Ireland: Selected Issues
Author: International Monetary Fund. European Dept.
Publication Date:
28 Jun 2018
eISBN:
9781484363874
Language:
English
Keywords:
ISCR; CR; commercial real estate; price; gap elasticity; unemployment rate gap; Phillips curve relationship; output gap estimation; labor force cycle; Wages; Output gap; Global
Download PDF (1 MB)

Selected Issues

Abstract

Selected Issues

Business Cycle in Ireland: Accounting for Open Labor Market and Multinationals1

Assessing the business cycle in Ireland is complicated by the open character of its labor market and large presence of globally active multinationals. This chapter takes into account these features of the Irish economy when estimating potential output and the cyclical position. Results point to a strong recent rebound of potential growth as well as a currently advanced cyclical position, with a positive and widening output gap.

A. Introduction

1. This chapter estimates the cyclical position of the Irish economy. The recent publication of the GNI* and other related indicators, which help to identify the large global activities of multinationals and provide a better measure of the size of the domestic economy, has invited for a deeper analysis of the cyclical position of the Irish economy.2 A better understanding of the cyclical position is essential for macroeconomic diagnostics, including assessing external sustainability and determining an appropriate fiscal policy stance.

2. Despite helpful supplementary data releases, several features of the Irish economy continue to pose challenges for business cycle identification. First, a large multinational sector, accounting for an important share of Irish GDP, tends to behave differently from the domestic economy and arguably it is not responsive to domestic demand management. Second, given the large trade with the United Kingdom, the headline inflation in Ireland often reflects movements in the Euro/Sterling exchange rate and thus does not provide a reliable measure of domestic inflation pressures as signs of slack in the economy. And finally, the usual cyclical signals from wage growth or the unemployment rate gap are likely to be often weakened by the very open and flexible character of the Irish labor market.

3. Given these challenges and general uncertainty around potential output estimates, a suite of adapted models has been employed for potential output estimation. Namely, univariate (HP) and multivariate filters, and the production function have been applied to the GNI* in constant prices.3 The multinational sector has been assumed to continuously operate at full capacity and added to potential GNI* to derive potential GDP. The central element in the structural models is the labor market, namely the wage Phillips curve and Okun’s law (multivariate filter), and potential employment growth (production function). Equilibrium in the labor market is estimated separately by factoring in the importance of migration in the Irish labor market.

4. The Irish economy is in the midst of a cyclical upswing. All methods suggest a positive output gap in 2017, while the labor market shows signs of upward wage pressures as net immigration has been weak so far. These signs are consistent with a cyclical upswing, amid strong estimated potential output growth, and point to risks of a boom-bust cycle, should the economy continue to push the growth momentum.

B. Equilibrium in a Small Open Labor Market

5. Labor market conditions in Ireland are affected by both internal and external job mobility. Wage dynamics, a yardstick for assessing the labor market cycle, is typically associated with the unemployment rate gap (through the Phillips curve), where a negative gap speeds up wage growth and vice versa.4 However, this approach captures only internal job mobility, when most variation in the unemployment rate stems from changes in unemployment while the labor force is a steadily-moving variable. In Ireland, the labor force exhibits procyclical patterns due to migration swings, showing signs of a large external job mobility (Figure 1).

6. External job mobility is a particularly important element in the Irish labor market and may have effects on wage dynamics. As in other countries with small open labor markets (for instance Cyprus, Malta, and Luxembourg), the Irish labor market is prone to cyclical swings in the labor force (Figure 1). Labor force fluctuations have been driven by the age group of 14–44, that is, the age group of most net migrants. This can also be seen in the high correlation between changes in the labor force and net migration. In addition, the high correlation between changes in the labor force and employment reveals the jobs-driven nature of net migration, which also affects the overall labor participation rate (see also Byrne and O’Brien, 2016).5, 6 And finally, native Irish account for an important part of labor force swings. These features suggest that a part of the labor market cycle is not captured by the unemployment rate (because the labor force and employment move jointly) and that the labor force cycle may have additional effects on wage dynamics, on top of the unemployment rate gap.

Figure 1.
Figure 1.

