Selected Issues

Abstract

Selected Issues

Long-Term Growth Prospects and the Output Gap in Montenegro1

The methodologies used in this paper to estimate medium-term and long-term growth—the production function approach and filtering—typically assume stable population growth and an economy not too exposed to external shocks. These assumptions usually apply to relatively large and closed economies. They are not fully satisfied in Montenegro, a small and relatively new country, but this offers opportunities and pitfalls at the same time. In a scenario where Montenegro joins the EU, implements growth-enhancing structural reforms, improves governance, builds human and physical capital, growth could easily exceed the illustrative “high-growth policy scenario” described below. Moreover, in such a virtuous circle, population could exceed the “best” scenario derived from UN population forecasts, which in their median variant drives modest growth in the baseline scenario. The paper shows that the difference between virtuous and vicious circles are mainly driven by policies rather than by model uncertainty.

A. Introduction

1. Projections of long-term growth allow policy makers to assess future available resources, and they have implications for today’s policy choices for fiscal sustainability, pensions, health care, and education. Updating long-term projections regularly and drawing conclusions from this exercise should improve the quality of decision making even if the exercise itself includes significant elements of uncertainty. Similarly, potential output and the output gap are unobserved, yet they are key for policy making. Output gap estimates are used to measure the amount of slack in the economy, help identify the fiscal stance, and gauge the impact of structural reforms. These estimates serve as the basis for macroeconomic policy discussions and guide the appropriateness and timing of consolidation or stimulus policies.

2. Long-term growth is best considered in a production-function context using labor, capital and total factor productivity as inputs. We calculate an estimate of long-term growth based on current policies—the baseline projection—and estimates around the baseline that are due to model uncertainty and those due to policy choices. We also consider upside and downside scenarios for each factor to get a better view of the possible distribution of outcomes. Variations due to model uncertainty are reasonably small, and there is substantial room for policy maneuver, even if the policies themselves are not necessarily easy.

3. For medium-term projections, there are different methods to estimate potential output and the output gap, and it is difficult to assess which is more accurate. For small open economies, the challenges are even bigger as large external shocks and structural breaks create larger fluctuations in the growth profile. In addition, in Montenegro the available time series are short and not all data that would be necessary for a complete analysis are available. Conventional methods in the literature are statistical filtering approaches (Hodrick and Prescott (1997), Baxter and King (1999), Christiano Fitzgerald (2003), Clark (1987), Marcet and Ravn (2004)), econometric approaches (Borio et. el. (2013), Bens et. el. (2010), Blagrave (2015)) and methods based on economic models (Justiniano et. el. (2012), Smets and Wouters (2007), Rabanal and Taheri Sanjani (2015)). Each of the three standard approaches―univariate filters, multivariate filters, and production functions―has advantages and disadvantages. In this paper, we focus on statistical filtering, i.e. two variations of the Hodrick-Prescott (HP) filter, and production function-based methods because of data availability.

4. Montenegro has had a volatile historic growth profile, which adds to the complexity of the computation of long-run growth. Montenegro is a small, open, euroized economy that is highly dependent on tourism and external financing. It has experienced volatile growth since independence in 2006, driven by domestic and external shocks. The short period of reliable GDP data, starting only in 2006, also makes the analysis more challenging.2 After a protracted period of consolidation following the global financial crisis, the economy is growing, bolstered by large investment projects, including the construction of the first section of the Bar-Boljare highway.3 The first section has contributed to growth through demand effects since 2015. It is expected to be completed by 2019, after which some supply side effects are expected.

uA01fig01

Real GDP Growth

(Percent)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Sources: Central Bank of Montenegro; and IMF staff estimates.

5. Results suggest that long-term growth is likely to decline in the absence of major policy efforts due to population dynamics. Our baseline projects growth declining from about 3 percent in 2018 to around zero in 2050. Our policy-based scenarios imply a range of long-run growth paths of 1¼ percentage points above or below our baseline estimates, and possibly more for the upside if productivity improves. Productivity growth has the largest potential impact and could be increased through education and improvements in the business environment. Improvements to labor markets also have potentially large positive effects, as do incentives towards more capital accumulation. Finally, we estimate potential GDP and the current output gap. The analysis suggests that there currently is a small positive output gap.

B. Estimating Long-Term Growth, Production Function Approach

6. The production function (PF) method is based on the standard neoclassical growth model. Output is obtained after combining three inputs in a production function: labor, physical capital, and total factor productivity (TFP). Real GDP is defined by the familiar Cobb-Douglass growth accounting formula with constant returns to scale as: Yt=AtKt1αLtα

7. In the equation above Yt is the level of output, Lt is the level of employed labor, Kt is the level of physical capital, and At is the so-called Solow residual, which accounts for technological and human-capital factors. The parameter alpha (α) is the labor share of the economy. Output growth is broken down into contributions from capital (K), labor (L), and TFP (A). For the historical period 2006–17, we compute A as the residual of the growth accounting formula, while for the projection period 2018–50 we make assumptions about the growth rates of capital, labor, and TFP. The capital stock dynamics follow the standard accumulation process: Kt = Kt-1(1 – δ) + It, with It indicating the level of real investment at every period and δ the depreciation of the previous period’s capital stock. In the baseline model, the labor share (α) is set at the EU-average value 65 percent—following the discussion in D’Auria et el. (2010) and Bosworth and Collins (2003). We use a rate of 6 percent for capital depreciation (delta), and we assume the capital stock ratio to GDP (K/Y) in 2005 was 1.8, as provided by the authorities and validated by comparisons. We explore how sensitive the results are to the assumptions.

