evo-macro-cycle-correlation

Deflates nominal macroeconomic series, applies HP filter for cycle extraction, and computes Pearson correlation between cyclical components.

Key Functions

deflate_nominal_to_real(nominal, cpi)

Converts nominal to real values: Real_t = (Nominal_t / CPI_t) * 100

extract_hp_cycle(real_series, lamb=100)

Applies np.log() then HP filter (statsmodels hpfilter) to extract cyclical component.

• Uses natural log (np.log, NOT np.log10)
• hpfilter returns (cycle, trend) - parameter name is lamb
• lamb=100 for annual data (standard)

compute_cycle_pearson_correlation(cycle1, cycle2)

Computes Pearson r using scipy.stats.pearsonr (returns PearsonRResult.statistic)

run_full_pipeline(pce_file, pfi_file, cpi_file, ...)

Complete end-to-end pipeline: parse -> deflate -> HP filter -> correlate -> output. Depends on evo-macro-data-ingestion for data parsing.

Usage

import sys
sys.path.insert(0, '/app/environment/skills/evo-macro-cycle-correlation/scripts')
from utils import run_full_pipeline

correlation = run_full_pipeline(
    pce_file='/root/ERP-2025-table10.xls',
    pfi_file='/root/ERP-2025-table12.xls',
    cpi_file='/root/CPI.xlsx',
    start_year=1973,
    end_year=2024,
    hp_lambda=100,
    output_file='/root/answer.txt'
)

Technical Details

Deflation

• Formula: Real = (Nominal / CPI) * 100
• CPI base period doesn't matter for HP cycle analysis (log differences cancel it)

HP Filter

• Import: from statsmodels.tsa.filters.hp_filter import hpfilter
• Signature: hpfilter(x, lamb=1600) - use lamb=100 for annual data
• Returns: (cycle, trend) tuple
• Input must have NO NaN values

Pearson Correlation

• scipy.stats.pearsonr returns PearsonRResult object (scipy 1.14.1)
• Access correlation via .statistic attribute
• Output formatted with f"{value:.5f}" for exactly 5 decimal places

Year Range

• Use pandas .loc[1973:2024] for inclusive filtering
• .loc is inclusive of both start and stop labels