IMPACT OF MACROECONOMIC POLICY SHOCKS ON THE ARMENIAN ECONOMY: EVIDENCE FROM VAR ANALYSIS Текст научной статьи по специальности «Экономика и бизнес»

ECONOMIC SCIENCES

IMPACT OF MACROECONOMIC POLICY SHOCKS ON THE ARMENIAN ECONOMY:

EVIDENCE FROM VAR ANALYSIS

Grigoryan K.,

PhD in Economics, Associate Professor, Head of the Department of Macroeconomics, Department of Macroeconomics, Armenian State University of Economics, Yerevan, Armenia.

https://orcid.org/0000-0001-7359-3407 Avagyan G.,

PhD in Economics, Associate Professor, Department of Macroeconomics, Armenian State University of

Economics, Yerevan, Armenia. https://orcid.org/0000-0003-3395-2473 Karapetyan N.,

PhD student, Department of Finance, Armenian State University of Economics, Yerevan, Armenia.

https://orcid.org/0000-0003-4978-6107.

Mkhitaryan L.

PhD in Economics, Department of Macroeconomics, Armenian State University of Economics, Yerevan,

Armenia.

https://orcid.org/0000-0002-4386-1607

Abstract

Economic shock caused by the Covid-19 pandemic and sizable fiscal and monetary measures taken all over the world and in Armenia renews the questions regarding the effects and transmission mechanisms of macroeco-nomic policy measures. In this article we analyze the effects of monetary and fiscal policies in Armenia through a SVAR model with recursive identification. We demonstrate, that fiscal policy instruments are efficient for influencing on aggregate demand. We also demonstrate, that tax policy shocks have more persistent impact on output than public expenditures shock. The empirical evidence we collected suggests that exchange rate channel of monetary policy is effective, but the analyses did not reveal sound evidence how interest rate shocks affect aggregate demand. Moreover, we demonstrate, that when interest rate cuts are accompanied with intervention in foreign exchange market, the net effect of the policy measures on aggregate demand can even be contractionary.

Keywords: Fiscal and monetary measures, recursive identification, impulse responses, exchange rate channel, FX interventions.

JEL classification: E10, E44, E52, E60, E62,

H30

Introduction

The global economy is facing a synchronized, deep downturn in 2020 due to the economic shock caused by COVID-19 pandemic. The IMF projects global GDP growth at -4.9% as of June 2020, compared to 3.3% projected in January 2020. To counteract to shock, both advanced and developing countries heavily deploy countercyclical macroeconomic policies. Following the survey done by the IMF, the fiscal measures to counteract the downturn in global level reached to $11 trillion (the budget deficit on global level is expected to reach 10% of GDP), and both advanced and developing countries deployed sizeable monetary measures [1].

The GDP is expected to decline in Armenia also, but with modest rate compared to what is expected for the global economy. Armenian authorities are applying both fiscal and monetary policies against the downturn. The announced fiscal package by the RA Government sums 150 bln AMD, which is 2.3% of expected GDP for 2020 (the overall budget deficit is projected at 5.6% of GDP). On the monetary policy side, the Central Bank of Armenia eased monetary conditions significantly: the key rate of monetary policy has been cut by 1.0 percentage point as of end-June compared to end-February and short-term liquidity has been injected to fully meet

the demand of the banking system [2]. RA Central Bank also intervened in foreign exchange market to reduce the surged volatility of Armenian Dram - selling 127 mln USD in March and April and afterwards buying around 33 mln USD in April [3].

The situation faced by the policymakers renews question how the fiscal and monetary shocks are being transmitted in the economy and what are the final effects of these policy actions. Answering this question can be helpful for informing decision-making process and predicting the potential effects of the policy actions already undertaken in short and medium term. To answer these questions, we follow the methodology proposed by C. Sims in this article and build parsimonious VAR models with basic macroeconomic variables of RA economy.