8. Growth over 2006–17 was 2.5 percent on average and 2.4 percent annualized. Due to the variability of growth the annualized growth figures are more representative, but there is little difference in practice. In the baseline model, capital contributed 2.3 percentage points on an annualized basis, labor 0.7 percentage points, and TFP subtracted 0.5 percentage points per year. The negative contribution of TFP is unusual and only at the 20th percentile of 10-year average TFP growth rates of emerging and advanced economies measured over 2005–14. It is mechanically mainly explained by the global financial crisis (GFC) and the Europe-related 2012 crisis. Abstracting from the crisis years 2009 and 2012, TFP growth was 0.6 on average, which is more in line with other countries at similar levels of development. However, the low historical TFP growth could also reflect the ongoing process of moving from a planned to a market economy, which renders some of the capital stock obsolete. Capital stock obsolescence above the assumed depreciation rate would reduce measured TFP because capital would be overestimated (see sensitivity analysis). However, it could also reflect more general productivity problems.

uA01fig02

Contributions to Growth – Baseline Scenario

(Percent)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Sources: Monstat; UN Population Prospects 2017; and IMF staff estimates.

Contributions to Growth – Baseline Scenario, 2007–17

article image
Sources: Monstat; and IMF staff estimates.

9. TFP is very sensitive to changes in model parameters. To study the sensitivity of historical TFP growth to underlying assumptions about the model parameters, we consider variations of the following parameters around the baseline: (i) the labor share; (ii) the depreciation rate; and (iii) the initial capital stock. The text table summarizes these parameter assumptions, and results for TFP are summarized further below. When the labor share goes up to 75 percent, the average TFP growth rate rises to 0.0 percent relative to −0.5 percent in baseline. A higher labor share reduces capital’s historical contribution and increases TFP and labor’s contribution. More depreciation increases measured TFP because the capital stock does not grow as fast, implying greater output/efficiency per unit of capital. If the depreciation of capital is assumed at 8 percent, the average TFP growth rate increases to −0.1 relative to −0.5 in the baseline. A larger initial stock of capital also increases measured TFP to 0.0 percent per year because capital accumulation is relatively smaller when the same amount of investment is added to a higher base. Depreciation and the initial capital stock only redistribute between capital accumulation and TFP because the contribution from employment is not changed. Interestingly, the historical contribution from labor only varies by 0.2 percentage points per year, whereas capital and TFP vary by 1.4 and 1.2 percentage points, respectively. In sections C, D, and E, we discuss how various factors contribute to Montenegro’s long-run growth outlook and suggest some policies that can boost those factors’ contribution.

Baseline Assumptions and Sensitivity Analysis

article image
Source: IMF staff estimates.

Sensitivity of Historical Growth, 2007–17

(Average)

article image
Sources: Monstat; and IMF staff estimates.

C. Employment Projections

10. Montenegro’s labor markets are rigid, and weak demographics and low labor force participation constrain growth. Nevertheless, Montenegro’s aging is a little less pronounced than in Southern Europe or Europe as a whole and in line with developments in Eastern Europe. Besides demographics, feeble labor force participation and high structural unemployment add to labor market rigidities (panel chart below). Montenegro’s youth unemployment is high relative to its peer countries, and unemployment spells last longer on average. Labor utilization has recently supported growth. In 2015–17, unemployment and labor participation improved, but long-term unemployment remained high (estimated at about 10 percentage points) and the labor force participation rate for the 15+ year-old population remained relatively low at around 50 percent. Montenegro’s low participation rate is partly explained by high remittances and low female participation.

uA01fig03

Labor Force Participation Rate

(Percent)/1

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

1/ 15+ population data used for labor force participation rate.2/ 2015 data used for 2016 BLR data.Sources: International Labour Organization; and IMF staff calculations.
uA01fig04

Unemployment Rate by Duration

(Percent)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Sources; Monstat, Labor Force Survey; and IMF staff calculations.

11. Montenegro’s population is projected to decline significantly over the medium term with likely negative effects on growth. The population fell during 1995–2002, since the end of the socialist regime, mainly due to emigration. The pace of emigration has slowed significantly, but continued emigration together with the aging of the working population and declining birth rates will contribute to a shrinking working-age population. The United Nations projects that Montenegro’s population will decline from an estimated 629 thousand inhabitants in 2017 to 588 thousand by 2050 in the baseline medium fertility variant (a decline of almost 7 percent). Even in the high fertility variant, the population would increase only marginally to 653 thousand. Assuming no net emigration between 2016–50, would limit the decline in population from 588 thousand to 610 thousand (3 percent decrease from 2017).

uA01fig05

Montenegro: UN Total Population Forecast

(Thousands)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

1/ Median variant of UN Population Prospects 2017.Source: UN Population Prospects 2017.

12. As a result of population aging, the working population is projected to decline even faster than the overall population. In the baseline, the working age population is projected to decline from 421 thousand to 352 thousand, a decline of 16 percent. A change in fertility has a delayed impact on the working-age population because the newly born enter the working-age population only 15 years from now and would raise/lower the projected number for 2050 by some +/- 30 thousand (+/- 9 percent). The zero-migration scenario has a greater initial impact on the working-age population because it is mostly people of working age who leave.

