Abstract
This Selected Issues paper on the Republic of Kosovo’s 2013 Article IV Consultation highlights growth and Kosovo’s external environment. In the wake of the global financial crisis, Kosovo’s economic growth slowed but remained positive, while most other Western Balkans slipped into recession. Moreover, the annual average growth rate has been among the highest in the Western Balkans since the onset of the financial crisis in 2007. Kosovo’s tax-to-GDP ratio is comparable to the average of Southeastern Europe, although its tax system relies significantly more on indirect taxation—including a high share of trade taxes. Kosovo’s reliance on trade taxes may create budgetary pressures in the event of further trade liberalization.
Growth And Kosovo’s External Environment1
Robust remittances and FDI inflows from a Diaspora, hosted mainly in Germany and Switzerland, and limited financial and export linkages to crisis countries, account for much of Kosovo’s resilience to external turbulence. Kosovo’s dependence on—and vulnerability to—developments in host countries is confirmed by vector autoregressions. Econometric analysis shows that remittance transfers to Western Balkan countries have been driven mostly driven by host country conditions.
A. Recent Growth Performance in a Regional Context
1. Kosovo’s economy has performed well in the face of headwinds from the global financial crisis and euro area turbulence. In the wake of the global financial crisis, Kosovo’s economic growth slowed but remained positive, while most other Western Balkans slipped into recession. Moreover, the annual average growth rate has been among the highest in the Western Balkans since the onset of the financial crisis in 2007.
2. Simple cross-country growth correlations for 2001-2011 show that Kosovo’s economic activity is correlated with that of Germany and Switzerland—where about two-third of Kosovo’s Diaspora live—but uncorrelated with economic activity in the rest of the euro area.2 This pattern differs markedly from the other Western Balkan countries whose GDP growth is highly correlated not only with Germany, but also with activity in the rest of the euro area.
3. This chapter analyzes in more depth the link between activity in Kosovo and the external environment. Section B asseses spillovers with a vector auto regression model (VECM), linking Kosovo’s imports—a proxy for domestic demand—to external conditions. Section C looks in more detail at remittances and FDI from the Diaspora and their role for growth.
Real GDP growth in the Western Balkans
Citation: IMF Staff Country Reports 2013, 223; 10.5089/9781484307854.002.A001
Real GDP growth in the Western Balkans
Citation: IMF Staff Country Reports 2013, 223; 10.5089/9781484307854.002.A001
Real GDP growth in the Western Balkans
Citation: IMF Staff Country Reports 2013, 223; 10.5089/9781484307854.002.A001
Nominal GDP growth correlations with Europe
Citation: IMF Staff Country Reports 2013, 223; 10.5089/9781484307854.002.A001
Sources: World Economic Outlook and IMF staff calculations.Nominal GDP growth correlations with Europe
Citation: IMF Staff Country Reports 2013, 223; 10.5089/9781484307854.002.A001
Sources: World Economic Outlook and IMF staff calculations.Nominal GDP growth correlations with Europe
Citation: IMF Staff Country Reports 2013, 223; 10.5089/9781484307854.002.A001
Sources: World Economic Outlook and IMF staff calculations.B. Assessing Spillovers from External Shocks
4. Spillovers to Kosovo’s economic activity from shocks to Diaspora host countries are being assessed using a vector error correction (VECM) model for readily available, high-frequency indicators for the Kosovar economy (imports, private sector credit and bank deposits), and Germany’s GDP (as a proxy for economic conditions in Diaspora host countries). The VECM model restricts the long-run behavior of the endogenous variables to converge to their cointegrating relationship while allowing for short-run adjustment dynamics:3
where yt is a vector of endogenous variables, β is the cointegrating vector of coefficients, α is the coefficient vector measuring the speed of adjustment to equilibrium, Γ s are the coefficient matrices of the lagged Δyt and et is a vector of residuals. Unit root tests indicate that all variables (imports, German GDP, private sector credit and bank deposits) are integrated of order one. Furthermore, Johansen’s test does not reject the presence of at least one cointegrating relationship. The finding is supported by the Engel and Granger test, rejecting the null hypothesis that the residuals of the cointegrating vector have unit root. Under the (provisional) assumption of cointegration between the series, the importance of German GDP for Kosovo’s imports is confirmed by the Granger causality test.4
5. The results indicate that imports—used as a proxy for domestic demand—are linked in a long-term, dynamic equilibrium relationship with German GDP and bank deposits. The estimated long-run elasticity of imports to German GDP—which proxies Diaspora’s host country conditions—is positive and significant under both models: Johansen’s and the two-step model by Engel and Granger.5 Moreover, the long run coefficient on bank deposits is also significant and positive under both methodologies.6 Results also indicate that transmission of shocks to domestic demand is fairly rapid, as evident by the highly significant adjustment coefficient on the error-correction term in the equation for the dynamics of imports.
