ITFP Growth, The Balassa-Samuelson Hypothesis, and Competitiveness in the Baltics
Author:
International Monetary Fund. European Dept.
Search for other papers by International Monetary Fund. European Dept. in
Current site
Google Scholar
Close

Kalman filter-based estimates suggest that potential GDP growth in Estonia has declined steadily since the Global Financial Crisis. To some extent, this is a common trend for the Baltic region. However, in Estonia this adjustment has been primarily driven by a fall in the growth of Total Factor Productivity. Falling TFP growth, combined with an appreciated real exchange rate, has likely reduced Estonia’s ability to absorb recent shocks, taking a toll on its external performance and exacerbating its current economic downturn compared to other countries in the region.

ITFP Growth, The Balassa-Samuelson Hypothesis, and Competitiveness in the Baltics1

Kalman filter-based estimates suggest that potential GDP growth in Estonia has declined steadily since the Global Financial Crisis. To some extent, this is a common trend for the Baltic region. However, in Estonia this adjustment has been primarily driven by a fall in the growth of Total Factor Productivity. Falling TFP growth, combined with an appreciated real exchange rate, has likely reduced Estonia’s ability to absorb recent shocks, taking a toll on its external performance and exacerbating its current economic downturn compared to other countries in the region.

A. Introduction

1. Since regaining independence in the early nineties, Estonia experienced fast economic growth and convergence of per capita income towards more advanced countries in Europe. During the process, the real effective exchange rate (REER) was expected to appreciate reflecting faster total factor productivity (TFP) growth compared to trading partners (the Balassa-Samuelson effect).2 A corollary of this proposition is that lagging productivity growth combined with continued REER appreciation can lead to loss of competitiveness.

2. This Selected Issues Paper (SIP) provides estimates of TFP, assesses the Balassa-Samuelson hypothesis, and constructs a TFP-consistent REER for Estonia, which can be used as a benchmark for competitiveness. Estimates of TFP are obtained from a standard Cobb-Douglas production function applied to quarterly data of real GDP, labor, and capital inputs over the period from 1995Q1 to 2023Q2. The estimated TFP series is used to assess the Balassa-Samuelson hypothesis in a cointegrating equation with the REER.

3. In the process, the SIP also provides estimates of the factors driving both actual and potential GDP. Signals from high-frequency indicators of economic slackness—confidence indices, industrial production, unemployment rate, and capacity utilization—are used in a multivariate Kalman filter to identify the business cycle and help estimate potential GDP. The same production function used to estimate TFP is applied to the filtered series (i.e., smooth trends) of the labor and capital inputs to decompose potential GDP. A structural (i.e., trend) TFP is estimated in the process. The decompositions of both actual and potential GDP distinguish between cyclical and “structural” (i.e., low frequency) drivers of real GDP over the sample period.

4. The same methodology is extended to the other two Baltic countries, aiming at answering the following research questions. How has potential GDP evolved over the last three decades in the Baltics? What are the roles of the main production factors and TFP in driving the dynamics of actual and potential GDP? Can the secular REER appreciation observed in the Baltics be fully accounted for by the TFP dynamics? Are there distinct paths of TFP and potential GDP growth among the three Baltic countries? Moreover, focusing on Estonia, can we disentangle the role played by both structural and cyclical factors during the recent economic downturn?

B. Methodology

5. Step 1: Impose a standard Cobb-Douglas production function on quarterly data to estimate a series for TFP. In particular, the TFP series is obtained as a residual from:

In at = In yt (1 — a) ln kt — a In lt

where at,yt,kt and lt are the levels of TFP, real GDP, the stock capital, and the labor input, respectively. Components yt and kt are both measured in millions of constant 2015 euros, while lt is measured in thousands of hours-worked per quarter. The stock of capital was obtained by applying quarterly investment flows to data of the (annual) capital stock from the European Commission’s Annual Macro-Economic Database (AMECO) and estimates of its depreciation rate. The resulting series (Figure 1) was then multiplied by a measure of industrial capacity utilization to produce an estimate of the effective capital stock, kt. The labor input was constructed by multiplying the number of employees (et) by the average number of hours worked per employee (ht). For the labor share, αt, a smooth trend of the ratio of compensation of employes to GDP (i.e., the labor share) was used.3

Figure 1.
Figure 1.

