Long-horizon projection models assume stable growth-determinant relationships, but the 2008 crisis, the strongest post-war shock to rich economies, provides a test case for this assumption. This thesis examines the impact of the 2008 crisis on the relationship between growth determinants and output, using a sample of 24 OECD members from 1996 to 2019, based on the Penn World Table 11.0. A balanced panel with two-way fixed effects is employed, incorporating augmented-Solow regressors and break tests at 2008 and 2010, as well as a sup-Wald scan across years. Mean growth fell from 2.5 percent a year before the crisis to 0.6 percent after, with every country experiencing slower growth. Every projection model must account for this shift. Projections that assume a return to pre-crisis trends may overestimate future growth.
Figure 1
Growth fell after 2008
Mean per-capita growth across 24 OECD countries, 1996–2019 (shaded band: p25–p75). The panel's own post-2008 flag splits the average from 2.46% (1996–2007) to 0.64% (2008–2019).
Figure 2
Every country slower
Pre- vs. post-2008 mean growth, one point per country. Every one of the 24 countries sits below the y = x line — no country matched its pre-crisis pace afterward, and the country average fell from 2.5% pre-2008 to 0.6% post-2008. Labelled points are the three lowest post-2008 averages — Greece, Italy, Finland — and the two highest — Ireland, Türkiye.
Data
OECD members before 1995 are selected for the sample, which spans from 1996 to 2019, and is based on the Penn World Table 11.0. Growth, the dependent variable, is measured as annual growth in real GDP per head. Across the sample, the average growth rate is 1.6 percent, with a spread of 2.6 percentage points. Regressors include lagged log income, investment share, log human-capital index, and population growth, with averages of 25.6 percent of GDP for investment and 0.7 percent for population growth.
Method
A two-way fixed effects model is used, with country effects and year effects, and standard errors are clustered by country with a CR2 correction and Satterthwaite degrees of freedom. Known-date tests are performed for each regressor times a post-crisis dummy, with joint zero tests on the four interactions and candidate years 2008 and 2010. Where the break date is unknown, the sup-Wald scan is used, refitting the model at every year, with 15 percent trimming at both ends. Because the sample has only 24 clusters, the CR2 correction is necessary to obtain reliable standard errors.
Findings
Before 2008, the average growth rate was 2.5 percent a year, compared to 0.6 percent after. Two of the four regressors, convergence and investment, are significant within countries. Convergence is estimated to be minus 0.074, with a p-value of 0.009. Investment is estimated to be 0.163, with a p-value of 0.014, indicating that a ten-point increase in investment is associated with approximately 1.6 percentage points more growth. At 2008, the interactions are tested, and none are significant alone, although the investment interaction is 0.108, with a p-value of 0.075. No break test rejects stability, with joint F-statistics of 1.30 and 1.03 for 2008 and 2010, respectively.
Figure 3
No candidate year rejects stability
Wald statistic (4 break-interaction restrictions) for a break at each candidate year. The maximum, at 2001, is 11.1 — below the Andrews (1993) 10% critical value of 14.3. (Per-year p-values aren’t valid for a max-selected statistic; compare to the critical values shown, not a chi-squared reference.)
Figure 4
Only investment moves, and not far enough
Post-2008 shift in each coefficient, in standard-error units (two-way FE panel, CR2 cluster-robust SEs, 24 clusters). The shaded band (±2.07 SE) marks shifts not distinguishable from zero. Investment share shifts most (0.108, 1.96 of 2.07) — nearest the edge. Log GDP per capita, lagged (−1.42 of 2.07), population growth (1.37 of 2.07), and log human capital (1.11 of 2.07) stay further inside.
Figure 5
Same slopes, nearly the same path
Illustrative projection from the full-sample slopes, 2019–2045, indexed to 2019 = 100, with other regressors frozen at their 2015–2019 country means. The pre- and post-crisis common components produce almost the same path — not a forecast.
Robustness
Dropping Ireland from the sample results in a catch-up coefficient of minus 0.069, with a p-value of 0.085, and an investment coefficient of 0.141, with a p-value of 0.060. Replacing population growth with employment growth yields a within R-squared of 0.091 to 0.275, and a coefficient of 0.635, with a p-value below 0.001. Lagging investment by one year results in a coefficient of 0.057, with a p-value of 0.34. Difference GMM estimation yields a Sargan test p-value of 0.93, and an AR(2) test p-value of 0.53, although the catch-up coefficient is imprecise.
Limitations
Power calculations indicate that the test has low power to detect shifts in the investment coefficient, with a threshold of 0.169, and a standard error of 0.055. With only 24 clusters, the test has limited ability to detect breaks. Failure to reject stability does not prove that the relationships are stable. Schooling and population growth have little within-country movement year to year, making it difficult to detect significant effects. Investment contemporaneousness may be subject to reverse causation, although the lagged result confirms the baseline finding. GMM estimation yields valid diagnostics, but the results are too imprecise to be reliable.
Version 1.0 — submitted version, June 2026; a revised version will replace this file.
Read the full paper (PDF, opens in new tab)
Get new essays and research by email. Subscribe on Substack (opens in new tab)