Journal Issue
Share
Article

Norway: Selected Issues

Author(s):
International Monetary Fund. European Dept.
Published Date:
August 2014
Share
  • ShareShare
Show Summary Details

Potential Output and Immigration in Norway1

Norway has experienced a sharp increase in immigration inflows since mid-2000. This chapter examines migration patterns in Norway and their implications for estimates of potential output. The data show that immigration inflows into Norway vary across source countries. Immigration patterns in Norway contain both cyclical and structural elements, but the latter seems dominant at least for now. Empirical results also suggest that immigration plays some role in determining potential output, but its impact is quite small, consistent with the view that the recent immigration patterns are structural.

A. Introduction

1. Potential output is widely used in macroeconomic policy making. Despite their importance, estimates of potential output are subject to significant uncertainty because it is not observable ex ante or ex post. Standard approaches use some type of smoothing, either directly smoothing output (e.g. HP filter) or by smoothing inputs (e.g. production function approach). Others take more of a structural approach using additional information such as inflation and unemployment.

2. Various idiosyncratic factors complicate generating reliable estimates of potential output. In the case of Norway, challenges arise from its unique economic structure. As a major oil and natural gas exporter, the Norwegian economy is greatly influenced by commodity price movements. The mainland economy has become more dependent on the offshore activity through increased demand for inputs and services from the mainland economy. High oil prices thus stimulate both the offshore sector and the mainland economy. Moreover, attracted by high wages in Norway, inward labor mobility has increased since the mid-2000s. Labor supply in Norway may therefore depend on economic cycles, and simply applying smoothing to data could bias the degree of overheating if these cycles are not appropriately accounted for.

3. The nature of changes in labor supply would affect estimated potential output. If one believes the increase in labor supply in recent years is cyclical, then simply smoothing the series could overestimate potential output, as the additional labor supply will boost growth but common filters will treat this increase in output as a permanent increase in the potential output. In this case, common filters will underestimate output gap in upturns and overestimate output gap in downturns.

4. This study examines immigration patterns in Norway and estimates potential output using various methodologies. In particular, it applies a new methodology proposed by Borio and others (2013) to estimate potential output by drawing on information about immigration and oil price movements. The next section provides an overview of the recent trend in immigration in Norway. Section C discusses various estimates of potential output using standard approaches. Section D will estimate “immigration neutral” potential output. Section E concludes.

B. Immigration and Economic Cycles

Immigration patterns

5. Norway has seen a surge in immigration in recent years. Recent growth in immigration deviates substantially from its historical trend. During the 1990s, roughly thirty thousand immigrants arrived in Norway annually, and this number more than doubled by 2010-2012. Immigration has been also contributing to the growth in labor force in Norway. Net migration was roughly 1 percent of total labor force during 1990-2012, and this ratio has been rising in recent years.

6. This surge in immigration stands out even compared with experiences of other OECD countries. Other Nordic countries experienced acceleration of immigration inflows since mid to late 2000s, but Norway saw the most significant increase among these comparators (Figure 1).

Figure 1.Labor Market Indicators: Cross-Country Comparison Immigration

Sources: Haver Analytics and IMF staff calculations.

1/ Selected OECD countries are Australia, Canada, US, and UK.

7. The largest share of immigrants is from Poland, accounting for 15 percent of the total in 2012. Immigration from Poland and Lithuania started to grow rapidly since 2004 when they joined the European Union (EU). This in part accounts for a rapid increase in immigration in mid 2000s, but inflows of immigrants from other countries also increased substantially during this time period. Other source countries include neighboring countries such as Sweden and Denmark.

Labor Market and Immigration

(Millions of persons)

Sources: IMF World Economic Outlook, Statistics Norway and Fund staff calculations.

The Origin of Immigration, 2012

(Share ot total immigration, percent)

Sources: Statistics Norway and IMF staff calculations.

8. Net migration patterns vary across source countries in part because of different emigration patterns (Figure 2). Net migration from Sweden seems be explained mostly by economic cycles in Norway and Sweden. On the other hand, net migration from Poland and Lithuania is dominated more by inward movement into Norway. For these countries, movements of immigrants are more one-way than two-way flows. Lastly, immigratns are employed in a wide range of industries in Norway.

Figure 2.Net Migration to Norway from Selected Countries

Sources: Statistics Norway and IMF staff calculations.

Employment of Immigrants by Industry, 2012

(Share of total immingration, percent)

Sources: Statistics Norway and IMF staff calculations.

Factors affecting immigration

9. Overall, the oil price hike seems to be behind the surge in immigration in recent years. Oil prices have been rising since early 2000s, and it continued to increase with a brief interruption during the global financial crisis. During 2000-2013, oil prices grew by more than 10 percent annually. The timing of acceleration in oil price growth corresponds to the surge in immigration into Norway.2 The mainland economy also grew at a robust rate during this time period, growing at 4.8 percent on average during 2004-2007. At the same time, legal changes also played a role, as the EU enlargement took place in 2004, allowing new EU members to move freely across EU/EEA countries.

