dadams/texas-district-analysis
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

drafts are done. going to convert to docx and share

Browse Source
This commit is contained in:
dadams-AU
2026-02-18 20:03:21 -08:00
parent 64777d15e1
commit 61b70c7133
2 changed files with 22 additions and 27 deletions
Download Patch File Download Diff File Expand all files Collapse all files

28
analysis/draft.md
View File

@@ -57,32 +57,24 @@ $$
Y_{dt} = \alpha_d + \beta_1 \mathrm{YearNum}_t + \beta_2 \mathrm{Post2019}_t + \beta_3 \mathrm{PostTrend}_t + \varepsilon_{dt} Y_{dt} = \alpha_d + \beta_1 \mathrm{YearNum}_t + \beta_2 \mathrm{Post2019}_t + \beta_3 \mathrm{PostTrend}_t + \varepsilon_{dt}
$$ $$
Where $(PostTrend_t = \max(0, t-2018))$. This distinguishes an immediate post-2019 level shift $(\beta_2)$ from a post-policy slope change $(\beta_3)$. Where $PostTrend_t = \max(0, t-2018)$. This distinguishes an immediate post-2019 level shift $\beta_2$ from a post-policy slope change $\beta_3$.
This follows interrupted time-series logic for a common policy shock, separating immediate and gradual responses (Biglan, Ary, & Wagenaar, 2000; Bernal, Cummins, & Gasparrini, 2017; Linden, 2015). This follows interrupted time-series logic for a common policy shock, separating immediate and gradual responses (Biglan, Ary, & Wagenaar, 2000; Bernal, Cummins, & Gasparrini, 2017; Linden, 2015).
### Model 2: District heterogeneity (H2) ### Model 2: District heterogeneity (H2)
$$ Model 2 estimates district-specific post-policy effects by interacting district indicators with the post-2019 period and including year fixed effects. This yields district-level post coefficients and a joint heterogeneity test. In the current run, the omnibus test of district-by-post terms is not statistically significant (Wald chi-square = 0.670, p=0.4130), so H2 evidence is interpreted primarily from the pattern and magnitude of district-specific estimates rather than a single global rejection.
Y_{dt} = \alpha_d + \gamma_t + \sum_{d} \theta_d \bigl(\mathrm{District}_d \times \mathrm{Post2019}_t\bigr) + \varepsilon_{dt}
$$
This yields district-specific post-policy effects and a joint heterogeneity test. In the current run, the omnibus test of district-by-post terms is not statistically significant (Wald chi-square = 0.670, p=0.4130), so H2 evidence is interpreted primarily from the pattern and magnitude of district-specific estimates rather than a single global rejection.
Because all districts are exposed in the same year, this is not a staggered-adoption DiD problem. Still, recent DiD work highlights that pooled average effects can mask meaningful treatment-effect heterogeneity, so we estimate district-specific post effects directly rather than rely on a single pooled interaction (de Chaisemartin & D'Haultfœuille, 2020; Goodman-Bacon, 2021; Sun & Abraham, 2021). Because all districts are exposed in the same year, this is not a staggered-adoption DiD problem. Still, recent DiD work highlights that pooled average effects can mask meaningful treatment-effect heterogeneity, so we estimate district-specific post effects directly rather than rely on a single pooled interaction (de Chaisemartin & D'Haultfœuille, 2020; Goodman-Bacon, 2021; Sun & Abraham, 2021).
### Model 3: Offshore moderation (H5) ### Model 3: Offshore moderation (H5)
$$ Model 3 adds an offshore-by-post interaction term to Model 2 to test whether offshore-regulating districts differ systematically from other districts after conditioning on district-specific post-policy shifts.
Y_{dt} = \alpha_d + \gamma_t + \sum_{d} \theta_d \bigl(\mathrm{District}_d \times \mathrm{Post2019}_t\bigr) + \phi\bigl(\mathrm{Post2019}_t \times \mathrm{Offshore}_d\bigr) + \varepsilon_{dt}
$$
Where `Offshore_d = 1` for districts 02/03/04.
This specification tests whether offshore-regulating districts differ systematically from other districts after controlling for district-specific post-policy shifts.
### Spatial diagnostic (H4) ### Spatial diagnostic (H4)
After estimating district treatment effects, we test for global spatial autocorrelation using permutation-based Moran's I (Anselin, 1995). The statistic is computed from a manually specified district contiguity matrix and evaluated with randomization inference (5,000 permutations), assessing whether high- and low-response districts are geographically clustered in ways consistent with diffusion or regional administrative spillovers. After estimating district treatment effects, we test for global spatial autocorrelation using permutation-based Moran's I (Anselin, 1995). The statistic is computed from a manually specified district contiguity matrix and evaluated with randomization inference (5,000 permutations), assessing whether high- and low-response districts are geographically clustered in ways consistent with diffusion or regional administrative spillovers.