Labor Market

Citation: IMF Staff Country Reports 2018, 195; 10.5089/9781484363874.002.A001

7. There are signs that the labor force cycle may have weakened the Phillips curve relationship, hereby suppressing wage signals about the labor market cycle. For instance, in the 1990s, wage growth accelerated only marginally when unemployment plummeted 10 percentage points also due to outward migration.7 In 1999–2000, wage growth accelerated strongly amid a rather steady unemployment rate. Reversely, wage growth has been trend decelerating during the early 2000s, while the unemployment rate stayed low and did not change much. Migration patterns also affect Okun’s Law. During the early 2000s, especially after opening the labor market to new EU member states in 2004, GDP and employment growth has been strong while the unemployment rate was stable (Figure 2). As a result, although the labor market may not show wage pressures and declining unemployment, employment and GDP growth can be strong and may lead to unsustainable dynamics. 8, 9

Figure 2.
Figure 2.

Weakened Cyclical Signals

Citation: IMF Staff Country Reports 2018, 195; 10.5089/9781484363874.002.A001

8. Extending the Phillips curve specification with the labor force cycle may account for the small open labor market in Ireland. Based on the premise that only the cyclical part of the unemployment rate affects the cycle in wage growth, a change in wage growth is modeled as a function of the unemployment rate gap, represented by the difference between the actual unemployment rate and the “non-accelerating wage growth rate of unemployment (NAWRU)” (see for instance Elmeskov and MacFarland, 1993). This concept has become a key element of estimation models (Blagrave, et al., 2015; Denis, et al., 2006). Nevertheless, such a specification is mostly applicable to a closed economy setting. Small, highly open labor markets with job-driven moves of the labor force across national borders calls for extending the Phillips curve with a labor force gap – the percentage deviation of the actual labor force from the “non-accelerating wage growth labor force growth (NAWLF)”, to account for wage pressures stemming from swings in the labor force.

9. Estimating the NAWRU and NAWLF requires modeling the components of the unemployment rate, which needs an alternative estimation approach. The usual estimation of NAWRU (ut*) as a state variable using a standard structural Kalman filter does not provide a theoretical base to assure consistency between the filtered NAWRU and its components, i.e., the trends in employment and the labor force. Moreover, as can be seen on an expanded Phillips curve model below, the need for modeling these trends inhibits application of the usual structural Kalman filter since the trends (state variables) enter the signal equation in a non-linear way (as a ratio Et*LFt*), which is not permitted:

ΔWt=α+θΔWt−1−β(ut−ut*)−ξ(LFtLFt*)+εt.

Noting that the unemployment rate gap (ut – ut*) is equal to (Et*LFt*−EtLFt), the Phillips curve model for wage W can be written as:

ΔWt=α+θΔWt−1−β(Et*LFt*−EtLFt)−ξ(LFtLFt*)+εt,

Where E and LF are employment and labor force, respectively, and ε is an i.i.d. error term. Therefore, an alternative multivariate Kalman filter model is needed.

10. The trend-cycle decomposition of employment and the labor force series is modeled using a simple, multivariate Kalman filter. Stochastic trends and cycles of employment (E) and the labor force (LF) are modeled using the Watson (1986) specification and extended to a multivariate Kalman filter setting by the assumption that the cycles in both series are co-determined by a common factor – net migration flows (NM) – and the estimated variance (σ) of the trends is assumed to be 500 times smaller than that of the cycles (similar to the noise-to-signal ratio in case of the Hodrick-Prescott filer). The model reads as follows:10

Et=Et*+EtcLFt=LFt*+LFtcEt*=α+Et−1*+εtε∼iidN(0,σ)Etc=γEt−1c+δNMt+υtυ∼iidN(0,σ*500)LFt*=λ+LFt−1*+ψtψ∼iidN(0,σ)LFtc=μLFt−1c+θNMt+φtφ∼iidN(0,σ*500)

Other estimated parameters: α, γ, δ, λ, μ, σ and θ. Superscript c denotes cycle; * denotes trends.

11. Resulting trends represent sustainable paths in employment and the labor force (Table 1 and Figure 3). Plotted against the trends, actual values point to the cycle of the labor force and employment. The cycle turned out to be unsustainable during 2005–2008 and the subsequent correction in employment has undershoot the potential employment. The significant recent hiring brought employment back close to the sustainable trend.