13. In the baseline projections, we assume that employment grows broadly at the rate of the working-age population (15–65-year-old cohort). We also assume that the labor force participation (LFP) ratio relative to the 15–65-year-old cohort will increase by 0.2 percentage points per year, reflecting increased female labor market participation and increases in the effective pension age. We also assume that the unemployment rate improves by 0.2 percentage points per year reflecting a gradual reduction in structural unemployment from its high current level. In the baseline, which could be considered somewhat optimistic because of the labor force participation rate and unemployment assumptions, employment would stay essentially stable in the 225–230 thousand range through 2050. In a no-change scenario with constant labor force participation and unemployment rates, employment would fall from 229 thousand in 2017 to 192 thousand (16 percent decline).

uA01fig06

Montenegro: UN Working Age Population Forecast

(Thousands; 15–64)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

1/ Median variant of UN Population Prospects 2017.Source: UN Population Prospects 2017.
uA01fig07

LFP and Unemployment Rates – Baseline Scenario

(In percent)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Sources: UN Population Prospects 2017; and IMF staff estimates.

14. The employment scenarios use various assumptions for labor force participation, unemployment, fertility, and migration, which are factors the authorities can influence to some extent. A 0.25 percentage point per year increase/decrease in the labor force participation rate would raise/lower employment by some 25 thousand by 2050 (+/- 11 percent). A 0.25 percentage point per year decrease/increase in the unemployment rate would raise/lower employment by some 20 thousand by 2050 (+/- 9 percent). Using the high/low fertility rate variants of the UN population projection would raise/lower employment by some 20 thousand by 2050 (+/- 9 percent). Finally, the zero-migration variant of the UN population projection would raise employment by 9 thousand by 2050 by itself (4 percent) and some 20 thousand in a combined scenario. In a high-growth upside scenario, which includes a 0.25 percentage point per year increase in labor force participation, a 0.25 percentage point per year decrease in the unemployment rate, the UN high fertility rate, and no net outward migration, employment would increase by 87 thousand to 316 by 2050 (38 percent). In the worse-case scenario, combining low LFP, high unemployment and low fertility employment would fall by 63 thousand to 160 thousand by 2050 (28 percent).

uA01fig08

Employment, 2006–50

(Thousands)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Sources: Monstat; UN Population Prospects 2017; and IMF staff projections.

15. Labor market policies can expand labor supply and lift potential growth.4 To increase the level of employment, we recommend the following labor market policies: (1) make hiring and firing processes more flexible while preserving adequate employee protection; (2) reduce the effective labor tax wedge; (3) support policies that increase female participation (parental leave, child care); (4) increase the effective pension age through pension reform; (5) reduce the skill mismatch by improving the education system; and (6) discourage emigration and encourage immigration by improving domestic opportunities.

D. Growth of the Capital Stock

16. The initial capital stock and the depreciation rate are difficult to estimate. Two publicly available estimates for Montenegro are the Penn World Tables (PWT, version 9.0) and the IMF Fiscal Affairs Department’s (FAD’s) Investment and Capital Stock Dataset. Both databases have starting dates for the capital stock that are earlier than reliable GDP estimates, which only start in 2006. The PWT estimate for the capital-to-GDP ratio in 2006 is 2.3 calculated in constant national prices (2.8 in current US$ PPP terms) whereas the FAD database shows 1.2. Most of the difference is explained by different depreciation rates. The PWT implicit depreciation rate is 4 percent on average, whereas the FAD database implicitly uses 6.8 percent on average. The authorities estimate that the capital-to-GDP ratio was 1.8 in 2005, and they assume a depreciation rate of 5 percent. There is some evidence that depreciation rates are increasing over time because the capital stock today consists more of short-lived IT investments and because product life cycles are declining.5 We chose a depreciation rate of 6 percent per year, which implies a half-life for the average capital good of 11 years versus 17 years for 4 percent depreciation and 8 years for 8 percent depreciation. We choose a ratio of 1.8 for the capital stock in 2005 because it provides a reasonable evolution of the capital stock and matches the authorities’ numbers.

uA01fig09

Real Capital Stock by Source, 2006–17

(Ratio to GDP)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Sources: Penn World Table (PWT), version 9.0; Monstat; FAD Investment and Capital Stock Dataset (ICSD); and IMF staff estimates.
uA01fig10

Real Capital Stock, 2006–17

(Ratio to GDP)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Sources: Statistical Office of Montenegro; and IMF staff estimates.

17. Baseline projections of the capital stock are based on staff’s macro-framework investment projections until 2023 and the average historical investment-to-GDP ratio (2006–17) for the long-term. In the baseline case, the investment ratio post-2023 is set at 28 percent, which is a relatively high in absolute terms but within the highest and lowest 5-year rolling averages between 2006–17 (32 percent and 22 percent respectively).6 We also consider high and low depreciation rates and high and low initial capital ratios but using the same real investment path as in the baseline. We also use a high investment case with 2 percentage points of GDP higher investment over the projection period and a low case with 2 percentage points lower investment.

uA01fig11

Straight Line Depreciation Rates

(Index)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Source: IMF staff calculations,

