Dating Bosnia and Herzegovina’s business cycle

Part 3 shifts from forecasting the level of GDP to identifying the business cycle. The business cycle is the movement of economic activity around its longer-term trend. This is different from simply asking whether GDP grew or fell in a given quarter. An economy can grow and still be below trend, or slow down while remaining above trend. For that reason, business-cycle dating based on filters gives a more nuanced view of macroeconomic phases.

This analysis compares three filters: Hodrick-Prescott, Baxter-King and Corbae-Ouliaris. Each separates trend and cycle differently. The aim is not to treat any one filter as the absolute truth, but to see which episodes are robust across methods.

The Hodrick-Prescott filter shows a long and readable cycle over the full sample. It classifies Bosnia and Herzegovina as being in contraction for 58 quarters and in expansion for 46 quarters. The early 2000s are mostly below trend, followed by a stronger expansion from 2006Q2 to 2008Q4. The global financial crisis then produces a long below-trend phase from 2009Q1 to 2012Q4. Later cycles are shorter and more frequent. The period from 2018Q1 to 2019Q3 is above trend, followed by the pandemic contraction from 2019Q4 to 2020Q2. The recovery from 2020Q3 to 2022Q2 is above trend, while the latest period from 2022Q3 to 2025Q4 is classified as below trend.

The Baxter-King filter gives a smoother picture but covers a shorter effective sample because of endpoint losses. It identifies 46 expansion quarters and 34 contraction quarters. The key phases are broadly similar: expansion before the global financial crisis, contraction during and after it, another above-trend phase from 2016Q2 to 2019Q2, contraction during the pandemic and a post-pandemic rebound from 2020Q4 to 2021Q4. Since the Baxter-King filter is designed to isolate business-cycle frequencies, it is less sensitive to short fluctuations, but less useful for the latest observations.

The Corbae-Ouliaris filter gives a full-sample classification with 53 expansion quarters and 51 contraction quarters. It agrees with the other filters on the main episodes: pre-crisis expansion, global-financial-crisis contraction, pandemic contraction and post-pandemic recovery. It identifies contraction from 2019Q3 to 2020Q3, expansion from 2020Q4 to 2021Q4, and then a sequence of shorter phases after 2022. Its latest classification is contraction from 2024Q4 to 2025Q4.

The three filters therefore agree on the major story, even when they differ on exact turning points. The clearest common episodes are the mid-2000s expansion, the 2008–2009 crisis, the pandemic contraction and the post-pandemic rebound. The main difference is the treatment of the final years. HP suggests a more persistent below-trend phase after 2022Q3, while Corbae-Ouliaris shows some short above-trend intervals before returning to contraction. Baxter-King is less informative at the end because of its endpoint limitation.

Table 4. Business cycles dating for Bosnia and Herzegovina time series

For interpretation, the HP filter is the most convenient main reference because it covers the whole sample and is easy to explain. However, it should be checked against Baxter-King and Corbae-Ouliaris. Where all three filters agree, the evidence is strong. Where they differ, it is better to speak of a turning-point zone rather than a precise quarter. The cautious reading is that Bosnia and Herzegovina recovered strongly after the pandemic, but the most recent phase is closer to moderation around or below trend than to renewed acceleration.

Methodological appendix to Part 3

Business-cycle dating in this analysis starts by separating GDP into a trend component and a cyclical component. If the cyclical component is above zero, GDP is interpreted as being above trend. If it is below zero, GDP is interpreted as being below trend. The resulting positive and negative phases are then treated as expansions and contractions.

The Hodrick-Prescott filter estimates a smooth trend and treats deviations from that trend as the cycle. The Baxter-King filter is a band-pass filter designed to isolate movements at business-cycle frequencies. The Corbae-Ouliaris filter is another method for extracting cycles from non-stationary data. None of these filters observes the true cycle directly. Each imposes assumptions. For that reason, robust interpretation should focus on episodes that appear across several filters, not on isolated turning points produced by one method only.