Table 1.

Ireland: Trend and Cycle Estimation

article image
Source: IMF staff calculations.
Figure 3.
Figure 3.

Stochastic Trends

Citation: IMF Staff Country Reports 2018, 195; 10.5089/9781484363874.002.A001

12. According to estimates, the equilibrium unemployment rate has been close to 9 percent over the last two decades (Figure 4). The derived stochastic trends of the labor force and employment determine the estimate of NAWRU=1−Et*LFt*. The estimated NAWRU’s path of a mild decline during the boom of early 2000s and an increase during the crisis seems intuitive, as hiring less skilled workers during booms reduces structural unemployment, while layoffs during the crisis lead to gradual deterioration of skills of the unemployed and increase the equilibrium level of unemployment. In comparison to structural measures of unemployment, such as the non-employment index, the estimated NAWRU of 8 percent at end-2016 seems to be a plausible estimate as it is close to the non-employment index of 9 percent.11

Figure 4.
Figure 4.

Unemployment Rates

(percent)

Citation: IMF Staff Country Reports 2018, 195; 10.5089/9781484363874.002.A001

Sources: CSO and IMF staff.

13. The plausibility of the implied unemployment rate gap and the labor force gap is tested in a Phillips curve estimation (Figure 5 and Table 2).12 Using a standard wage Phillips curve, we test the unemployment rate gaps by OECD, EC, and the here derived stochastic gap. Gaps showing higher amplitudes tend to perform better than others and imply that Sources: CSO and IMF staff. the equilibrium level is likely sluggish. This is also intuitive as the Irish labor market is one of the most flexible in the euro area13 and therefore would be expected to exhibit relatively large movements of unemployment compared to its equilibrium level. In addition, the Phillips curve extended with the labor force gap, shows a significant improvement in the Phillips curve specification14 and allows separating the effects of the labor force cycle from the unemployment rate cycle.

Figure 5.
Figure 5.

Labor Market Gaps

Citation: IMF Staff Country Reports 2018, 195; 10.5089/9781484363874.002.A001

14. Wage pressures have been importantly affected by both the unemployment rate gap and swings in net migration. Based on the regression results (Table 2, column II), observed wage growth appears to be dwarfed by the large net immigration during the boom of 2005–08 (Figure 6). If the labor force would have grown at a sustainable pace (in line with the above derived stochastic trend), wage growth would have been higher during the boom, exceeding the in-sample average wage growth and in line with the positive unemployment rate gap. Similarly, in recent years, with a broadly closed labor force gap, rising wage pressures reflect the sharply falling unemployment rate gap.

Figure 6.
Figure 6.

Wage Growth Decomposition

Wage Per Employee

(y/y; percentage points)

Citation: IMF Staff Country Reports 2018, 195; 10.5089/9781484363874.002.A001

Sources: Haver; OECD; and IMF staff.
Table 2.

Ireland: Wage Phillips Curve Estimation

article image
Note: Wage growth as dependent variable; using quarterly data series; OLS.Source: IMF staff calculations.

C. Potential Output of a Globally-Integrated Economy

15. Potential output for Ireland is modeled as the sum of separately evaluated domestic and multinational parts. While the multinational sector is assumed to be continuously operating at full capacity (Figure 7), the usual methods to derive potential output – univariate and multivariate filters as well as the production function approach, are applied to the domestic economy, measured by GNI* in constant prices, in quarterly frequency. Given significant uncertainty surrounding potential growth estimates in general, the use of several methods should increase confidence in the estimates. The potential output (GDPp) is therefore a sum of the potential output for the domestic economy (GNI*p) and the multinationals (GDPMNE):15

Figure 7.
Figure 7.

GDP Growth Contributions

(y/y percent change)

Citation: IMF Staff Country Reports 2018, 195; 10.5089/9781484363874.002.A001

Sources: CSO and IMF staff.
GDPp=GNI*p+GDPMNE.

16. The univariate filter (Hodrick-Prescott). The quarterly GNI* series in constant prices (see footnote 2) has been filtered with a smoothing parameter of 1600, commonly used for quarterly data. The choice of using quarterly data for estimation, while presenting annual series is likely to reduce the influence of the end-point biases of univariate filters on inference about potential growth.