Estimates of TFP

(In logarithmic form)

Citation: IMF Staff Country Reports 2024, 178; 10.5089/9798400278334.002.A002

Sources: Statistics Estonia; Eurostat; Haver Analytics; and IMF staff calculations.
Figure 2.
Figure 2.

Estimates of Structural TFP

(In logarithmic form)

Citation: IMF Staff Country Reports 2024, 178; 10.5089/9798400278334.002.A002

Sources: Statistics Estonia; Eurostat; Haver Analytics; and IMF staff calculations.

6. Step 2: Apply a multivariate Kalman filter on quarterly real GDP data and indicators of economic slack to estimate potential GDP. The TFP series obtained in step 1 captures the effect of both cyclical (reflecting, for example, short-term developments such as labor hoarding, short-term skills mismatch etc.) and structural (i.e., low-frequency effects of institutions, business environment and practices, education, R&D etc.) elements that are not explained by labor and capital. To isolate the structural component of TFP, an estimate of potential GDP (y¯t) was independently obtained from a state-space decomposition of cycle and trend in GDP using monthly confidence indices (i.e., consumer, industry, construction, and retail sector), the unemployment rate, and industrial capacity utilization as signal variables to help pin down the cyclical components.4 A trend TFP series (Figure 2), āt, was obtained from the production function equation applied to HP-filter trends for effective capital (k¯t), employment (ēt), and hours-worked (h¯t). For comparison, Figure 2 also displays the HP-trend of the estimated TFP series.

7. Step 3: Estimate a cointegration relationship between the TFP and the REER to assess the Balassa-Samuelson Hypothesis. After confirming that both series are integrated,5 Johansen cointegration tests found at least one cointegrating relationship between the two variables according to different specifications regarding exogenous regressors in the cointegration vector and/or short-term dynamic equations (Table 1). This result, confirmed for all three Baltic countries, is consistent with the Balassa-Samuelson hypothesis, indicating a long-run relationship between TFP and REER. It also indicates that a cointegration relationship between TFP and the REER can be estimated (Table 2), and the fitted values can be used to construct a measure of the TFP-based REER as the implied long-term relationship between the two series. The comparison between actual and TFP-based REER can indicate how large deviations from the Balassa-Samuelson hypothesis are. Such deviations have implications for competitiveness. For instance, negative (positive) gaps between actual and TFP-based REER indicate a price-competitiveness advantage (disadvantage).

Table 1.

Estonia: Johansen Cointegration Tests

article image
Note: Rank selected at 0.05 level using critical values from MacKinnon-Haug-Michelis (1999) Remarks: Case 7: No deterministic terms; Case 2: Cointegrating relationship includes a constant; Case 3 (Johansen-Hendry-Juselius): Cointegrating relationship includes a constant. Short-run dynamics include a constant; Case 3: Short-run dynamics include a constant; Case 4 (Johansen-Hendry-Juselius): Cointegrating relationship includes a constant and trend. Short-run dynamics include a constant; Case 4: Cointegrating relationship includes a trend. Short-run dynamics include a constant; Case 5 (Johansen-Hendry-Juselius): Both the cointegrating relationship and short-run dynamics include a constant and trend; Case 5: Short-run dynamics include a constant and trend. Source: IMF staff estimates
Table 2.

Estonia: Cointegration Between REER and TFP

article image
Source: IMF staff estimates

C. Results

Potential GDP Growth Has Declined Since the GFC, Driven by a Fall in TFP Growth

8. Estimates of potential GDP levels point to scarring effects. Simple extrapolations in Figure 3 suggest significant output losses following the GFC and, to a lesser extent, the pandemic. The latter may be related to supply chain disruptions and labor hysteresis which started with the pandemic and were compounded by Russia’s war on Ukraine.