18. The scenarios with high/low depreciation rates and high/low initial capital stocks should be considered different models of the economy because they are not policy variables. The initial capital stock obviously is not a variable that can be altered by policy choices, and the depreciation rate probably is also not responsive to policy changes. The variations in these parameters assume that the model of the economy is different both for the past and the future. A high initial capital stock of 2.8 times GDP in 2005 (22 percent higher than the baseline) results in in a capital stock growth that is only 0.2 percent lower than the baseline over the projection period (due to higher depreciation of a larger stock) (and vice-versa), suggesting that the initial capital stock assumption is not very crucial for capital accumulation. This can also be seen by the fact that both the high and low initial capital stocks converge with the baseline over the projection period. By contrast, a 2 percentage point higher/lower depreciation rate results in a capital stock growth rate that is 0.34/0.67 percent lower/higher than in the baseline, suggesting that the depreciation rate assumption is important. The only policy variable for capital accumulation is an increase in the investment ratio, where a 2 percentage point increase/decrease results in a capital accumulation rate that is 0.2/0.2 percent higher/lower than the baseline over 2018–50, but with more of an impact at the beginning of the period. The projections of the capital stock show more of a differentiation than the growth rates. The biggest difference occurs for the scenarios with a higher/lower depreciation rate.

uA01fig12

Real Capital Stock Growth, 2015–50

(Percent)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Sources: Statistical Office of Montenegro; and IMF staff estimates.
uA01fig13

Real Capital Stock, 2015–50

(Billions of 2006 euro)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Sources: Monstat; and IMF staff estimates and projections.

19. Domestic savings may not be sufficient to finance investment over the long term. Montenegro’s private savings is well below the average of its Western Balkan peers but has been on an upward trend since 2008, on the back of higher remittances, growth rates and income. Currently, a large part of investment in Montenegro is financed by the government, which temporarily endangered debt sustainability and cannot continue at the high present level. The gap between investment needs and total saving ratios thus raises concerns for the sustainability of future investment and should be solved through an increase in domestic savings and be complemented by better prospects for domestic investment.

uA01fig14

Private Saving

(Percent of GDP)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Sources: IMF, WEO database; and IMF staff estimates.
uA01fig15

Gross Domestic Savings

(Percent of GDP)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Source: IMF staff estimates.

20. Appropriate policies can improve the capital accumulation channel for growth. The following policies are recommended: (1) create conditions conducive for domestic or FDI projects with significant impacts on economic activity; (2) increase domestic savings; (3) create fiscal space to maintain/increase public investment; (4) improve public investment selection and management frameworks; and (5) strengthen financial markets to facilitate domestic and foreign investment.

E. Total Factor Productivity Projections

21. Montenegro’s historical productivity growth has been slow. Productivity has been low and its growth weak due to structural and institutional reasons. Institutional obstacles prevent the diffusion and efficient use of available technologies (e.g. high risks or adverse business climate that discourage FDI).7 Moreover, structural features of Montenegro, such as the high share of tourism and other labor-intensive industries, may have contributed to low productivity growth.

22. In the baseline model, we assume zero contribution from TFP over the projection period. Our baseline assumption is based on the historical experience, as explained in section B, but adds 0.5 percentage point per year to reflect higher future TFP growth. Also, TFP growth has been positive under most model specifications since 2015, so this would be in line with the recent past. There is no satisfactory way to model TFP growth, since it is the residual in the data. Therefore, we also conduct sensitivity analysis for different model assumptions and for high and low TFP growth rates. The high-growth scenario assumes 0.5 percent TFP growth, which is more in line with the experience of other emerging market economies. The latest median 10-year trailing TFP growth rate for emerging markets was 0.5 percent for the 2005–14 period. The low-growth scenario assumes minus 0.5 percentage points contribution from TFP growth, in line with the historical baseline estimation and around the current 20th percentile for emerging markets. The text-chart illustrates the level of TFP under different scenarios/models. Note that the TFP contribution projections for all the other models also add 0.5 percentage points to their historical averages, which are the base for the projections.8 That is, the TFP for the high (low) labor share, low (high) initial capital stock, and high(low) depreciation scenarios is set at the historical TFP growth for that scenario plus 0.5 percent, the difference between the baseline historical average TFP contribution and zero. Thus, all models get the same expected TFP boost for the projection period as the baseline scenario. As we will see below, the differences in the historical TFP averages across models drive the overall model-based growth projections much more than the other assumptions.

uA01fig16

Emerging and Developing Countries: TFP Growth

(1960–2014, 10-year trailing average)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Sources; Penn World Table, version 9.0; and IMF staff calculations.
uA01fig17

TFP, 2006–50

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Sources: Statistical Office of Montenegro; and IMF staff estimates.

Average TFP Growth, 2007–17

article image
Source: IMF staff estimates.

23. Equalizing the TFP contribution across models results in smaller differences in projection outcomes. Using the same TFP assumption across models answers the question of what would happen if, for example, the labor share in the economy was 65 percent in the past, but then decreases to 55 percent for the projection period. This is somewhat unrealistic but could happen for the labor share, where there is some evidence that the labor share has been decreasing over time. In addition, for the depreciation rate it is plausible that capital goods become obsolete faster than in the past. The realism of the different TFP assumptions will be discussed in the context of the overall growth projections.

24. Structural reforms could raise long-run productivity. Despite structural constraints, productivity is expected to increase as Montenegro progresses with the EU accession process, spurring important institutional improvements. Domestic reforms such as improvements to the credit registry and collateral registration will facilitate financial deepening and help allocate resources to more productive sectors. Furthermore, diversification into green and renewable sources of energy could reduce input costs. Montenegro has made steady progress in improving its business climate, although there are countries that have improved more. The enforcing of contracts and resolving insolvencies are areas where further improvements would be desirable.

uA01fig18

Doing Business Indicator

(Higher score means improvement)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Sources: World Bank, Doing Business 2018; and IMF staff calculations.
uA01fig19

Business Environment, Distance to Frontier

(100 represents the frontier)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Sources: World Bank, Doing Business 2018; and IMF staff calculations.