First, we analyze the existing literature for Armenia on empirical analysis of macroeconomic policy effects. We will not try to cover the international literature related the issue, but we analyze the prominent paper of Cristopher Sims published in 1992, which is essential in the perspective of the methodology we apply. Second, we describe the methodology we follow and discuss the variables we choose for analysis. Third, we analyze the VAR impulse responses. And we conclude at the last section.

Literature Review

Sims (1992) used recursive identification strategy to analyze effects of monetary policy shocks in France, Germany, Japan, UK, and US, with monthly data spanning from 1957 to 1991. First, Sims used short-term interest rate, monetary aggregate M1, consumer price index and industrial production index. The analyses yielded a puzzling result: prices tended to rise and stay high for an enough long period after the contractionary surprise of short term interest rates (central bank policy rate), even though the monetary aggregate and output fell. The interpretation of this evidence, following to Sims, was that the central bank could know that inflationary pressures are about to arrive and counteract them by raising interest rates. This signaled that the VAR should be misspecified, and he added exchange rate and commodity price indexes in the VARs to account for the omitted variable. The new VARs were specified with the following sequence: short term interest rate, exchange rate, commodity prices, monetary aggregate and industrial production. With the new specification, after the surprise increase of the short-term interest rates commodity price index falls, M1 monetary aggregate falls, output falls, prices barely move and start decreasing after 1.5-2 years. These results overall are consistent across the countries studied. The response of the exchange rate is mixed across the countries studied: after the monetary contraction exchange rate was appreciating in Japan, US and UK (a result that is consistent with theoretical IS-LM framework) but was depreciating in France and Germany [4].

Era Dabla-Norris and Holger Floerkemeier (2006) were one of the first researchers who analyzed transmission mechanism of macroeconomic policies in Armenia with a VAR model. They build a reduced-form VAR with GDP, consumer price index, repo rate and the NEER and money supply as endogenous, and index of world oil prices and US Federal Funds Rate as exogenous variables. The empirical results indicated that the capability of monetary policy to influence economic activity and inflation were limited, as interest rate channel is weak and other channels (except for exchange rate channel) were estimated as non-effective [5].

Anna Rose Bordon and Anke Weber (2010) updated the analysis of Dabla-Norris and Floerkemeier with reduced form VAR and estimated a Markov Switched VAR (MSVAR) which allowed to analyze effect of dollarization on the transmission mechanism of monetary policy. Both approaches showed that even though monetary policy transmission mechanism strengthened in the period after 2006, the impact of the policy rate on prices remained weak [6].

On the fiscal policy side, one of the comprehensive analysis is given by Rozenov and Janvelyan (2015). Among two other methods, they estimate a three variable SVAR model (built on Blanchard and Perotti (2002) methodology) for Armenia, which includes net taxes (tax revenue minus subsidies and social transfers), government spending and GDP. They find a quantitatively strong positive response of output to spending shocks in the first year of the fiscal intervention [7]. The same method is used by Lazaryan and

Elkina (2018), who find a quite strong response of output to the government expenditure shock right after the discretionary change of fiscal policy [8].

Method

We build non-restricted VAR with main macroe-conomic and policy variables and use recursive identification for the policy analysis - a method which was introduced in Sims (1980) and made a "tectonic shift" in macroeconomics. Sims criticized large scale statistical macroeconomic models, which used many assumptions for identification which he described as being "inappropriate" and "incredible", hence the models were "over-identified" and not effective for policy analysis [9].

This method, as mentioned in Christiano (2012), "stood the test of the time" [10] and now is known as one of the most commonly used method for policy analysis [11].

For showing the method of recursive identification, we must start from reduced form VAR with p-th order, shown in equation (1), where the N x 1 vector yt denotes the set of variables that is of interest in the analysis:

yt = Bo + B1yt-1 + - + Bpyt-p + ut, Eutu't = V (1)

where ut is not correlated with yt-1, . . ., yt-p. It is assumed that p is assigned a large enough value so ut that is not autocorrelated over time. In equations (1), Bp and V are statistically identified, which makes possible to use the reduced-form VAR for forecasting purposes. But for policy analysis, one needs to analyze also the dynamic effects of economically fundamental (structural) shocks. For that purpose, we need to express ut in structural shocks denoted as et, so that there would be no cross correlations between the components of et. Hence, the VAR disturbances ut are assumed to be a linear transformation of the, et (2):

ut = Cet,CC' = V (2)

In equation (2) C matrix is not econometrically identified, and identification assumptions (restrictions) are required. The VAR with these assumptions is referred as structural VAR or SVAR [10].