9. Potential GDP growth has declined steadily since the GFC. Figure 4 shows that potential growth fell from above 5 percent pre-GFC period to around 2 percent post-GFC. Actual growth has dropped even lower after 2020 and well below potential growth, suggesting also a significant cyclical component in the current downturn.

Figure 3.
Figure 3.

GDP

(Mil. Euros)

Citation: IMF Staff Country Reports 2024, 178; 10.5089/9798400278334.002.A002

Source: Statistics Estonia; and IMF staff calculations.

10. Capital has been the major driver of GDP growth in Estonia. Figure 4 also shows the contributions to GDP of TFP, capital, labor, and the change in the time-varying labor share for the three different periods. After acting as a drag on GDP growth in the pre-GFC period, labor has supported growth in the following periods. However, the contribution of labor has remained less significant than that of capital even in recent years, despite the large migration flows following Russia’s war on Ukraine.

11. The fall in TFP growth accounts for most of the decline in GDP growth. After providing a significant contribution to GDP growth in the pre-GFC period, TFP growth faded post GFC and has turned negative more recently. When comparing the years after 2020 with the pre-GFC period, the decline in average TFP growth surpasses that of actual GDP growth and explains almost 90 percent of the fall in estimated potential GDP growth (Table 3). The fall in TFP growth is also significant relative to the post-GFC period, and explains the current downturn, while the contribution of both capital and labor to GDP growth has increased (Table 4).

Figure 4.
Figure 4.

Decomposition of Actual and Potential GDP Growth

(Percentage points per year)

Citation: IMF Staff Country Reports 2024, 178; 10.5089/9798400278334.002.A002

Source: Statistics Estonia; Eurostat; Haver Analytics; and IMF staff calculations.

12. Capital and labor have dampened the fall in growth in the recent downturn. Table 4 offers a more granular breakdown of the various components, including the specific drivers of capital and labor, shedding more light on the current downturn. Capital accumulation and capacity utilization have both cushioned the decline in GDP growth after 2020, although capacity utilization remains on a downward trend. While demographics, labor participation and hours-worked have all contributed positively to GDP growth recently, the structural contribution of the unemployment rate has turned negative, suggesting an increase in the natural rate of unemployment.

Table 3.

Estonia: Changes in GDP Growth Rates Between 1995–2008 and 2020–2023Q2

(Percentage points)

article image
Source: IMF staff estimates
Table 4.

Estonia: Decomposition of GDP Growth

(Percentage points)

article image
Source: IMF staff estimates Note: growth rates calculated as difference in natural logarithms of original series.

Declining TFP Growth Has Compounded with Real Exchange Rate Appreciation in the Current Downturn

13. Fast TFP growth pre- GFC gave Estonia a competitive edge. Figure 5 shows that before the GFC actual REER was below the TFP-based REER (i.e., negative REER gaps) most of the time, indicating that the REER was undervalued relative to the (counterfactual) TFP-based equilibrium REER, providing Estonia with a price-competitiveness edge.

Figure 5.
Figure 5.

REER and TFP-Based REER

(LHS – in logarithmic form; RHS – in percent)

Citation: IMF Staff Country Reports 2024, 178; 10.5089/9798400278334.002.A002

Source: Statistics Estonia; Eurostat; Haver Analytics; and IMF staff calculations.

14. But Estonia’s competitive advantage was significantly eroded after the GFC. After 2008, periods of negative REER gaps gave way to periods of positive gaps. Negative REER gaps became significantly smaller and shorter-lived, while the positive gaps were larger and more frequent relative to the previous period (Figure 5), indicating some erosion of competitiveness during a phase of declining GDP growth.

15. The current downturn has coincided with both sharp real exchange rate appreciation and a decline in TFP, exacerbating existing problems. A significant divergence between actual and TFP-based REER has started in late 2021, at the onset of the current downturn, as steady TFP deceleration turned into a decline and compounded with real exchange rate appreciation. The overvalued exchange rate has likely reduced Estonia’s ability to absorb recent shocks, leading to a decline in net exports and economic activity.