17. The multivariate filter is an adapted model of the IMF’s Research Department that relates the output gap to inflation and the unemployment rate gap (Blagrave, et al., 2015). The core of the model is the Phillips curve, relating the output gap (an unobservable variable yt, assumed to follow a random walk) to observable data on wage inflation (πtw) instead of headline inflation, due to large external effects on headline inflation in the small open Irish economy. The specification is extended for the labor force gap (gt) to account for the wage growth effects from labor force swings:

πtw=ρπt−1w+αgt+βyt+εtπ.
Table 3.

Ireland: Structural Multivariate Filter

article image
Source: IMF staff calculations.

The Phillips curve is complemented with Okun’s law, governing the relationship between the unemployment rate gap (ut) and the output gap:

ut=θut−1+δyt+εtu.

The unemployment rate gap is not an unobserved variable, but it is estimated separately (using the stochastic trend, derived in the previous section) and becomes an observed variable in the model. εt denotes random errors. The model uses quarterly data series. Table 3 shows the results of estimation.

18. The production function methodology uses the commonly applied Cobb-Douglass production function with capital and labor inputs. What remains unexplained after accounting for capital and labor inputs is attributed the total factor productivity Bt (the Solow residual). The specification for quarterly real output γ (GNI*) is as follows:

Yt=BtKt(1−α)(EDUtLt)α,

where Bt=AtCUt(1−α)(AHWt)α. By accounting for the cyclical utilization of production factors pertaining to capital and labor, namely, capital utilization and average hours worked, we isolate the “structural” part of the total factor productivity At, and further:

  • K denotes the capital stock, derived using the usual perpetual inventory model (Epstein and Macchiarelli, 2010, and Teixeira de Silva, 2001) as Kt = (1-ρ) Kt-1 + lt, where ρ is the depreciation rate calibrated using the historical average (taken from Penn World Tables 9.0) and / stands for the modified real domestic investment, filtering out the investments of IP-related MNE investment. The capital stock of 1997 was taken as the initial capital stock, assuming that the capital stock then has not been affected by IP-related MNE investments.

  • L denotes the number of employed persons, using the national accounts concept.

  • EDU is the human capital index, taken from the Penn World Tables 9.0.

  • CU stands for capacity utilization in manufacturing industry. It is a survey-based measure expressed as a balance of responses. Due to the lack of publicly available Ireland-specific series, it is approximated by the Euro Area average.16

  • AHW stands for the average hours worked corresponding to national accounts.

  • α stands for the labor share in the production function. The labor share has been calculated as the ratio of compensation of employees to gross value added, taken from the Penn World Tables 9.0.

Potential output has been derived using the calibrated labor share, smoothed trends in employment (that is, the stochastic trend derived in the previous section) and the total factor productivity A (by HP filter with smoothing factor 1600), and the actual capital stock. The average hours worked and capital utilization were assumed to be at their long-term values.

19. Estimated potential output growth has recovered to its pre-crisis rate. All methods point to a strong recovery of potential GNI* growth in the post-crisis period (Figure 8). The GDP potential growth remained positive even during the crisis, thanks to the positive effects of the multinational sector. Current rates of potential GNI* growth are strong and comparable to those recorded during the 1990s, when Ireland was dabbed “The Celtic Tiger” (Dermot McAleese, 2000).

Figure 8.
Figure 8.

Potential Growth

Citation: IMF Staff Country Reports 2018, 195; 10.5089/9781484363874.002.A001

20. The rebound in potential growth has been driven by gains in productivity and investment. Using the production function method, which allows a decomposition into factors, the major drivers of potential GNI* growth in recent years were increases in total factor productivity and non-MNE’s related capital accumulation (Figure 9). The MNE sector has been consistently adding momentum to potential GDP growth, particularly helpful in smoothing the impact of the crisis.

Figure 9.
Figure 9.

Potential Growth Decomposition

(y/y percent change)

Citation: IMF Staff Country Reports 2018, 195; 10.5089/9781484363874.002.A001

Sources: Haver; Penn World Table 9.0; and IMF staff.

