Placebo test: shuffle sentiment within ticker

Shuffle score_mean within ticker (destroys time ordering) and re-run predictive panel FE.

Updated: 2026-04-05

Study

Robustness • Status: LIVE

Updated: 2026-04-05

β real

-0.000133

β placebo

-8.95e-05

Seed

Tickers

507

Sample ticker

—

Tickers (panel)

507

Obs (panel)

261179

R² (TS)

—

R² (FE)

—

Key findings

Placebo test: shuffling sentiment should remove any true predictive relationship.
Baseline β=-0.00013323464208029765 vs placebo β=-8.950510769708544e-05.
Interpretation: if placebo collapses toward zero while baseline does not, the signal is less likely to be driven only by fixed cross-sectional differences.

Specification

Compare baseline panel FE vs placebo with permuted score_mean.

Data

Universe: S&P 500 tickers in your snapshot pipeline.
Frequency: Daily.
Returns: log(P_t) − log(P_{t−1}) from ticker JSON prices.
Sentiment: score_mean from your sentiment pipeline (ticker JSON).

Limitations

Timing: after-close articles can contaminate same-day results; predictive tests mitigate this but do not fully solve it.
Causality: results are descriptive; omitted variables and measurement error remain.
Trading: no transaction costs / slippage / capacity modeled.

References (minimal)

Newey, W. K., & West, K. D. (1987). A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix.
Petersen, M. A. (2009). Estimating standard errors in finance panel data sets.

Methodology

Within each ticker, randomly permute the sentiment time series (deterministic seed).
Re-estimate y_ret(t+1) ~ score_mean(t) (+ controls) with panel FE and ticker-clustered SE.

Figures

Charts are from the sample ticker (not the full panel).

Series (sample)

No series available.

Sentiment (sample)

score_mean

No sentiment series available. Export results.series.score_mean.

Tables

Placebo (shuffle within ticker): panel FE comparison

Spec	β(score_mean)	p	Sig
Baseline	-1.332e-4	0.19931	0
Placebo (seed=7)	-8.951e-5	0.422631	0

Note: exported by the Python builder; formatting is intentionally compact.

Models

This study is primarily descriptive (figures/tables). No regression model outputs were exported.

Appendix

Raw exported objects (reproducibility / debugging).

Raw exported JSON (collapsed)

{
  "results": {
    "n_tickers": 507,
    "n_obs_panel": 261179,
    "panel_fe_baseline": {
      "params": {
        "score_mean": -0.00013323464208029765
      },
      "bse": {
        "score_mean": 0.00010380408330189446
      },
      "tvalues": {
        "score_mean": -1.2835202416152551
      },
      "pvalues": {
        "score_mean": 0.1993098726609117
      },
      "nobs": 260672,
      "rsquared": 0.000006083300493009425,
      "rsquared_adj": 0.00000224707046869721,
      "cov_type": "cluster(ticker)"
    },
    "panel_fe_placebo": {
      "params": {
        "score_mean": -0.00008950510769708544
      },
      "bse": {
        "score_mean": 0.00011162110265869739
      },
      "tvalues": {
        "score_mean": -0.8018654677759655
      },
      "pvalues": {
        "score_mean": 0.42263078337295856
      },
      "nobs": 260672,
      "rsquared": 0.00000274536813915649,
      "rsquared_adj": -0.0000010908746903570687,
      "cov_type": "cluster(ticker)"
    },
    "tables": [
      {
        "title": "Placebo (shuffle within ticker): panel FE comparison",
        "columns": [
          "Spec",
          "β(score_mean)",
          "p",
          "Sig"
        ],
        "rows": [
          [
            "Baseline",
            -0.00013323464208029765,
            0.1993098726609117,
            ""
          ],
          [
            "Placebo (seed=7)",
            -0.00008950510769708544,
            0.42263078337295856,
            ""
          ]
        ]
      }
    ]
  },
  "methodology": [
    "Within each ticker, randomly permute the sentiment time series (deterministic seed).",
    "Re-estimate y_ret(t+1) ~ score_mean(t) (+ controls) with panel FE and ticker-clustered SE."
  ],
  "sections": [
    {
      "title": "Specification",
      "bullets": [
        "Compare baseline panel FE vs placebo with permuted score_mean."
      ]
    },
    {
      "title": "Data",
      "bullets": [
        "Universe: S&P 500 tickers in your snapshot pipeline.",
        "Frequency: Daily.",
        "Returns: log(P_t) − log(P_{t−1}) from ticker JSON prices.",
        "Sentiment: score_mean from your sentiment pipeline (ticker JSON)."
      ]
    },
    {
      "title": "Limitations",
      "bullets": [
        "Timing: after-close articles can contaminate same-day results; predictive tests mitigate this but do not fully solve it.",
        "Causality: results are descriptive; omitted variables and measurement error remain.",
        "Trading: no transaction costs / slippage / capacity modeled."
      ]
    },
    {
      "title": "References (minimal)",
      "bullets": [
        "Newey, W. K., & West, K. D. (1987). A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix.",
        "Petersen, M. A. (2009). Estimating standard errors in finance panel data sets."
      ]
    }
  ],
  "conclusions": [
    "Placebo test: shuffling sentiment should remove any true predictive relationship.",
    "Baseline β=-0.00013323464208029765 vs placebo β=-8.950510769708544e-05.",
    "Interpretation: if placebo collapses toward zero while baseline does not, the signal is less likely to be driven only by fixed cross-sectional differences."
  ]
}