Figure 6.
Figure 6.

Contribution to GDP Growth in the Baltics 1/

(Percentage points per year)

Citation: IMF Staff Country Reports 2024, 178; 10.5089/9798400278334.002.A002

Source: IMF staff calculations.1/ Growth rates calculated as difference in natural logarithims of original series.2/ Latvia: 200202–2008; Lithuania: 1998Q3–2008

The Interplay Between TFP Growth and the Real Exchange Rate May Explain Some Regional Differences

16. Capital accumulation drove the Baltics’ convergence process. Figure 6 shows the factors driving GDP growth across the three Baltics. Prior to the GFC, all three countries experienced rapid GDP growth, mainly driven by capital accumulation. This is typical of a period of capital deepening and fast income convergence.

17. After 2008, TFP growth diverged across the Baltics. Post-GFC the capital contribution to GDP growth has declined, especially for Latvia. In contrast, our analysis suggests that TFP has increasingly become a driver of growth for Latvia and especially Lithuania. Estonia has been the exception with declining and eventually negative productivity growth, as discussed in the previous section. Differences in TFP dynamics may be at the root of Estonia’s underperformance after 2021.

18. Diverging cyclical and structural forces are also behind recent regional developments. A more granular decomposition of the drivers of GDP growth between structural and cyclical components offers shed more light on recent developments in the region (Figure 7). While cyclical components appear to be playing an important role for Latvia and Estonia—in this case especially TFP—Lithuania’s expansion is largely driven by a structural contribution of TFP and, to a lesser extent, capital.

19. Similar to Estonia, Latvia appears to have had a competitive edge that faded post-GFC. Figure 8 shows actual and estimated TFP-based REER for the Baltics. Prior to the GFC, high TFP relative to REER explained consistently negative REER gaps for Estonia and Latvia. In Estonia, negative REER gaps were largely driven by high TFP, whereas in Latvia a depreciating real exchange rate was the main factor. In Lithuania, the gap turned negative only in 2005–2007 and to a smaller extent As a result, Latvia and Estonia likely experienced competitiveness advantages during this period, while Lithuania did not. However, post-GFC this advantage faded. Declining productivity growth in the case of Estonia and real exchange rate appreciation for Latvia drove REER gaps into positive territory.

Figure 7.
Figure 7.

Contribution to GDP Growth in the Baltics, 1995–2023

Citation: IMF Staff Country Reports 2024, 178; 10.5089/9798400278334.002.A002

Source: Statistical Authorities; Eurostat; Haver Analytics; and IMF staff calculations.Note: Growth rates are the difference in natural logarithms. Latvia: 2022Q2–2023Q2; Lithuania: 1998Q3–2023Q2
Figure 8.
Figure 8.

Actual and TFP-Based REER in the Baltics, 1995–20231

(LHS – natural logs; RHS – percent)

Citation: IMF Staff Country Reports 2024, 178; 10.5089/9798400278334.002.A002

Source: IMF staff estimates1 Latvia: 2002Q2–2023Q2; Lithuania: 1998Q3–2023Q2.

20. Estonia’s TFP-based competitive disadvantage observed more recently started earlier and became stronger than in Latvia and Lithuania. Figure 8 shows a decoupling between actual and TFP-based REER in recent years for all three Baltic countries. The REER gap is larger for Estonia, which is consistent with the country’s underperformance. Figure 9 zooms in the REER gaps during this period. The competitiveness disadvantage has started earlier for Estonia, and it has been more pronounced.

Figure 9.
Figure 9.

REER Gaps in the Baltics

(Percent)

Citation: IMF Staff Country Reports 2024, 178; 10.5089/9798400278334.002.A002

Source: Statistical Authorities; Haver Analytics; and IMF staff calculations.

