A law of motion, if you respect breaks, slack, and the limits of prediction.

1. Croatia’s Okun story in one sentence, and the caveats hidden inside it

The report’s bottom line is deceptively simple: in Croatia’s annual data, output movements matter for unemployment, and the direction of influence runs more clearly from GDP to the labour market than the other way around, especially once you take structural shocks seriously. But the caveats are doing real work. The sample is short (2000–2024), the period is turbulent, and the results depend on whether you frame the relationship as year-to-year changes, long-run levels, or cyclical slack.

That combination, clear intuition, fragile inference, is precisely why Okun’s Law remains useful. It is not a mechanical rule that always delivers the same coefficient. It is a disciplined way to ask: when the economy runs hot or cold, how much of that temperature shows up in joblessness, and how quickly?

Croatia gives that question unusually sharp edges. Not because the relationship is absent, but because the economy’s recent history contains the kind of breaks that can make even a genuine relationship look inconsistent unless you model them as breaks rather than “noise”.

2. What the combined evidence actually says about Okun in Croatia

Across Parts I and II, the report builds its case in the intended sequence: visuals, then stationarity, then breaks, then cointegration, then dynamic modelling, then causality. The findings can be summarised without drowning in test statistics, because the report itself provides a clear “triangulated” conclusion.

Start with the foundations. Standard unit-root testing largely supports a familiar macro pattern: GDP per capita in logs (LGDP) and the unemployment rate (UR) behave like I(1) processes in levels, while differences behave more like stable, mean-reverting series. Once structural breaks are allowed for, the picture becomes more nuanced: the report notes evidence consistent with trend-stationarity for some variables when breaks are considered. The lesson is not “unit roots are everywhere.” It is that Croatia’s data are shaped by disruptive episodes, and a model that assumes one smooth regime is likely to misread the properties of the series.

Then the report asks the question that turns an Okun intuition into something policy-relevant: is there a long-run relationship tying output and unemployment together, or do they merely co-move episodically? On this, the report makes an important distinction. It concludes with high certainty that there is no long-run equilibrium relationship between the GDP growth rate and the unemployment rate in the research period, so the “growth-rate version” of Okun is not a stable long-run anchor in Croatia. At the same time, it finds stronger evidence for a long-run linkage in levels: Johansen cointegration identifies a long-run relationship between LGDP and UR (under a specification that includes intercept and trend), and ARDL bounds testing provides further support, particularly when unemployment is treated as the dependent variable. The Bayer–Hanck meta evidence reinforces the cointegration conclusion in the report’s narrative summary.

That combination is a subtle but important policy message. Croatia’s Okun relationship is not best understood as a long-run mapping from annual growth rates to annual unemployment rates. It is more credibly understood as a relationship that emerges in levels and in slack frameworks, and therefore as a mechanism that may require persistence to show up in the data.

The break evidence is the third pillar of the story. The report identifies structural shifts around 2009–2010 and 2019–2021, aligning with the global financial crisis and the pandemic era. That matters because it implies the relationship is not just a stable “law” operating uniformly; it is a channel whose strength can change when the economy is hit by shocks large enough to alter behaviour, institutions, or adjustment speed.

Finally, the causality results give the story a directional spine. Using the causality approach employed in the report (it labels it as a Toda–Yamamoto Granger-causality framework), the key finding is unidirectional: GDP helps predict unemployment, while unemployment does not predict GDP. The report explicitly notes the absence of reverse causality and draws a practical implication: unemployment is not an especially effective leading indicator for forecasting GDP in this setting; output-side indicators are more useful for anticipating labour-market movements. In differences and in gap series, the report does not detect strong causality, and it interprets that as consistent with Okun’s Law being a relationship that may be clearer in long-run or structural terms than in noisy short-run annual dynamics, especially in small samples.

If you want the report’s “meaning-first” synthesis in one line: Croatia’s Okun channel is real, but it is not a clean short-run forecasting tool; it is a medium-run adjustment mechanism that becomes visible once you respect trend, breaks, and the right variable framing.

3. Where policy transmission looks strongest, and where it leaks

The report’s results effectively separate two transmission stories—both compatible with Okun’s Law, but with different practical implications.

Transmission Story A: Output drives labour-market outcomes over the medium run.
This is the story supported by the cointegration evidence in levels (Johansen; ARDL bounds; Bayer–Hanck) and by the unidirectional causality from GDP to unemployment. In this story, Croatia’s labour market responds to output conditions, but not necessarily in the same year, and not necessarily with a stable growth-rate “rule of thumb.” Policy implication: if you want durable reductions in unemployment, output support must be sustained and structured; one good year is not a cure.

Transmission Story B: Cyclical slack is the cleanest policy lens, but short-run dynamics are noisy.
The report repeatedly emphasises the usefulness of the gap model for interpreting cyclical conditions: output gap and unemployment gap translate the Okun idea into slack language. Yet causality in gaps is weak in the report’s tests, and the short-run (differenced) causality is also weak. This is not a contradiction; it is an empirical reality in a small annual sample. Cyclical co-movement can be strong enough to matter and still not strong enough to pass strict short-run predictability tests when the sample is short and shocks are large.