Immigration Inflows in Norway

(Persons, in thousands, USD, left; percent change, right)

Sources: Statistics Norway and IMF staff calculations.

10. Labor market conditions have been persistently favorable in Norway compared with its peers (Figure 1). Real wages grew at a faster pace than comparator countries, and continued to grow even after the global financial crisis. Unemployment rates have been low while employment was least affected by the global financial crisis. These favorable labor market conditions seem to have worked as a strong pull factor for immigration.

11. There are multiple factors as well as uncertainties that need to be taken into account on the role of immigration in potential output. For example, the recent surge in immigration is correlated with oil price movements, but it could be too soon to tell whether this trend will end when oil prices drop because immigration from countries like Poland and Lithuania may have both cyclical and structural components. Experiences from other countries also suggest that how immigrants would respond to a large economic downturn could differ across countries (Box 1). The rest of the paper will estimate potential output with various methodologies and discuss the results in light of these observations.

Box 1.Experience of Immigration Patterns from Spain and United Kingdom

Spain and the United Kingdom have both experienced a sharp increase in immigration inflows from Poland in the mid-2000s. Subsequently, both countries saw economic downturn to different degrees due to the global financial crisis. The immigration pattern proved to be quite different in the two countries as the immigration inflows returned to pre-crisis levels in Spain after the housing bust while in the U.K., the immigration remained relatively high even after the global financial crisis.

Spain: Immigration from Poland

(Thousands/Percent, left; Thousands, right)

Sources: Eurostat and Fund staff calculations.

U.K.: Immigration from Poland

(Thousands/Percent, left; Thousands, right)

Sources: Eurostat and Fund staff calculations.

C. Estimating Potential Output Using Standard Approaches

12. Output gap estimates vary across different institutions.3 The table below reports mainland output gap estimates from Norges Bank, Statistics Norway, OECD, and the Fund staff as reflected in the World Economic Outlook.4 These estimates suggest output gaps ranging from -1.0 to -0.3 percent of potential output in 2014. A comparison of these estimates reveals the uncertainty associated with the underlying estimates. This section will estimate potential output for the mainland economy using various standard approaches and discuss the implications of these estimates.

Mainland GDP Output Gap Estimates(Percent of potential output)
2012201320142015
Norges Bank 1/0.30.0−0.6−0.7
OECD−0.4−0.9−0.8−0.4
Statistics Norway 1/−0.3−0.5−1.0−0.9
Fund staff0.2−0.1−0.3−0.3
Sources: Norges Bank Monetary Policy Report 1/14, OECD, Statistics Norway, World Economic Outlook April 2014.

Average of quarterly estimates.

Sources: Norges Bank Monetary Policy Report 1/14, OECD, Statistics Norway, World Economic Outlook April 2014.

Average of quarterly estimates.

13. Two standard methods are used to estimate potential output: A simple HP filter and a production function approach are employed on annual data. For the univariate HP filter, the smoothing parameter is chosen to cover ranges suggested in the literature. For the production function approach, an HP filter was used to obtain smoothed time series of inputs: (i) employment, computed as the total hours worked in the mainland economy; (ii) mainland capital stock; and (iii) total factor productivity, with factor intensity calibrated to the Norwegian economy.

14. Estimates point to a negative output gap for the mainland economy in 2014. These results suggest that the output of the mainland economy has been close to potential for the past few years. The average estimate suggests that output gap was turned slightly negative in 2013 and it is estimated at -0.32 percent in 2014. The output gap in the mainland economy is projected to begin to narrow in 2015.

15. Potential growth of the mainland economy is estimated to be lower than the pre-crisis level. Estimated potential growth was robust in mid-2000 before the crisis reflecting the strong real output growth of the mainland economy during 2004-2007 with average growth at 4.8 percent. The estimate suggests that potential growth of the mainland GDP is about 2.3 percent in 2014.

Mainland Economy: Output Gap Estimates

(Percent of potential output)

Sources: Eurostat, Haver Analytics, IMF World Economic Outlook, OECD, Statistics Finland and Fund staff calculations.

Mainland GDP: Potential Growth Estimates

(Percent)

Sources: Eurostat, Haver Analytics, IMF World Economic Outlook, OECD, Statistics Finland and Fund staff calculations.

16. The production function approach suggests an increasing role of labor in accounting for potential output growth in recent years. A large part of potential growth came from TFP growth in late 1990s to early 2000s, but the contribution from TFP has been declining while that from labor started to rise in mid 2000s. This timing coincides with the timing of a surge in immigration.

Contribution to Potential Growth

(Percentage points unless otherwise indicated)

Sources: Eurostat, Haver Analytics, IMF World Economic Outlook, Norges Bank, OECD, Statistics Finland and Fund staff calculations.

17. These output gap estimates implicitly assume that employment growth in recent years is structural rather than cyclical. Section B has shown, however, that part of the immigration into Norway could be cyclical. The next section will use a different approach to examine whether the cyclical component of immigration would matter empirically for output gap estimation.