D. Conclusions

21. Potential GDP growth in Estonia has fallen since the GFC, largely due to a steady decline in TFP growth. Some decline in potential GDP growth typically accompanies the process of income convergence, as capital accumulation decelerates. However, the largest contributor to the reduction in potential GDP growth for Estonia has been a decline in TFP growth. And the drop in TFP growth has been significantly more pronounced than in Latvia and, especially, Lithuania, which appears to experience an acceleration in TFP growth in recent years.

22. Differently from Latvia and Lithuania, the level of TFP has declined in Estonia since 2020. Not only has TFP growth declined in Estonia since the GFC, but it has become negative more recently. The decline in TFP growth has a structural component, which is probably associated with the scarring effects of recent shocks.

23. Differences in TFP dynamics across the Baltic countries have implications for competitiveness. Pre-GFC, fast TFP growth underpinned Estonia’s competitive advantage, despite real exchange rate appreciation. Post-GFC, decelerating TFP growth has eroded Estonia’s competitive advantage. More recently, significant real exchange rate appreciation has compounded the effect of declining TFP, turning into a competitive disadvantage and left the country more vulnerable to recent shocks. Loss of competitiveness may be a factor in Estonia’s current, more severe economic downturn relative to other Baltic countries. In Estonia, the (positive) wedge between the actual and TFP-based REER started earlier, evolved faster, and became wider than in the other Baltics, reducing the country’s external competitiveness by a much larger factor than in Latvia and Lithuania.

References

  • Balassa, B. (1964). “The Purchasing-Power Parity Doctrine: A Reappraisal.” Journal of Political Economy, 72 (6), 584596.

  • Blagrave, P., Garcia-Saltos, R., Laxton, D., and Zhang, F. (2015). “A Simple Multivariate Filter for Estimating Potential Output.” IMF Working Paper WP/15/79 https://www.imf.orq/external/pubs/ft/wp/2015/wp1579.pdf

    • Search Google Scholar
    • Export Citation
  • Benes, J., and N’Diaye, P. (2004). “A Multivariate Filter for Measuring Potential Output and the NAIRU: Application to The Czech Republic.” IMF Working Paper WP/04/45.

    • Search Google Scholar
    • Export Citation
  • Durbin, J., and Koopman, S. J. (2001). Time Series Analysis by State Space Methods. Oxford University Press.

  • Hamilton, J. D. (1994). Time Series Analysis. Princeton University Press.

  • Ito, T., Isard, P., and Symansky, S. (1997). “Economic Growth and Real Exchange Rate: An Overview of the Balassa-Samuelson Hypothesis in Asia.” NBER Chapters, in: Changes in Exchange Rates in Rapidly Developing Countries.

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund (2006). “Methodology for Estimating Potential Output and the Output Gap,” in World Economic Outlook: Globalization and Inflation.

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund (2017). Multivariate Filter Estimation of Potential Output for the United States. IMF Working Paper WP/17/106 https://www.imf.orq/-/media/Files/Publications/WP/2017/wp17106.ashx

    • Search Google Scholar
    • Export Citation
  • Kuttner, K. N. (1994). “Estimating Potential Output as a Latent Variable.” World Bank Policy Research Working Paper 1293.

  • Samuelson, P. A. (1964). “Theoretical Notes on Trade Problems.” The Review of Economics and Statistics, 46 (2), 145154.

1

Prepared by Carlos de Resende, Alice Fan, and Sadhna Naik.

3

For Estonia, at is relatively stable and mean-reverting but it has been increasing markedly in both Latvia and Lithuania since around 2015. For that reason, in the estimation of total factor productivity, instead of a fixed calibrated value for α (e.g., at its historical average or last observed value) a smooth, time-varying labor share was used for the three countries.

5

Using Augmented Dickey-Fuller tests with a test specification that includes both a constant and a deterministic linear trend, with lags selected automatically based on Schwartz information criteria.

  • Collapse
  • Expand
Republic of Estonia: Selected Issues
Author:
International Monetary Fund. European Dept.