D. The Implied Business Cycle Position

21. Despite the strong potential growth in recent years, the output gap has become positive. Figure 10 shows the average and the range of the output gap for GNI* and GDP, as implied by the three methods employed for the potential output estimation. While potential growth has rebounded strongly, the output gap switched from negative to positive in 2015. The output gap for the domestic economy (GNI*) tends to be more volatile compared to that for GDP, since the multinationals are assumed to continuously operate with a closed output gap.

Figure 10.
Figure 10.

Output Gap Estimates

Citation: IMF Staff Country Reports 2018, 195; 10.5089/9781484363874.002.A001

22. Changes in the output gap are driven by cyclical employment and capacity utilization. According to the decomposition of the output gap for GNI*, derived through the production function approach, the major drivers of the recently positive and widening output gap are cyclical employment and stretched capacity utilization (Figure 11). These factors were also behind the boom-bust cycle of late 2000s. Average hours worked gradually approach long-term values and reduce the drag on the output gap as well.

Figure 11.
Figure 11.

Output Gap Decomposition

GNI* Output Gap

(percentage points)

Citation: IMF Staff Country Reports 2018, 195; 10.5089/9781484363874.002.A001

Sources: IMF staff.

23. These results suggest that the Irish economy has firmly moved into the upswing phase of a business cycle. Notwithstanding the general uncertainty surrounding output gap estimations, the currently positive output gap is confirmed with reasonable confidence as the range of the three applied methods has recently narrowed. The economy is in the midst of a cyclical upswing.

References

  • Byrne, S., M. O’Brien, 2016, “Understanding Irish Labour Force Participation”, Research Technical Papers 01/RT/16. Central Bank of Ireland.

    • Search Google Scholar
    • Export Citation

  • Blagrave, P., R. Garcia-Saltos, D. Laxton, and F. Zhang, 2015, “A Simple Multivariate Filter for Estimating Potential Output,” IMF Working Paper 15/79.

    • Search Google Scholar
    • Export Citation

  • Clancy, Daragh, 2013, “Output Gap Estimation Uncertainty: Extracting the TFP Cycle Using an Aggregated PMI Series”, The Economic and Social Review, Vol. 44, No. 1, Spring, pp. 1–18.

    • Search Google Scholar
    • Export Citation

  • Denis, Cecile, Daniel Grenouilleau, Kieran Mc Morrow, and Werner Roger, 2006, “Calculating Potential Growth Rates and Output Gaps – A revised Production Function Approach”, Economic Papers No. 247, ECFIN, European Commission.

    • Search Google Scholar
    • Export Citation

  • Elmeskov, J. and MacFarland, M. (1993), “Unemployment Persistence”, OECD Economic Studies 21, pp. 59–88.

  • Epstein, N. and C. Macchiarelli, 2010, “Estimating Poland’s Potential Output: A Production Function Approach”, IMF Working Paper No. 15.

    • Search Google Scholar
    • Export Citation

  • IMF, 2010, “Decomposing Irish Growth”, Box 1, IMF Country Report No. 9/195, International Monetary Fund.

  • IFAC, 2018, “Inside the “Upside Down”: Estimating Ireland’s Output Gap,” Working Paper 5, Irish Fiscal Advisory Council, January 2018.

    • Search Google Scholar
    • Export Citation

  • Linehan, Suzane, Reamonn Lydon, Tara McIndoe-Calder, Paul Reddan, and Diarmaid Smyth, 2017, “The Labour Market and Wage Growth after a Crisis”, Quarterly Bulletin 04, October 2017, Central Bank of Ireland.

    • Search Google Scholar
    • Export Citation

  • Lawson, Robert A., Edward Bierhanzl, 2004, “Labor market flexibility: An index approach to cross-country comparisons”, Journal of Labor Research, Volume 25, Issue 1, pp 117–126, March 2004.

    • Search Google Scholar
    • Export Citation

  • Lozej, Matija, 2017, “Economic Migration and Business Cycles in a Small Open Economy with Matching Frictions”, Manuscript, Central Bank of Ireland, 2017.

    • Search Google Scholar
    • Export Citation

  • McAleese, Dermot, 2000, “The Celtic Tiger: Origins and Prospects”, Policy Options, July-August 2000, pp. 46–50.

  • Szeto, K. L., M. Guy, 2004 “Estimating New Zealand NAIRU”, New Zealand Treasury, WP 04/10.