F. Long-Term Growth—Putting Together the Factors

25. Growth falls substantially in the baseline projection, reflecting a declining labor force and decreasing contributions from investment. We project potential output from 2018 to 2050 using a growth accounting framework. The average growth rate for 2018–23 is projected at 1.7 percent per year but would decline to only 0.1 percent per year in the 2041–50 decade. In the baseline scenario, employment will likely have a negative contribution on average, unless both labor force participation and unemployment improve significantly since the baseline already assumes a moderate improvement in both. The highway project will contribute to capital accumulation in the near term. However, in the absence of TFP growth, even the relatively projected high investment rates add increasingly less to a growing capital stock in the long run. In the absence of structural reform, the contribution from TFP is likely limited and probably the main bottleneck for long-run growth prospects in Montenegro. The baseline projections could be considered moderately optimistic because the scenario with no change in the labor force participation and unemployment rate projections yields considerably lower growth rates (see graph). On the other hand, one might also question the TFP contribution assumption, which is below the recent median of emerging and advanced countries.

uA01fig20

Contributions to Long-Term Growth – Baseline Scenario

(Percent)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Sources: Monstat; UN Population Prospects 2017; and IMF staff estimates.

Baseline Contributions to Growth, 2018–50

(Percent)

article image
Source: IMF staff estimates and projections.

26. Model uncertainties do make a difference, but they are not very large initially. Among the different model assumptions (depreciation, labor share, and initial capital stock), the initial capital stock has the largest impact long-term impact and the depreciation has the smallest impact. The impact of the model assumptions increases over time because employment and capital growth decline in the baseline, and thus the different TFP assumptions for these different models dominate. For the medium-term 2018–23 projections, the worst model assumptions combined (low depreciation rate, low labor share, and low initial capital stock) compared to the best model assumptions only results in a difference of 1.1 percentage points per year on average, but the difference increases to 2.5 percentage points on average for 2041–50. The sensitivity analysis implies that the maximum model average growth rate over the 2018–50 projection period is 1.5 percent. This would assume that the economy operates with a high depreciation rate, a higher labor share, and a high initial capital stock. Given these assumptions, the projected factor accumulation results in a higher growth rate. However, most of this is driven by TFP, which would contribute almost 1 percentage point per year on average in this model.

uA01fig21

Potential Output Growth, Model Uncertainty Scenario

(TFP = based on 2007–17 historical average, in percent)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Sources: Monstat; UN Population Prospects 2017; and IMF staff projections.

Model Uncertainty Contributions to Growth (Based on 2006–17 historical TFP)

(Percent)

article image
Source: IMF staff estimates and projections.

27. The TFP assumptions explain the counter-intuitive result that a high depreciation rate and a high initial capital stock results in the highest overall growth rate because the slower capital accumulation rates are more than offset by a higher TFP growth. Similarly, a higher labor share results in a higher growth rate despite employment’s negative contribution because the TFP effect dominates. In fact, the difference in growth rates between the high-growth model and the low-growth model of 2.2 percent over 2018–50 is almost as large as the difference in the TFP assumptions of 3.0 percentage points between these two models. To the extent that the estimates of the historical differences in TFP growth are overestimated, the differences in the model projections would also be overestimated. It appears likely that the estimated difference between the high and low growth model of 3.0 percentage points based on the 2006–17 data is too high. This is probably partly explained by the short data period for estimation.

28. Equalizing the TFP assumptions results in smaller and more intuitive model differences. If the TFP contributions were held at zero for all the models as in the baseline, the results answer the following question: “suppose the economy had a depreciation rate of 6 percent in the past, but now increases to 8 percent, what would be the effect of this change?” Under this assumption, the effect on the projections would be smaller overall and reverse the order. The depreciation rate would have the highest impact and the initial capital stock would have the smallest impact, which is intuitive.9 Also, the impact would decline over time because differences in capital stocks would decline. The difference in the model assumptions would decline from 1.6 percentage points in 2018–23 to 0.5 percentage points in 2041–50.

uA01fig22

Potential Output Growth, Model Uncertainty Scenario

(TFP = Zero, in percent)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Sources: Monstat; UN Population Prospects 2017, and IMF staff projections.

Model Uncertainty Contributions to Growth (Zero TFP growth)

(Percent)

article image
Source: IMF staff estimates and projections.

29. Policy changes can result in important changes in income over time. The best combination of policy levers could raise average growth over the projection period from 0.7 percent to 2 percent per year, which is significant. The policy levers available to the government are mostly related to TFP (including human capital), labor force participation, unemployment, fertility, migration, and investment. The government has little leeway to influence the depreciation rate or the labor share. The most important policy lever is productivity growth, which also includes improvements in human capital. The projections assume scope for upside of 0.5 percentage points per year, but in principle higher improvements are possible and have been achieved in other countries.10 The other two most important policy levers are increasing labor force participation and reducing the unemployment rate. A ¼ percentage point increase in the labor force participation rate per year would increase growth by 0.22 percentage points per year on average with a slightly higher initial impact. Similarly, a ¼ percentage point reduction in the unemployment rate would increase growth by 0.17 percentage points per year, again with a higher initial impact. Increasing the fertility rate to the level of the UN high fertility variant would also have an average impact of 0.17 percentage points per year, but would only take effect in 2031–50. Reducing net emigration to zero would only have an effect of 0.09 percentage points per year, which would be fairly evenly distributed across time. Finally, increasing the investment ratio by 2 percentage points per year over the projection period would only raise growth by 0.10 percent per year, but this is likely underestimated because a permanently higher investment ratio would probably endogenously increase TFP. In that sense, the short-term impact, which is 0.2 percentage points over the 2018–23 period, is probably more representative.

uA01fig23

Potential Output Growth, Policy-Based Scenario

(Percent)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Sources: Monstat; UN Population Prospects 2017, and IMF staff projections.

