Pre-registration
Across cross-country liberalisation episodes since 1971 (trade opening, product-market deregulation, state-owned-enterprise divestiture, financial account liberalisation), output growth follows the reform rather than preceding it — i.e. the observed growth pickup is policy-caused and not driven by reverse causation where successful economies reform because they can afford to. Tested via event-study across liberalisation-episode panel with pre-trend checks.
Falsification criterion — what would disprove this
This hypothesis is considered falsified if:
PRIMARY (dispositive): the hypothesis is SUPPORTED if event-study coefficients on real GDP-per-capita log-growth (excess over the country sample-mean baseline) are each at least +0.5pp/yr AND statistically positive at p<0.10 (country-block bootstrap) at horizons h=3, h=5, AND h=10, AND the pre-trend coefficients at h=-3 and h=-5 are NOT significantly positive (i.e. either p>=0.10 against the zero null or magnitude < +0.5pp). REFUTED if any post-reform coefficient at h=3,5,10 fails the magnitude OR the significance test, OR if either pre-trend at h=-3,-5 is significantly positive (the reverse-causation indicator). INFORMATIVE: per-horizon coefficients, SEs, and number of cells in diagnostics.json. METHOD_VALID: at least 30 EFW-jump episodes detected across the 41-country sample.
formal test & threshold
test: liberalisation_episode_event_study_dual_threshold
threshold: PRIMARY: coef[h] >= +0.005 AND p[h] < 0.10 for h in {3,5,10}; AND not (coef[h] >= +0.005 AND p[h] < 0.10) for h in {-3,-5}.Method
- Template
event_study- Fixed effects
country, year- Clustering
country- Sample
- 41 countries · 1971 – 2019
- Evidence type
- associational
Stacked event-study across liberalisation-episode panel since 1971. Outcome: real GDP per capita growth at horizons h ∈ {-5,-3,0,3,5,10}. Pre-trend tests at h=-5,-3 distinguish policy causation from reverse causation. Episodes coded from Sachs-Warner trade-openness, IMF capital-account, and product-market deregulation indicators.
Data
| Variable | Source | Transform |
|---|---|---|
log_gdp_pc_ppp outcome | world_bank_wdi:NY.GDP.PCAP.PP.KDtier 2 | log |
real_gdp_growth outcome | world_bank_wdi:NY.GDP.MKTP.KD.ZGtier 2 | level |
liberalisation_episode_indicator treatment | fraser_efw:aggregate_scoretier 4 | indicator |
trade_openness treatment | world_bank_wdi:NE.TRD.GNFS.ZStier 2 | level |
log_initial_gdp_pc control | world_bank_wdi:NY.GDP.PCAP.KDtier 2 | log |
log_population control | world_bank_wdi:SP.POP.TOTLtier 2 | log |
rule_of_law control | wgi:RL.ESTtier 4 | level |
gross_capital_formation_pct_gdp control | world_bank_wdi:NE.GDI.TOTL.ZStier 2 | level |
● ready · ● pending · ● reconstruct-needed
Detailed result card
Liberalisation episodes and growth trajectory
Verdict: refuted — Both the post-reform growth pickup and the clean-pre-trend conditions fail. h=3 -0.47pp (p=0.30); h=5 +1.99pp (p=0.17); h=10 +0.34pp (p=0.68). Pre-trends: h=-3 -1.28pp; h=-5 +5.62pp.
Summary
- Episodes detected: 31 across 31 countries (year of largest 5-year jump in Fraser EFW aggregate score per country, conditional on jump ≥ +0.5 points).
- Method validity: OK — event study estimates GDP-per-capita log-growth deviation from country mean at horizons [-5, -3, 0, 3, 5, 10].
Event-study coefficients (excess log-growth over country baseline)
| Horizon h | Coef | SE | p | n cells | |---:|---:|---:|---:|---:| | -5 | +5.62pp | 2.29pp | 0.004 | 31 | | -3 | -1.28pp | 0.58pp | 0.028 | 31 | | +0 | +4.39pp | 1.72pp | 0.000 | 31 | | +3 | -0.47pp | 0.47pp | 0.300 | 31 | | +5 | +1.99pp | 1.56pp | 0.170 | 31 | | +10 | +0.34pp | 0.66pp | 0.676 | 29 |
Method
- Treatment: liberalisation episode at the country-year of the LARGEST 5-year jump in Fraser EFW aggregate score (
efw_summary) per country, conditional on that jump being ≥ +0.5 points. One episode per country (the EFW panel is 5-year-spaced 1970-2000 and annual after, so 5 yr is the natural unit; one-per-country prevents stacked-event-study double-counting of the same multi-decade reform wave). - Outcome: real GDP-per-capita log-growth (year-on-year), preferring NY.GDP.PCAP.PP.KD (PPP) and falling back to NY.GDP.PCAP.KD where PPP is missing. 2020 dropped (COVID).
- Event study: per (country, episode, h), average growth in years e+h ± 1, demean by country sample mean, then average across (country, episode) cells. Country-block bootstrap (1000 reps) for SE / two-sided p-values.
- Falsification rule (sharpened): SUPPORTED requires h ∈ {3,5,10} all ≥ +0.5pp/yr AND p<0.1; AND pre-trend at h ∈ {-3,-5} not significantly positive. REFUTED if either condition fails.
Data
- fraser_efw:efw_panel (Economic Freedom of the World, aggregate score)
- world_bank_wdi:NY.GDP.PCAP.PP.KD (PPP per-capita GDP)
- world_bank_wdi:NY.GDP.PCAP.KD (constant-USD per-capita GDP, fallback)
Caveats
- Episode catalogue is derived from EFW jumps, which is an indicator-based (not narrative) coding. Spec mentions Sachs-Warner and Wacziarg-Welch dating; those would refine the catalogue but are not on disk.
- Event horizons constrained by EFW coverage (most countries from 1980; pre-1980 episodes therefore underweighted).
- Country baseline-demeaning absorbs slow-moving country-fixed differences; richer two-way FE would need iterative within-estimation. Block-bootstrap by country mimics country-clustered SEs.
Notes
Seeded from a classical-liberal claim that growth follows liberalisation rather than precedes it. Stacked event study; episode catalogue (inclusion criteria for what counts as a liberalisation episode) is the central judgement call requiring human review before pre-registration.