![Sample our Economics, Finance,Business & Industry journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5931/TOC-Subject-Sample-2021-Economics-Finance-Business-Industry.jpg"})
ABSTRACT
We study the effect of a (standard) monetary policy shock in the euro area on the Lithuanian economy. We employ a structural vector autoregressive model incorporating variables from both the euro area and Lithuania. The model exhibits a block exogenous structure to account for the fact that Lithuania is a small economy. In general, we find that a monetary policy shock in the euro area has a stronger effect on the Lithuanian than it does on the euro area economy, though the effects are not statistically significant, preventing firm conclusions. We further broaden our analysis employing a panel vector autoregression (PVAR) model for the three Baltic states. PVAR model results suggest a stronger impact of monetary policy than that estimated using the Lithuanian model and a quite considerable degree of variation over time in the strength of monetary policy transmission.
KEYWORDS:
Acknowledgments
We would like to thank Lenno Uusküla, Rūta Rodzko, Sigitas Šiaudinis, Mihnea Constantinescu, Aurelijus Dabušinskas, Tomas Reichenbachas and the participants of the internal Bank of Lithuania seminars for their helpful comments and suggestions. The views expressed and the conclusions reached in this publication are those of the authors and do not necessarily represent those of the Bank of Lithuania or the Eurosystem.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes on contributors
Julius Stakėnas obtained his master in econometrics from Vilnius University, Lithuania. He is currently working as an economist in the Economics department of the Bank of Lithuania. His main interests lie in the issues of short-term forecasting: factor modelling, mixed-frequency data, forecast aggregation, global economy modelling.
Rasa Stasiukynaitė obtained her Ph.D. in Economics from Tilburg University, The Netherlands. Currently she works as an economist in the Economics department of the Bank of Lithuania. Her main interests are unconventional monetary policy transmission and policy stance assessment.
Notes
1. The assets of the financial system in Lithuania, by the end of 2014, constituted around 70% of GDP, while the same measure stood around 310% for the euro area. Source: ECB Statistical Data Warehouse and own calculations.
2. By the end of 2014, market capitalization in Lithuania was 9% of GDP, while in the euro area – around 59%. Source: ECB Statistical Data Warehouse and own calculations.
3. In 2014, the top three banks held 70% of the system's assets and over 70% of total loans, see Statistical Annex 2 of the Financial Stability Review, 2015, Lietuvos bankas.
4. Allowing the data to ‘speak’ can be perceived both as an advantage and disadvantage: without theoretical relations included in the model the outcomes might be counterintuitive, for example, a well-known ‘price puzzle’, and the results can be sensitive to various identification schemes (see, for example, Van Aarle, Garretsen, & Gobbin, Citation2003).
5. Estonia and Lithuania had currency boards, while Latvia – a quasi-currency board. The Estonian national currency was pegged to the DEM from 1992 and later to the euro from 1999 until the euro adoption in 2011; the Latvian currency was pegged to the euro from 2005 until the euro adoption in 2014.
6. In an earlier study, using the data for 20 industrialized countries, Georgiadis (Citation2014) also points to financial structure, labour market rigidities and industry mix being the main determinants for differences in monetary policy transmission.
7. Note that during the period Latvia was in a fixed exchange regime.
8. In 2014 the share of Lithuanian exports to the euro area countries constituted 29% and imports — 39%. Source: Eurostat and own calculations.
9. For brevity reasons and due to the fact that every VAR(p) model can be written as VAR(1), we write the model as VAR(1). We elaborate on the lag structure later in the text.
10. For detailed data description, together with the applied data transformations, see Table in Appendix 1.
11. Excluding the energy component from consumer prices helped to obtain more reasonable impulse responses to a monetary policy shock, whereas accounting for the oil price impact on HICP in the VAR model's framework proved to be a more difficult task. Nevertheless, we agree that when the Euribor equation in the euro area block is interpreted as the Taylor rule, it may be more realistic to condition the Euribor rate on the complete HICP, jointly with the commodity prices (and other indicators).
12. For the credit variable we use credit stock (nominal values). Although the data on new loans might be more relevant (it might react more to changes in the Euribor rate), due to shorter new loans data series we opted to use credit stock variable. We also tried to use credit deflated by consumer prices; however, it did not considerably change the results.
13. OLS estimates were sensitive to the inclusion of the Q1 2009 data point. We argue, that due to the various different factors (which are also difficult to account for) making this data point an outlier, it is reasonable to exclude this data point from the estimation of the Lithuanian block.
14. Credit margins are calculated as follows:
where loan interest contains average interest rates on new loans (both in litas and in euros) to non-financial institutions and households;
was the average 3-month interbank lending interest rate in litas.
15. In this regard our estimation strategy is reminiscent of the global VAR methodology (see Pesaran, Schuermann, & Weiner, Citation2004), which also uses individual country VAR estimates to build a joint model.
16. All estimations in the paper were carried out using R statistical package.
17. To illustrate this argument, we also present impulse response estimates for Lithuanian block specification in levels.
18. We use a longer sample for the euro area block because a model with levels requires longer time series. However, despite the fact that there is much more data available for the euro area block, we choose a period which is reasonable in a sense that we expect data within our chosen period to have the same structure.
19. The impulse response functions are plotted together with 68% confidence intervals. The confidence intervals are obtained using standard bootstrapping procedure with 10,000 draws.
20. Due to lack of data, we do not have the original data series for the credit margin in Latvia. However, due to largely the same main banks operating in the region (also experiencing similar economic environment), we might expect the credit margins to be very similar in the three Baltic states. Therefore, in the subsequent analysis, we will approximate the Latvian credit margin using the mean of the Lithuanian and Estonian credit margins.
21. The specific dummy variable treatment, for example, using the same dummy of Q1 2009 for all three countries, does not have a significant influence on the results.
22. Due to a fairly large T, the so-called Nickell bias should not be a cause of concern in this case. We thank the referee for pointing this out.
23. For PVAR shock identification, we applied the same identification scheme based on short-term zero restrictions as in Section 2.
24. Georgiadis (Citation2015) points to differences in the industry mix as one of the main factors in explaining asymmetries in monetary policy transmission: economies with larger share of aggregate output accounted for by manufacturing and construction, all else being equal, exhibit larger declines in GDP in case of monetary tightening. Note that the manufacturing share in Lithuania is the largest among the three Baltic states (Figure ), hence, the industry structure does not indicate that responses to monetary policy shock in Lithuania should be smaller than in other Baltic countries.
25. In our framework, the euro area variable responses to a 100 bp Euribor shock are the same for all rolling window regressions and are equal to the ones presented in Figure . Note that although this strategy helps to control for changes in monetary policy transmission in the euro area, we employed it due to the necessity to obtain reliable estimates. Panel treatment does not increase the number of observations for the EA model; therefore, it would suffer from small sample problems if it were estimated using the rolling window sample.
26. Barth and Ramey (Citation2002) highlight the supply-side channel of monetary policy, noting that contractionary monetary policy shock may have positive effect on prices due to rising cost of firms' working capital.