  • Teixeira de Silva, T., 2001, “Estimating Brazilian Potential Output: A Production Function Approach”, Working Paper, Research Department, Central Bank of Brazil.

    • Search Google Scholar
    • Export Citation

  • Watson, M. W., 1986, “Univariate detrending methods with stochastic trends”, Journal of Monetary Economics, Vol. 18, No. 1, pp. 49–75.

    • Search Google Scholar
    • Export Citation
1

Prepared by Jiří Podpiera. This version benefitted from comments and suggestions by the discussant Michael O’Grady and other participants at the Central Bank of Ireland’s Workshop on May 8, 2018, in Dublin.

2

The GNI* aims to measure primarily the domestic economy, however, it does so by an approximation: GNI* = GDP – (Net Factor Income) – (Factor Income of Re-Domiciled Companies) – (Depreciation on R&D-related IP imports) – (Depreciation on Aircraft Related to Leasing). Therefore, GNI* may still contain some activities of multinationals. In this chapter, we use GNI* as a proxy for the mostly-domestic part of the economy, henceforth referred to as ‘the domestic economy’.

3

The GNI* series is available only in nominal values and at annual frequency. The constant-prices, quarterly GNI* has been derived by applying the seasonality of modified domestic demand to GNI* and deflating by the deflator of real modified domestic demand.

4

The unemployment rate gap is the difference between the unemployment rate and the so called “non- accelerating wage rate of unemployment (NAWRU)”. The underlying assumption behind this concept is that only cyclical unemployment affects labor cost developments.

5

The joint co-determination of the labor force participation rate, the labor force, and employment by net migration motivates modeling potential employment directly, rather than involving the working age population and the labor force participation rate.

6

The increase in the overall labor force participation rate during the 2000s has been in part due to one-time catchup in female labor force participation.

7

Wages per employee, using OECD data http://stats.oecd.org/Index.aspx?DataSetCode=EO#

8

See, for instance, Lozej (2017) for evidence that migration increases business cycle fluctuations.

9

See also Box 1, IMF Country Report No. 9/159.

10

Watson, M. W., 1986, “Univariate detrending methods with stochastic trends”, Journal of Monetary Economics, Vol. 18, No. 1, pp. 49–75.

12

Similar to testing of various unemployment gaps in Table 3 in K.L. Szeto, M. Guy, 2004 “Estimating New Zealand NAIRU”, New Zealand Treasury, WP 04/10.

14

This is not the case for the labor force gap of the OECD (Table 2, column IV.), perhaps due to an inconsistency between trend labor force and employment in the NAWRU estimation. The unemployment rate gap of the European Commission turns out to be statistically and economically insignificant (Table 2, column V.).

15

Similar approach has been taken by the Irish Fiscal Advisory Council, see Working Paper 5, 2018, using the gross value added (GVA). Namely, splitting GVA into mostly domestic and mostly MNE parts; assuming that the mostly-MNE sector operates at full capacity at all times; and deriving overall potential output as the sum of the estimated potential output using the mostly-domestic GVA and the mostly-MNE’s GVA.

16

The choice of the proxy for capacity utilization may somewhat alter results of the output gap decomposition. Alternative proxies used include the Purchasing Managers’ Index (in Clancy, 2013) and the Irish capacity utilization measure produced by a proprietary Markit® Economics database (used by the Central Bank of Ireland).

Same Series

Other IMF Content

Other Publishers

Cover IMF Staff Country Reports
Ireland: Selected Issues
Author: International Monetary Fund. European Dept.
Volume/Issue:
Volume 2018: Issue 195
Publisher:
International Monetary Fund
ISBN:
9781484363874
ISSN:
1934-7685
Pages:
62
DOI:
https://doi.org/10.5089/9781484363874.002

Labor Market

Weakened Cyclical Signals

Stochastic Trends

Unemployment Rates

(percent)

Labor Market Gaps

Wage Growth Decomposition

Wage Per Employee

(y/y; percentage points)

GDP Growth Contributions

(y/y percent change)

Potential Growth

Potential Growth Decomposition

(y/y percent change)

Output Gap Estimates

Output Gap Decomposition

GNI* Output Gap

(percentage points)