Policy Contributions to Growth

(Percent)

article image
Source: IMF staff estimates and projections.

30. Policy levers likely have a more important role than model uncertainty. The policy levers we consider could improve growth by 1¼ percentage points or more per-year over the projection period. The projected impact of policy levers is probably underestimated because higher growth rates would induce higher investment and capital accumulation, TFP growth, and immigration, which is not included in the projections. By contrast, model uncertainty only adds between 0.5–0.8 percentage points on the upside and somewhat wider range on the downside (0.3–1.2 percentage points). In addition, the different model assumptions are unlikely to be truly additive since the difference in the parameter assumptions is relatively large (for example, for the depreciation rate) and the probability that all three model assumptions are in the worst or best case is probably quite low. By contrast, the policy lever assumptions are relatively reasonable and could easily go together. For example, high investment and TFP growth often go together and could very well be consistent with lower unemployment, higher labor force participation, and lower emigration.

uA01fig24

Potential Output Growth, Policies and Uncertainty

(Percent)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Sources: Monstat; UN Population Prospects 2017; and IMF staff projections.

31. In the baseline, Montenegro would likely not converge with the rest of EU but could do so with significant policy efforts. The convergence hypothesis states that economies with lower per capita GDP should converge to higher income levels due to decreasing returns on capital—assuming similar technologies, saving, demographic features, and human capital. EU accession, adoption of new technologies and enhanced institutions, higher FDI, and increased overall efficiency could help to close income gaps, especially if they are leveraged to achieve higher productivity growth in Montenegro. Montenegro’s legal, judiciary, and regulatory institutions are still catching up with EU standards. Per-capita growth in the baseline is modest, but it would increase to almost 2 percent on average in the best policy-based scenario, which would help reach EU level incomes over time. Real GDP per-worker would grow less because the best policy scenarios include higher fertility and no net immigration, which lowers the per-worker output.

uA01fig25

Selected CESEE: Expected Convergence

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Source: IMF staff estimates.
uA01fig26

Per Capita Output Growth 2015–50

(Percent)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Sources: Monstat; UN Population Prospects 2017; and IMF staff projections.
uA01fig27

Per Worker Output Growth

(Percent)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Sources: Monstat; UN Population Prospects 2017; and IMF staff projections.

G. Potential Output and the Output Gap Using HP Filter Approach

32. Univariate filters’ simplicity for estimating potential GDP and the output gap make them appealing, but they are not free of pitfalls. Unlike structural methods and multivariate filtering approaches, univariate filtering approaches like the Hodrick-Prescott (HP) filter simply assume that potential output is a smoothed trend around actual output. These approaches are easy to interpret and communicate, but they have some conceptual shortcomings. By construction, HP filters estimate a smooth and relatively stable trend with symmetric cyclical deviations from the actual data. These cyclical deviations are, on average, relatively small and corrected relatively quickly depending on the degree of smoothing. They also suffer from the “end-point problem,” which implies excessive sensitivity to the final observations. Essentially the filters assume that the beginning and end-point variations are mainly changes to the trend and not cyclical variations. The end-point limitation results in frequent revisions to historical output gap estimates, with the greatest degree of uncertainty reserved for the last and most needed data point for policy formulation.

uA01fig29

Potential Output Gaps

(Percent of potential GDP)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Source: IMF staff estimates.
uA01fig28

Actual and Potential Growth Rates

(Percent)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Source: IMF staff estimates.

33. We estimate variations of one-sided and two-sided HP filters using historical real GDP data (2006–17) and the macro-economic framework projection path (2018–23). The smoothing parameters are set at 100 and 6.25, reflecting the discussion in Ravn and Uhlig (2002). A lower value of lambda provides a more rapid adjustment of the estimated potential output with respect to the actual data. Our results suggest that lambda=6.25 fits Montenegro’s growth experience better, partially reflecting the relatively short available time series and thus greater uncertainty about the state of the economy. The main difference between one-sided and two-sided is that one-sided filters only use past data, while two sided filters use projected values as well. Onesided filters can be used for back testing and for real-time analysis (what is the current filtered value). Two-sided (or centered) filters are better-equipped for post-facto analysis, given that at any point in time they require future data. In our view, the two-sided HP filter with lambda=6.25 (HP2–6.25) seems to produce a more plausible result for the latest observation, 2017, and for the projection period 2018–23, illustrated in the chart above. The HP2–6.25 filter captures the periods of growth downturns well. It possibly underestimates the booms, which is mainly due to the “fast corrections” assumption described above. The output gap in 2017 and 2018 is estimated to be small and positive, reflecting the impact of the highway project. The output gap is closed at the end of our projection period, 2023.

uA01fig30

Nominal GDP and Nominal Potential

(LHS: million of euros, RHS: percent)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Source: IMF staff estimates.