D. “Immigration Neutral” Potential Output

18. This section will discuss the methodology proposed by Borio and others (2013) and apply the method to obtain “immigration neutral” potential output for Norway. Their methodology is motivated by the observation that inflation may not necessarily be the right indicator that signals overheating of an economy. As the euro area crisis has shown, much of the pre-crisis growth in the booming euro area economies was driven by finance, real estate, and construction with inflation showing no obvious sign of overheating. These activities declined substantially after the crisis. In light of this, Borio and others (2013) propose a statistical method to draw on information on other variables that are likely to capture financial cycles, such as real credit growth, property price growth, and real interest rate, to estimate “finance neutral” potential output. This section will extend their methodology by including immigration inflows and oil prices to estimate potential output for Norway.

19. The state space model was applied to quarterly mainland GDP to produce potential output estimates which are directly comparable with the HP filter (Box 2). The state space model is an augmented version of Borio et al (2013), which allows for directly estimating unobserved variables such as potential output.

20. The results indicate that immigration plays a small but statistically significant role in the estimation of potential output for Norway. Consistent with our prior, the HP filter underestimates the output gap in upturns and overestimate downturns, at least in the recent time period. But the effect is relatively small. As an alternative specification, oil prices were also employed as an explanatory variable. If significant, it would suggest that the potential estimates with immigration were still picking up some of the cyclicality from th4e exogenous oil prices, given the high correlation between the oil price and immigration. However, oil prices were not found to be a significant predictor of potential output.

Output Gap, Mainland

(Percent of potential GDP)

Sources: Haver Analytics, Statistics Norway, and Fund staff calculations.

E. Conclusion

21. Immigration patterns in Norway contain both cyclical and structural elements. Immigration from Sweden and Denmark (to lesser extent) seem cyclical in nature, following either strong growth in Norway, an uptick in oil prices, or a downturn in source countries’ local economic conditions. On the other hand, immigration inflows from countries like Poland and Lithuania appear to be more structural at least for now, likely driven by relative expected income differentials. One caveat is that the nature of immigration could turn out to be cyclical if the economy experiences a sharp downturn as seen in other countries.

22. The “immigration neutral” results suggest that immigration plays some role in the determination of potential output. Simple trend calculations would understate the actual degree of overheating given that some of the immigration is cyclical. However, the impact seems relatively small given that a larger share of immigration appears to be more structural than cyclical in recent years.

Box 2.Technical Details of Immigration Neutral Estimation

The model. The state space model, an augmented version of Borio et al (2013), expands the HP Filter by adding additional covariates that help identify the transitory part of GDP, albeit without structural constraints. Reducing the state space model, the estimating equation is as follows:

where y is real GDP, y* is potential output, and x is a vector of observables which contains information on transitory variables. Built on the HP filter in a state-space framework—a standard method to estimate unobserved variables—this equation includes an autoregressive output gap term and additional transitory variables (without the transitory variables, the equation reduces to the HP filter). The advantage of this approach is that estimates from the HP filter can be used as a baseline benchmark for comparison. To produce comparable results with the HP filter, the signal-to-noise ratios for equation (1) and the HP filter are equated so that the frequency cutoff, namely the length of the cycles, matches. This is achieved by imposing a restriction on the variance such that:

Estimation. While Borio et al (2013) employ a Bayesian approach, this paper uses maximum likelihood estimation (MLE) to estimate the model on quarterly data (Mrkaic, 2014). ρ and β are estimated in a two-step procedure. First, the autoregressive parameter ρ is estimated by running an AR(1) regression on the output gap obtained from the simple HP filter. Then ρ is substituted into (1) and estimated using MLE. All time series are demeaned to reduce pro-cyclicality and differenced to account for unit roots. Specifically, the measurement equation becomes:

where y − y* refers to the output gap, immigration is immigration inflows, and ε is a disturbance term.

To satisfy equation (2), the HP filter estimate is calculated using lambda=1600 (Maravell and del Rio, 2001; Ravn and Uhlig, 2002) and the signal to noise ratio is computed. Then the model is adjusted by calibrating a restriction on the variance to produce the same signal to noise ratio as the HP filter to ensure full comparability.

References

Prepared by Thomas Dowling and Kazuko Shirono (both EUR).

The correlation between net migration and oil prices during 1990-2012 was about 0.9.

Potential GDP of the mainland economy is more relevant for policy considerations in Norway. Monetary policy is assessed based on mainland economic activity, among other factors, and Norway’s fiscal impulse is measured in terms of a change in non-oil deficit as a share of potential mainland GDP. Thus this paper focuses on potential output of the mainland economy.

The OECD’s estimates are based on a production function approach (Girono and others, 1995). Norges Bank’s output gap assessment entails trend calculations of the mainland GDP, adjusted for various key factors including unemployment, capacity utilization and labor force participation. See Sturød and Hagelund (2012) for more detailed discussions on these factors.

Other Resources Citing This Publication