'] < -0.03 and row['cash_conversion'] > 1.05
        bad  = row['accruals_ratio'] >  0.05 or  row['cash_conversion'] < 0.80
        return 'HIGH QUALITY' if good else ('INVESTIGATE' if bad else 'AVERAGE')

    df['signal'] = df.apply(signal, axis=1)

    # Format output
    out = df.dropna(subset=['accruals_ratio']).tail(6).copy()
    out['net_income'] = out['net_income'].apply(lambda x: f"${x/1e9:.1f}B")
    out['cfo']        = out['cfo'].apply(lambda x: f"${x/1e9:.1f}B")

    print(f"n{'='*65}")
    print(f"  Earnings Quality Report: {label}")
    print(f"{'='*65}")
    print(out[['year','net_income','cfo','accruals_ratio',
               'cash_conversion','signal']].to_string(index=False))
    print(f"{'='*65}")
    return df

# -------------------------------------------------------
# Run analysis on multiple companies
# -------------------------------------------------------
for ticker in ['AAPL', 'MSFT', 'AMZN', 'NVDA', 'META']:
    analyze_earnings_quality(ticker)

What Good vs. Bad Looks Like

Apple typically shows cash conversion ratios above 1.1 — it consistently converts more than $1.10 of operating cash for every $1.00 of reported net income. That’s a hallmark of high-quality earnings. Contrast this with companies that report growing net income while operating cash flow stagnates: that divergence is one of the most reliable early warning signs of future earnings disappointment or restatement risk.

What’s Next

Part 5 closes the series by zooming out to the macro picture — building a regime detector using FRED’s free API that tells you whether the economy is in expansion, late-cycle, slowdown, or recession risk, and what that implies for your asset allocation.

Series: The Python Investor’s Toolkit
Part 1: Momentum Scanner
Part 2: Monte Carlo Portfolio Stress-Test
Part 3: Mean Reversion Alert System
Part 4: Earnings Quality Analyzer — SEC EDGAR (this post)
Part 5: Macro Regime Detector (FRED API)