34. The comparison between the baseline PF approach and the HP filter approach suggests that the framework assumptions may be somewhat optimistic. The difference between the real GDP growth rates over 2018–23 in the macro-economic framework and the ones implied by the baseline PF projections is explained by the TFP contribution, because the assumptions for employment growth and investment are the same. The text table below illustrates that the TFP contribution in the framework is larger compared to the TFP contribution in the baseline PF method. The difference arises mostly for the 2020–23, a period in which the contribution from TFP in the framework amounts to slightly more than 1 percentage points of GDP more than in the baseline PF approach. This could be consistent with the highway supply effect, which pushes up growth in the years after the highway investment is completed, and could be reflected in increased TFP during the years the supply impact takes effect. However, this effect should at least be partially reflected in a higher private investment ratio, which is the case, but it could also imply that the macro-economic framework assumptions are somewhat optimistic.

Feasibility of Framework Projections

(Percent)

article image
Source: IMF staff estimates and projections.

35. The production function method could also be used to derive an estimate of potential GDP and the output gap. Estimated TFP includes both underlying TFP growth and cyclical variations. The underlying TFP growth is likely relatively smooth. Smoothed TFP could be used to construct a potential output series and the difference between this potential GDP and actual GDP would represent the output gap. Therefore, subtracting the annualized TFP growth or a trailing moving average from the estimated TFP provides an estimate of growth contribution from the change in the output gap (see graph). We constructed a potential output series using the annualized growth in TFP from 2006–23, which results in plausible behavior through 2017 but diverging values for 2018–23. The latter is explained by the lower TFP growth in our baseline PF method and thus produces increasing output gaps for 2018–23. The output gap in 2017 is almost identical in this methodology, which gives some comfort. However, the diverging overall results suggests that estimates of the output gap most likely must include substantial elements of judgement.

uA01fig31

Contribution to Growth

(Percent)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Source: Monstat; and IMF staff estimates and projections.
uA01fig32

Potential and Actual Real GDP

(In 2006 Euros)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Source: Monstat; and IMF staff estimates and projections.
uA01fig33

Real GDP Growth and Output Gap

(Percent)

Citation: IMF Staff Country Reports 2018, 122; 10.5089/9781484357026.002.A001

Source: Monstat; and IMF staff estimates and projections.

H. Conclusion

36. We estimate long-run growth in Montenegro using a production function approach using employment, capital, and total factor productivity (TFP) growth and discuss possible drivers of growth. Historical growth was driven mostly by capital with some contribution from labor, while TFP contributed negatively. Going forward, in the baseline growth accounting framework with no reforms, employment will likely have a slightly negative contribution because of demographic dynamics unless both labor force participation and unemployment improve significantly. The highway project will contribute to capital accumulation in the near term, but the contribution from capital accumulation will likely fall despite relatively high investment ratios. Based on historical performance, the contribution from TFP is likely limited and constitutes the main bottleneck for long-run growth prospects in the no-reform baseline. Thus, structural reform efforts to increase TFP through human capital accumulation and improving the efficiency of the economy will be key for preventing growth rates from falling towards zero. The policy-based scenarios show that growth rates could increase by 2 percentage points per year in a high growth scenario and perhaps more if this high growth scenario leads to immigration and other positive feedback loops. We use a variant of the HP filter (two-sided HP approach) to produce an estimate of potential GDP and the output gap. The analysis shows that the output gap in both 2017 and 2018 is positive but small at less than 1 percent of GDP.

References

  • Acemoglu, A. and Johnson S., 2005Unbundling InstitutionsJournal of Political Economy, 113, 949995.

  • Alberola, E., Estrada, A., and D. Santabárbara, 2014, “Growth and Imbalances in Spain: A Reassessment of the Output Gap,” Journal of the Spanish Economic Association, Vol. 5(2–3), p. 333356.

    • 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 Papers 15/79 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Benes, J., Johnson M., Clinton K., Matheson T., Manchev P., Garcia-Saltos R., and Laxton D., 2010, “Estimating Potential Output with a Multivariate Filter,” IMF Working Paper 10/285

    • Search Google Scholar
    • Export Citation
  • Berger, H., Dowling, T., Lanau, S., Lian, W., Mrkaic, M., Rabanal, P., and Taheri Sanjani T., 2015, “Steady as She Goes–Estimating Potential Output During Financial “Booms and Busts”, IMF Working Paper 15/233 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Borio, C., Piti D., and Mikael K., 2013, “Rethinking Potential Output: Embedding Information About the Financial Cycle,” BIS Working Paper No. 404.

    • Search Google Scholar
    • Export Citation
  • Böwer, U. and Turrini, A., 2009. “EU accession: A road to fast-track convergence?”, European Economy – Economic Papers 2008 – 2015 393, Directorate General Economic and Financial Affairs (DG ECFIN), European Commission.

    • Search Google Scholar
    • Export Citation
  • Bosworth, B. P., and S.M. Collins, 2003. “The Empirics of Growth: An Update,” Brookings Papers on Economic Activity, Vol. 34(2003–2), pages 113206.

    • Search Google Scholar
    • Export Citation
  • Campos, N., Coricelli, F., and Moretti, L., 2014. “Economic Growth and Political Integration: Estimating the Benefits from Membership in the European Union Using the Synthetic Counterfactuals Method,” IZA Discussion Papers 8162, Institute for the Study of Labor (IZA).

    • Search Google Scholar
    • Export Citation
  • Christiano, Lawrence J, and Terry J Fitzgerald, 2003, “The Band Pass Filter,” International Economic Review, Vol. 44(2), p. 435465.