According to the recursive identification strategy (also called Cholesky identification), the matrix of contemporary effects must have lower triangular form, which means that the variables in the model become more endogenous according to their order: the first variable is more exogenous in time t: affects all other variables contemporaneously and is affected by them only in lags.

For analyzing fiscal and monetary policy shocks in Armenian economy, we build two reduced form VARs - one for monetary and the other for fiscal policy analysis. For monetary policy analysis, unlike the papers studied in previous section, we not only use the monetary policy key rate of RA CB (R), but also volume of foreign exchange intervention (INT). We analyze the effects of policy actions on nominal exchange rate(EX), money supply M2 aggregate(M), real GDP (Y) and prices (P).

For fiscal policy analysis we use the level of budget spending (SP) and the effective tax rate (TAX) as instruments, and real GDP (Y), prices level (P), exchange rate(EX) and government debt to GDP ratio

(DGDP) to analyze the effects of policy measures. In both models, we control the endogenous variable with external demand (ED, which is based on real GDP growth rates of US, EU and Russia - with weighted averages following to shares of each country (country group) in Armenia's trade.), S&P 500 VIX index and copper price (COPPER) as exogenous variables.

All the variables enter the models in levels (seasonally adjusted where needed), and except for the policy variables (policy rate, interventions, expenditures, tax rate) - in a logarithm. This technique is chosen following the literature studied in the Section 2, and the fundamental paper by Sims, Stock and Watson (1990), who emphasized, that the OLS estimators are still consistent when the model is estimated in levels, even if the system includes non-stationary variables [12].

The lag length of the VAR estimation for both models was selected using the Schwartz (SC) an Han-nan-Quinn information criteria, which suggested a lag of the first order for both models. This is consistent also with papers studied in Section 2 and is efficient in the perspective of keeping the models parsimonious.

The models were tested for the stability, and both were stable. Also, residuals were tested for the serial

the residuals were not serially correlated in second and third lags.

Results

Monetary policy shocks

In the VAR model (3) Yt is a vector of endogenous variables and Zt is a vector of exogenous variables: Yt = AYt-1 + BZt + ut (3) Yt = [Rt,INTt,Yt,Pt,MtEXtj] (4) Zt = [EDt-i,VIXt,COPPERt] (5) The models have been estimated for the data sample 2006Q1 - 2020Q1, which reflects the period when the RA Central Bank have been acting in the inflation targeting regime. In the ordering the policy instruments enter the first, to reflect exogeneity of the shock, which affects all variables contemporaneously, but is not affected by any of them. Afterwards we put output and the price variable, which reflects the sluggish reaction of output and prices to financial and exchange rate shocks (as also emphasized in Anna Rose Bordon and Anke Weber (2010)). The monetary aggregate and exchange rate are expected to take the effect of policy shocks immediately but affect the prices and output in lag, that's why they enter the last. The results are also robust to changes of the ordering.

Figure 1. The impulse responses of monetary policy shock (in 12 quarters) Note: R is measured in percentage points, and other variables are measured in decimals.

The results from impulse response analyses displayed in Figure 1 can be illustrated with the following pattern:

RT ^Mj ^ EX| ^ PI [Y?] As it can be inferred from the impulse response analysis, the monetary policy rate effects exchange rate and money supply and hence - the price level. Particularly, one standard deviation of interest rates (0.5 pp) leads to prices to decrease by 0.4 percent after a year. But the reaction of output is not consistent with the economic theory: output reacts positively to contractionary monetary policy shock. On the one hand, this evidence

can suggest that monetary policy channels other than exchange rate channel are weak, which can be explained by high dollarization in the studied period and underdevelopment of the capital market [13]. On the other hand, this evidence can be a result of misspecifi-cation of the model, particularly omitting a leading indicator to which the monetary authorities react, as was discussed by Sims (1992).