All models use district-clustered standard errors. All models use district-clustered standard errors. Full formal equations and notation are reported in the appendix.
## Results ## Results
@@ -99,8 +91,8 @@ Figure 1 visualizes these system-level changes across the regulatory pipeline. T
**Model 1 (timing outcome):** **Model 1 (timing outcome):**
- Immediate post-2019 level shift: **0.1514**, p=0.3294. - Immediate post-2019 level shift: 0.1514, p=0.3294.
- Post-2019 slope shift: **-0.3603**, p=0.0010. - Post-2019 slope shift: -0.3603, p=0.0010.
Interpretation: no statistically significant immediate level break in 2019, but a significant post-policy acceleration trend in enforcement timing. Interpretation: no statistically significant immediate level break in 2019, but a significant post-policy acceleration trend in enforcement timing.
@@ -160,9 +152,9 @@ The map indicates that large positive and negative effects coexist across region
In the conditional heterogeneity model (Model 3): In the conditional heterogeneity model (Model 3):
- Offshore-by-post-policy differential = **0.3819**, p<0.001. - Offshore-by-post-policy differential = 0.3819, p<0.001.
This estimand is the average post-2019 offshore differential conditional on district-specific post-policy effects. Because the outcome is logged enforcement delay, the coefficient implies an approximate $(e^{0.3819}-1 \approx 46.5\%)$ relative increase in time-to-enforcement for offshore-regulating districts, holding the rest of the specification constant. Read jointly with the annual offshore differential results (Table C3 in the appendix), this pooled estimate should be interpreted as an average over uneven yearly effects, with the strongest divergences concentrated in 2023-2024. Substantively, H5 is supported as a structured heterogeneity result within the all-district analysis, but it should not be interpreted as isolating a single offshore causal mechanism, given that offshore jurisdiction is concentrated in districts 02/03/04. This estimand is the average post-2019 offshore differential conditional on district-specific post-policy effects. Because the outcome is logged enforcement delay, the coefficient implies an approximate $e^{0.3819}-1 \approx 46.5\%$ relative increase in time-to-enforcement for offshore-regulating districts, holding the rest of the specification constant. Read jointly with the annual offshore differential results (Table C3 in the appendix), this pooled estimate should be interpreted as an average over uneven yearly effects, with the strongest divergences concentrated in 2023-2024. Substantively, H5 is supported as a structured heterogeneity result within the all-district analysis, but it should not be interpreted as isolating a single offshore causal mechanism, given that offshore jurisdiction is concentrated in districts 02/03/04.
### H3: Structural moderators ### H3: Structural moderators
@@ -204,7 +196,7 @@ Table 3 summarizes why structural accounts are only partially successful in this
Moran’s I on district effects: Moran’s I on district effects:
- \(I = -0.0493\), permutation p=0.8550. - $I = -0.0493$, permutation p=0.8550.
No evidence of statistically significant global spatial autocorrelation. No evidence of statistically significant global spatial autocorrelation.
In this run, the spatial evidence is inferential (permutation Moran’s I) rather than model-based spatial plotting; Figure 4 provides the relevant geographic context for district-level effect dispersion. In this run, the spatial evidence is inferential (permutation Moran’s I) rather than model-based spatial plotting; Figure 4 provides the relevant geographic context for district-level effect dispersion.
@@ -234,7 +226,7 @@ In this run, the spatial evidence is inferential (permutation Moran’s I) rathe
- Linear interrupted model: post-trend -67.0420 days (p=0.0100). - Linear interrupted model: post-trend -67.0420 days (p=0.0100).
- Winsorized interrupted model: post-trend -0.3147 (p=0.0016). - Winsorized interrupted model: post-trend -0.3147 (p=0.0016).
Across variants, the post-policy **slope** result is more stable than the immediate **level** effect. Across variants, the post-policy slope result is more stable than the immediate level effect.
**Table 4. Robustness summary (interrupted panel framework)** **Table 4. Robustness summary (interrupted panel framework)**