    • Search Google Scholar
    • Export Citation
  • D’Auria, F., Denis, C., Havik, K., Mc Morrow, K., Planas, C., Raciborski, R., Werner, R. and Rossi, A., 2010, “The Production Function Methodology for Calculating Potential Growth and Output gapsEuropean Economy – Economic Papers 2008 – 2015 420, Directorate General Economic and Financial Affairs (DG ECFIN), European Commission.

    • Search Google Scholar
    • Export Citation
  • Dobrinsky, R. and Havlik, P., 2014. “Economic Convergence and Structural Change: The Role of Transition and EU Accession,” wiiw Research Reports 395, The Vienna Institute for International Economic Studies, wiiw.

    • Search Google Scholar
    • Export Citation
  • Fernald, J., 2014, “Productivity and Potential Output before, During, and after the Great Recession,” NBER Working Paper No. 20248.

  • Havik, Karel, Kieran Mc Morrow, Fabrice Orlandi, Christophe Planas, Rafal Raciborski, Werner Röger, Alessandro Rossi, Anna Thum-Thysen, Valerie Vandermeulen., 2014, “The Production Function Methodology for Calculating Potential Growth and Output GapsEuropean Economy – Economic Papers 2008 – 2015 535, Directorate General Economic and Financial Affairs (DG ECFIN), European Commission.

    • Search Google Scholar
    • Export Citation
  • IMF, World Economic Outlook, April 2017, Chapter 2, “Road Less Traveled: Growth in Emerging Markets and Developing Economies in a Complicated External Environment”.

    • Search Google Scholar
    • Export Citation
  • IMF, “The Western Balkans: 15 years of Economic Transition”, Regional Economic Issues Special Report, March 2015.

  • IMF, Regional Economic Issues (REI): Central, Eastern, and Southeastern Europe, November 2016, “Effective Government for Stronger Growth”, Washington, DC.

    • Search Google Scholar
    • Export Citation
  • IMF, “Potential Output in Albania”, Selected Issues Paper, 2016.

  • IMF (2017), “FAD Investment and Capital Stock Database 2017: Manual & FAQ – Estimating Public, Private, And PPP Capital Stocks.” Washington, DC.

    • Search Google Scholar
    • Export Citation
  • Hausmann, R., Pritchett, L., and Rodrik, R., 2005. “Growth Accelerations.” Journal of Economic Growth 10 (4): 30329.

  • Lenain, P., and Rawdanowicz, L., 2004. “Enhancing Income Convergence in Central Europe after EU Accession,” OECD Economics Department Working Papers 392, OECD Publishing.

    • Search Google Scholar
    • Export Citation
  • Martin, C., Velazquez, F., and Funck, B., 2001. “European Integration and Income Convergence: Lessons for Central and Eastern European Countries,” World Bank Publications, The World Bank, number 13968, December.

    • Search Google Scholar
    • Export Citation
  • Okun, A., 1962, “Potential GNP: Its Measurement and Significance,” Cowles Foundation Paper 190.

  • Podpiera, J, Raei, F, and Stepanyan, A, 2017, “A Fresh Look at Potential Output in Central, Eastern, and Southeastern European Countries,” IMF Working Paper 17/37 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Rabanal, P., and Taheri Sanjani, M., 2015, “Financial Factors: Implications for Output Gaps,” IMF Working Paper 15/153 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Rapacki, R. and Próchniak, M., 2009. “The EU enlargement and economic growth in the CEE new member countries,” European Economy – Economic Papers 2008 – 2015 367, Directorate General Economic and Financial Affairs (DG ECFIN), European Commission.

    • Search Google Scholar
    • Export Citation
  • Ravn, M., and H. Uhlig, 2002, “On Adjusting the Hodrick-Prescott Filter for the Frequency of Observations,” The Review of Economics and Statistics, Vol. 84(2), pp. 371375.

    • Search Google Scholar
    • Export Citation
  • Svirydzenka, K. and Petri, M., 2014, “Mauritius: The Drivers of Growth—Can the Past be Extended?”, IMF working paper WP/14/134.

1

Prepared by Marzie Taheri Sanjani and Min Song.

2

There is some pre-2006 GDP data, but it is based on a different, less reliable methodology and data. The current methodology was first applied in 2006. For the purpose of this project, and in line with the authorities, we focus on post-2006 data for which the methodology is consistent with the ESA 2010 and NACE 2 standards.

3

See 2017 Montenegro – Selected Issues Paper on the impact of the highway, which is likely to have lower economic returns than standard investment projects.

4

Refer to chapter on labor market issues for details.

5

IMF (2017), “FAD Investment and Capital Stock Database 2017: Manual & FAQ – Estimating Public, Private, And PPP Capital Stocks.”

6

We are using the real investment-to-GDP ratio calculated from real GDP components. The real ratio is higher than the nominal investment-to-GDP ratio due to investment deflator issues. In 2017, the nominal ratio was 30 percent relative to a real ratio of 33 percent.

7

The Fall 2016 European Regional Economic Issues Paper “Effective Government for Stronger Growth” provides some analysis on these factors.

8

The high/low TFP growth assumptions are a policy scenario. Varying labor share, the initial capital stock and the depreciation rates are different models of the economy which assume that, for example, the labor share is different throughout the historical and projections periods.

9

Changing the initial capital stock retroactively makes little sense but would be equivalent to a natural disaster that destroys part of the capital stock (possible) or the economy inherited additional capital stock for free (not very realistic).

10

The most recent 10-year trailing TFP growth of the 75th percentile of emerging markets was 1.7 percentage points per year.

Montenegro: Selected Issues
Author: International Monetary Fund. European Dept.