ECM Estimation Results for the dynamics of Imports 1/, 2/, 3/
Estimated for quarterly data over 2002Q4 – 2012Q4. All the variables are seasonally adjusted to remove seasonality present in quarterly data and expressed in logarithmic terms to remove non-linearilities.
All the variables are integrated of order one in levels and stationary in first differences.
See Johansen (1991), (1995) and Engle and Granger (1987).
Estimates vector autoregressions (VAR) and the EC vector simultaneously. ADF test is used to test for stationarity of the EC series.
ECM Estimation Results for the dynamics of Imports 1/, 2/, 3/
| Trace test (Johansen)/ADF test (Engel & Granger) statistics | Slope coefficients β, of the cointegrating vector | Speed of adjustment, α, coefficient | Lagged GDP growth of Germany | |||||
|---|---|---|---|---|---|---|---|---|
| Imports | DEU_GDP | Deposits | Trend | dlog(Imports) | dlog(DEU(-1)) | dlog(DEU(-4)) | ||
| Johansen (VECM) | 26.843** | 1 | -0.964** | - 0.79*** | Linear | -0.604** | 1.351* | 1.776** |
| Engle-Granger (two-step procedure) | -4.906*** | 1 | -1.422*** | -0.389** | Linear | -0.537** | 0.837* | 1.402* |
| System estimation 4/ | -3.181** | 1 | -0.875** | -0.682*** | No | -0.481*** | 1.285*** | 2.276*** |
Estimated for quarterly data over 2002Q4 – 2012Q4. All the variables are seasonally adjusted to remove seasonality present in quarterly data and expressed in logarithmic terms to remove non-linearilities.
All the variables are integrated of order one in levels and stationary in first differences.
See Johansen (1991), (1995) and Engle and Granger (1987).
Estimates vector autoregressions (VAR) and the EC vector simultaneously. ADF test is used to test for stationarity of the EC series.
ECM Estimation Results for the dynamics of Imports 1/, 2/, 3/
| Trace test (Johansen)/ADF test (Engel & Granger) statistics | Slope coefficients β, of the cointegrating vector | Speed of adjustment, α, coefficient | Lagged GDP growth of Germany | |||||
|---|---|---|---|---|---|---|---|---|
| Imports | DEU_GDP | Deposits | Trend | dlog(Imports) | dlog(DEU(-1)) | dlog(DEU(-4)) | ||
| Johansen (VECM) | 26.843** | 1 | -0.964** | - 0.79*** | Linear | -0.604** | 1.351* | 1.776** |
| Engle-Granger (two-step procedure) | -4.906*** | 1 | -1.422*** | -0.389** | Linear | -0.537** | 0.837* | 1.402* |
| System estimation 4/ | -3.181** | 1 | -0.875** | -0.682*** | No | -0.481*** | 1.285*** | 2.276*** |
Estimated for quarterly data over 2002Q4 – 2012Q4. All the variables are seasonally adjusted to remove seasonality present in quarterly data and expressed in logarithmic terms to remove non-linearilities.
All the variables are integrated of order one in levels and stationary in first differences.
See Johansen (1991), (1995) and Engle and Granger (1987).
Estimates vector autoregressions (VAR) and the EC vector simultaneously. ADF test is used to test for stationarity of the EC series.
6. The model’s impulse response functions confirm Kosovo’s dependence on—and vulnerability to—developments in host countries. A shock to quarterly growth in Germany affects Kosovo’s imports by a factor of about three over two years (imports are used as a proxy for domestic demand). The total cumulative impact on imports has two components: (i) the long-run error-correction term (αβ), which contributes by a factor of 0.6 to the next quarter’s import growth, and (ii) the short term dynamics of imports, which largely and positively depends on the lagged GDP growth in Germany (the Γ coefficients that contribute a factor of 1.35 in the second quarter and of 1.78 in the fifth quarter).