21
analysis/draft_appendix.md
View File

@@ -1,11 +1,12 @@
# Appendix: Heterogeneous Enforcement of Transparency # Appendix: Heterogeneous Enforcement of Transparency
## Evidence from the Texas Railroad Commission ## Evidence from the Texas Railroad Commission
## Appendix A. Data Construction and Variables ## Appendix A. Data Construction and Variables
### A1. Data integration ### A1. Data integration
The analysis combines inspection and violation administrative records (2015-2025) into a district-year panel. The estimation sample contains 143 district-year observations across 13 districts (52 pre-policy; 91 post-policy). Well-level linkage is done via `api_norm`. The analysis combines inspection and violation administrative records (2015-2025) into a district-year panel. The estimation sample contains 143 district-year observations across 13 districts (52 pre-policy; 91 post-policy). Well-level records are linked across sources prior to district-year aggregation.
### A1b. Pipeline volume and sample flow ### A1b. Pipeline volume and sample flow
@@ -45,21 +46,21 @@ The specification sequence follows the main text: a common-shock interrupted pan
### B1. Interrupted panel (all districts; H1) ### B1. Interrupted panel (all districts; H1)
\[ $$
Y_{dt}=\alpha_d + \beta_1 \text{YearNum}_t + \beta_2 \text{Post2019}_t + \beta_3 \text{PostTrend}_t + \varepsilon_{dt} Y_{dt}=\alpha_d + \beta_1 \text{YearNum}_t + \beta_2 \text{Post2019}_t + \beta_3 \text{PostTrend}_t + \varepsilon_{dt}
\] $$
### B2. District heterogeneity (H2) ### B2. District heterogeneity (H2)
\[ $$
Y_{dt}=\alpha_d + \gamma_t + \sum_d \theta_d (\text{District}_d\times \text{Post2019}_t) + \varepsilon_{dt} Y_{dt}=\alpha_d + \gamma_t + \sum_d \theta_d (\text{District}_d\times \text{Post2019}_t) + \varepsilon_{dt}
\] $$
### B3. Offshore moderation (H5) ### B3. Offshore moderation (H5)
\[ $$
Y_{dt}=\alpha_d + \gamma_t + \sum_d \theta_d (\text{District}_d\times \text{Post2019}_t) + \phi(\text{Post2019}_t\times \text{Offshore}_d) + \varepsilon_{dt} Y_{dt}=\alpha_d + \gamma_t + \sum_d \theta_d (\text{District}_d\times \text{Post2019}_t) + \phi(\text{Post2019}_t\times \text{Offshore}_d) + \varepsilon_{dt}
\] $$
All models report district-clustered standard errors. All models report district-clustered standard errors.
@@ -124,18 +125,20 @@ These estimates indicate that offshore jurisdictions diverge from non-offshore d
| :--- | ---: | ---: | | :--- | ---: | ---: |
| Offshore-by-post-policy differential | 0.3819 | <0.001 | | Offshore-by-post-policy differential | 0.3819 | <0.001 |
See **Figure 4** in the main text (`district_treatment_effects_map_psj.png`) for the geographic distribution of district treatment effects. See Figure 4 in the main text (`district_treatment_effects_map_psj.png`) for the geographic distribution of district treatment effects.
Read alongside C3, this pooled interaction should be interpreted as an average offshore differential in the post period after district heterogeneity is already modeled, not as a claim that offshore status is the dominant driver of all district variation. Read alongside C3, this pooled interaction should be interpreted as an average offshore differential in the post period after district heterogeneity is already modeled, not as a claim that offshore status is the dominant driver of all district variation.
### C5. H3 moderator tests ### C5. H3 moderator tests
Main block: Main block:
- H3a Capacity: -0.0188 (p=0.9415) - H3a Capacity: -0.0188 (p=0.9415)
- H3b Baseline performance: -0.0884 (p=0.7144) - H3b Baseline performance: -0.0884 (p=0.7144)
- H3e Border proximity: -0.2768 (p=0.3082) - H3e Border proximity: -0.2768 (p=0.3082)
- H5 (same block estimate): 0.6317 (p=0.1055) - H5 (same block estimate): 0.6317 (p=0.1055)
Deep-dive block: Deep-dive block:
- H3c EJ: 0.1818 (p=0.4866) - H3c EJ: 0.1818 (p=0.4866)
- H3f Rurality: 0.2213 (p=0.4649) - H3f Rurality: 0.2213 (p=0.4649)
- H3e Border proximity: -0.3626 (p=0.1669) - H3e Border proximity: -0.3626 (p=0.1669)
@@ -199,7 +202,7 @@ Specification checks again point to the same empirical hierarchy: slope effects
## Appendix F. Interpretation Notes ## Appendix F. Interpretation Notes
1. The strongest system-wide evidence in this run is a **post-policy slope change**, not a one-time 2019 level shift. 1. The strongest system-wide evidence in this run is a post-policy slope change, not a one-time 2019 level shift.
2. District heterogeneity is substantial and statistically material. 2. District heterogeneity is substantial and statistically material.
3. Offshore jurisdiction contributes meaningfully in conditional models, but placebo behavior indicates caution in purely timing-based causal claims. 3. Offshore jurisdiction contributes meaningfully in conditional models, but placebo behavior indicates caution in purely timing-based causal claims.
4. Spatial diffusion is not supported by global autocorrelation tests. 4. Spatial diffusion is not supported by global autocorrelation tests.