So where does it “leak”? The report flags three main leak points, each with a policy analogue:

1. Breaks distort averages. If the labour market responds differently across crisis regimes and recovery regimes, a single “average Okun coefficient” is not a stable policy constant.
2. Short-run annual noise hides mechanisms. The absence of short-run causality in differences is consistent with the idea that annual changes are dominated by idiosyncrasies, measurement, or lag structure.
3. Not all unemployment is cyclical. The report explicitly warns that mixed evidence in levels and the strong role of breaks imply unemployment is not purely a cyclical mirror; structural components matter, and policy must address them directly.

4. EU and euro context, carefully and briefly

Croatia’s institutional context matters mostly as framing, not as evidence. As a relatively recent EU member and now a euro-area country, Croatia operates under a more integrated macro-policy environment than in the early part of the sample, while still managing domestic labour-market adjustment challenges shaped by shocks and structural change. The report’s results suggest that alignment and integration do not eliminate Okun’s mechanism, but they also do not guarantee that growth translates into jobs smoothly without complementary labour-market policies, especially after disruptive episodes.

5. Practical takeaways

The report’s policy implications read less like ideology and more like an operational checklist: if output predicts unemployment, growth policy matters; if breaks matter, resilience matters; if short-run predictability is weak, patience and structure matter. Distilled into the tightest “key results” language consistent with the report, the takeaways are:

• Output-led labour improvement is the dominant direction. The report’s causality evidence supports a GDP → unemployment direction, not the reverse.
• Breaks are not a nuisance; they are part of the mechanism. The crisis periods (2009–2010; 2019–2021) coincide with structural shifts that change how the relationship looks and how it should be modelled.
• The growth-rate Okun story is not a long-run anchor. The report rejects a stable long-run relationship between GDP growth rates and unemployment rates; the more credible long-run linkage is in levels and through dynamic frameworks.
• Slack framing is the most policy-intuitive lens, but not a short-run forecasting machine. The gap model clarifies cyclical alignment, yet strict short-run causality in gaps is not strongly detected in this sample.

6. What to watch next

If Part III is meant to leave readers with “what should we watch for,” the study points to a few high-value signals, not in the sense of forecasting gimmicks, but in the sense of where policy effectiveness is most likely to be gained or lost.

• Watch whether post-shock recoveries translate into broad labour absorption or narrow output rebounds. If unemployment adjustment is slower than output recovery, slack can persist even when GDP looks healthy.
• Watch whether the next downturn creates a new break, or merely a temporary deviation. The report’s break findings imply that some shocks reshape the mechanism itself.
• Watch the structural component of unemployment, not just the cyclical one. The report’s interpretation explicitly cautions that not all unemployment in Croatia is cyclical; long-run instruments (skills, matching, flexibility) matter alongside stabilisation.

And the most practical meta-lesson is the report’s own methodological warning turned into policy language: do not overlearn from any single model in a short sample. Triangulation is not a stylistic preference here; it is the only responsible approach.

References

Bai, J., & Perron, P. (1998). Estimating and testing linear models with multiple structural changes. Econometrica, 66(1), 47–78. DOI: https://doi.org/10.2307/2998540

Bayer, C., & Hanck, C. (2013). Combining non-cointegration tests. Journal of Time Series Analysis, 34(1), 83–95. DOI: https://doi.org/10.1111/j.1467-9892.2012.00814.x

Clemente, J., Montañés, A.,& Reyes, M. (1998). Testing for a unit root in variables with a double change in the mean. Economics Letters, 59(2), 175–182. DOI: https://doi.org/10.1016/S0165-1765(98)00052-4

Engle, R. F., & Granger, C. W. J. (1987). Co-integration and error correction: Representation, estimation, and testing. Econometrica, 55(2), 251–276. DOI: https://doi.org/10.2307/1913236

Gregory, A. W., & Hansen, B. E. (1996). Residual-based tests for cointegration in models with regime shifts. Journal of Econometrics, 70(1), 99–126. DOI: https://doi.org/10.1016/0304-4076(69)41685-7

Hodrick, R. J., & Prescott, E. C. (1997). Postwar U.S. business cycles: An empirical investigation. Journal of Money, Credit, and Banking, 29(1), 1–16. DOI: https://doi.org/10.2307/2953682

Johansen, S. (1991). Estimation and hypothesis testing of cointegration vectors in Gaussian vector autoregressive models. Econometrica, 59(6), 1551–1580. DOI: https://doi.org/10.2307/2938278

Okun, A. M. (1962). Potential GNP: Its measurement and significance. American Statistical Association, Proceedings of the Business and Economic Statistics Section, 98–104.

Pesaran, M. H., Shin, Y., & Smith, R. J. (2001). Bounds testing approaches to the analysis of level relationships. Journal of Applied Econometrics, 16(3), 289–326. DOI: https://doi.org/10.1002/jae.616

Shin, Y., Yu, B., & Greenwood-Nimmo, M. (2014). Modelling asymmetric cointegration and dynamic multipliers in a nonlinear ARDL framework. In Festschrift in Honor of Peter Schmidt (pp. 281–314). Springer. DOI: https://doi.org/10.1007/978-1-4899-8008-3_9

Zivot, E., & Andrews, D. W. K. (1992). Further evidence on the great crash, the oil-price shock, and the unit-root hypothesis. Journal of Business & Economic Statistics, 10(3), 251–270. DOI: https://doi.org/10.1080/07350015.1992.10509904