Spillovers from a shock to activity in Germany 1/
Citation: IMF Staff Country Reports 2013, 223; 10.5089/9781484307854.002.A001
Sources: Statistical Agency of Kosovo, IMF staff calculations.1/ Shows an impulse-response function tracing out the accumulated impact (in percent) of a one percent shock to the GDP growth rate of Germany on Kosovo’s imports (in logs).Spillovers from a shock to activity in Germany 1/
Citation: IMF Staff Country Reports 2013, 223; 10.5089/9781484307854.002.A001
Sources: Statistical Agency of Kosovo, IMF staff calculations.1/ Shows an impulse-response function tracing out the accumulated impact (in percent) of a one percent shock to the GDP growth rate of Germany on Kosovo’s imports (in logs).Spillovers from a shock to activity in Germany 1/
Citation: IMF Staff Country Reports 2013, 223; 10.5089/9781484307854.002.A001
Sources: Statistical Agency of Kosovo, IMF staff calculations.1/ Shows an impulse-response function tracing out the accumulated impact (in percent) of a one percent shock to the GDP growth rate of Germany on Kosovo’s imports (in logs).C. Determinants of Remittances in the Western Balkans
7. An understanding of the factors driving remittances is of critical importance for analyzing growth dynamics, given the role that remittances play in supporting incomes and economic activity in Kosovo. In this section, the elasticities of remittances to changes in host- and home country conditions are estimated, using panel data for 6 Balkan countries for 1999–2011. Specifically, remittances are regressed on measures of home and host country GDP per capita. Host country GDP is proxied by using euro area (EA) GDP, as Balkan emigrants are unevenly spread across the different EA countries. Germany was the main source for remittances to Croatia, Kosovo and to a lesser extent Bosnia and Herzegovina and Macedonia, while the majority of Serbia’s remittances originate from Austria. Remittances from the EA periphery (i.e. Greece and Italy) are high for Albania and Macedonia.
8. Host country per capita income plays a key role in driving remittances (as is often found in the literature). There is also evidence for an altruistic motive to remittances, as shown in the positive coefficient on the income gap variable (column 2) and the negative coefficient on home country GDP per capita (column 1). Emigrants’ remittances are thus stronger at times when home country income levels are lower. The results are maintained if one proxies for host country with Germany’s GDP per capita instead of EA GDP per capita.
Decomposition of Remittances by Source Country, 2011
(Percent)
Citation: IMF Staff Country Reports 2013, 223; 10.5089/9781484307854.002.A001
Sources: World Bank Bilateral Remittances Matrix and Central Bank of Kosovo.Decomposition of Remittances by Source Country, 2011
(Percent)
Citation: IMF Staff Country Reports 2013, 223; 10.5089/9781484307854.002.A001
Sources: World Bank Bilateral Remittances Matrix and Central Bank of Kosovo.Decomposition of Remittances by Source Country, 2011
(Percent)
Citation: IMF Staff Country Reports 2013, 223; 10.5089/9781484307854.002.A001
Sources: World Bank Bilateral Remittances Matrix and Central Bank of Kosovo.Determinants of Migrants’ Remittance Transfers in the Balkans, 1999-2011
*significant at 10 percent level; **significant at 5 percent level; ***significant at 1 percent level.
All variables are expressed in logarithmic terms. Regressions include constants and are estimated with robust standard erros. In equations including home country GDP on the right side, we use lagged values to attenuate reverse causality bias.
We control for home and host realinterest rates in regression (1) and for their differential (host-home) in regression(2) to capture the investment motive to remittances. The latter has the expected negative sign but is statistically insignificant.
Determinants of Migrants’ Remittance Transfers in the Balkans, 1999-2011
| Dependent variable: Remittances per Capitat | (1) | (2) |
|---|---|---|
| Home Country GDP per capita(t-1) | -0.49* | |
| Host Country GDP per capita | 2.67** | |
| Host-Home GDP per Capita diff(t-1) | 1.75** | |
| Observations | 68 | 68 |
| R-squared | 0.22 | 0.29 |
| Country FE | YES | YES |
*significant at 10 percent level; **significant at 5 percent level; ***significant at 1 percent level.
All variables are expressed in logarithmic terms. Regressions include constants and are estimated with robust standard erros. In equations including home country GDP on the right side, we use lagged values to attenuate reverse causality bias.
We control for home and host realinterest rates in regression (1) and for their differential (host-home) in regression(2) to capture the investment motive to remittances. The latter has the expected negative sign but is statistically insignificant.
Determinants of Migrants’ Remittance Transfers in the Balkans, 1999-2011
| Dependent variable: Remittances per Capitat | (1) | (2) |
|---|---|---|
| Home Country GDP per capita(t-1) | -0.49* | |
| Host Country GDP per capita | 2.67** | |
| Host-Home GDP per Capita diff(t-1) | 1.75** | |
| Observations | 68 | 68 |
| R-squared | 0.22 | 0.29 |
| Country FE | YES | YES |
*significant at 10 percent level; **significant at 5 percent level; ***significant at 1 percent level.
All variables are expressed in logarithmic terms. Regressions include constants and are estimated with robust standard erros. In equations including home country GDP on the right side, we use lagged values to attenuate reverse causality bias.
We control for home and host realinterest rates in regression (1) and for their differential (host-home) in regression(2) to capture the investment motive to remittances. The latter has the expected negative sign but is statistically insignificant.
9. In addition to remittances, there is evidence of complementarity between emigration and FDI.7 Several channels are at play. First, the presence of an educated Diaspora provides foreign investors with a much needed knowledge of local markets, which helps breaking barriers to long-term inward foreign investments. Second, unskilled migrants can also increase FDI by revealing workforce characteristics such as productivity of labor force and decreasing cross-border information costs and FDI-related country risk premium. Third, migration could also provide unskilled migrants with the necessary human and physical capital to invest in their home countries, an opportunity that would not be possible without migration. This seems to be the case for Kosovo: Germany, Switzerland and UK, which in 2011 accounted for over half of remittance transfers to Kosovo, have over the years contributed to about 40 percent of overall yearly FDI inflows since 2007.
Source countries of FDI
(Percent)
Citation: IMF Staff Country Reports 2013, 223; 10.5089/9781484307854.002.A001
Source: Central Bank of Kosovo.Source countries of FDI
(Percent)
Citation: IMF Staff Country Reports 2013, 223; 10.5089/9781484307854.002.A001
Source: Central Bank of Kosovo.Source countries of FDI
(Percent)
Citation: IMF Staff Country Reports 2013, 223; 10.5089/9781484307854.002.A001
Source: Central Bank of Kosovo.References
Engle, R. F., and C. W. J. Granger (1987): Co-integration and Error Correction: Representation, Estimation, and Testing, Econometrica, 55, 251-276.
Kugler, M., and H. Rapoport (2006): International Labor and Capital Flows: Complements or Substitutes, Economic Letters 94, 155-162.
Javorcik, B., C. Ozden, M. Spatareanu, C. Neagu (2010): Migrant Networks and Foreign Direct Investment, Journal of Development Economics 94, 231-241
Johansen, Søren (1991): Estimation and Hypothesis Testing of Cointegration Vectors in Gaussian Vector Autoregressive Models, Econometrica, 59, 1551–1580.
Johansen, Søren (1995): Likelihood-based Inference in Cointegrated Vector Autoregressive Models, Oxford: Oxford University Press
Prepared by Nina Budina, Ghada Fayad and Xingwei Hu.
Data are for the annual growth in nominal GDP, expressed in euro.
The cointegration term is known as the error correction term since the deviation from long-run equilibrium is corrected gradually through a series of partial short-run adjustments. All the variables are integrated of order one in levels and stationary in first differences. The choice of lag structure is based on lags exclusion test.
Granger-causality test, which provide preliminary insights into possible ‘causal’ links between the variables, strongly rejected the null hypothesis that German GDP does not granger cause Kosovo’s imports.
VECM results have been compared to the estimated single equation framework based on the two-step Engle-Granger (1987) methodology as a robustness check. Note that the null hypothesis of unitary elasticity of imports with respect to German GDP could not be rejected, suggesting that both methodologies yield very similar coefficients. VECM results have been also compared with the results of simultaneous autoregressions (VAR) and the EC vector system estimation. The system equation approach has added flexibility in that it can eliminate non-significant regressors.
This is also supported by the Granger causality test, which strongly rejects the null that imports do not Granger cause deposits.
See Javorcik and others (2010); Kugler and Rapoport (2007) for supportive empirical evidence.