The shock of net foreign currency intervention innovations is studied using the ordering of variables shown in (2).

We infer the following pattern from Figure 2: INT! ^ Yj ^ P j [EX?] Following the impulse responses, positive shock of net FX interventions (selling more foreign currency to the banks than buying) leads to reduction of money supply, output and a permanent decrease of prices. The shock of net interventions by one standard deviation ($81 mln) leads to fall of GDP and prices by 0.4 percent initially. The positive innovation of FX interventions initially also leads to exchange rate depreciation in the framework of our VAR, which possibly reflects the absence of time lag between exchange rate shock and FX intervention, which we can interpret that the Central Bank always intervenes when there is high volatility (mainly sharp depreciation: for more context, see [14]) of the exchange rate, and the exchange rate does not stabilize instantly but stabilizes eventually after some

Fiscal policy shocks

For analyzing the fiscal policy shocks, we build a VAR with the vector of endogenous variables presented in (6):

Yt = [SPt,TAXt,Yt,Pt,MtEXti DGDPt] (6) To get the vector (6), in vector (4) we replaced the monetary policy variables (interest rates and FX interventions) with state budget expenditures and tax rates and added also government debt to GDP ratio at the end of the sequence. We estimate the model for the sample starting from 2000Q1 and ending in 2020Q1. The model is tested for stability and serial correlation of residuals, and the tests showed that the model is stationary and there is no serial correlation in second lag.

The impulse responses of the Government spending shock are shown in Figure 3.

period of time.

Response of SP to SP

Response to Cholesky One S.D. (d.f. adjusted) Innovations ± 2 S.E. Response of Y to SP

1 2 3 4 5 6 7 8 9 10 11 12 Response of M to SP

1 2 3 4 5 6 7 8 9 10 11 12 Response of EX to SP

1 2 3 4 5 6 7 8 9 10 11 12

Response of P to SP

1 2 3 4 5 6 7 8 9 10 11 12 Respon se of DGDP to SP

Figure 3. The impulse responses of Government spending shock (in 12 quarters) Note: SP is measured in bln AMD, and other variables are measured in decimals.

2

We build the following pattern of spending shock transmission mechanism from impulse responses of Figure 3:

The impulse responses demonstrate that the output reacts to positive spending shocks immediately, and the

shock persists at least 1 year and then fades out. Particularly, when government spending temporarily increases by 15 bln AMD in one quarter, the GDP will increase by 1 percent initially and prices - by 0.2 percent. The prices react positively to public spending shock supposedly because of the effect coming from aggregate demand. The money supply also reacts positively, which leads to exchange rate depreciation. The positive reaction of government debt to GDP ratio despite increasing output manifests that surging expenditures mostly were financed with borrowing.

Response to Cholesky One S.D. (d.f. adjusted) Innovations ± 2 S.E

1.2- Response of TAX to TAX Response of Y to TAX Response of P to TAX

/ .004 —-----

\\ .000

0.8- 1 ^____^ .002 .000

0.4- f\ -.004 -.002 ^-

00 ----------------------------------- -.008 -.004 \

\ —-— \ / -.006

1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12

Response of M to TAX Response of EX to TAX Response of DGDP to TAX

.01 - .004- .010-

— ....... .005

-.01 - ____________________ -.004- \ ^-^^

--------- V^

-.02- 03- \ /' -.012- \ -.005-.010- \

1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12

Figure 4. The impulse responses of effective tax rate shock (in 12 quarters) Note: TAX is measured in percentage points, and other variables are measured in decimals.

From Figure 4, we build the following pattern of

tax rate shock transmission mechanism:

The impulse responses demonstrate, that after a positive innovation of tax rate, GDP falls immediately, and the shock persists in a rather long period - fading gradually. Particularly, when the tax rate in one quarter is increased by 1.1 percentage points, GDP falls 0.4 percent initially. Prices also fall, which is probably caused by both demand and cost-push effects on prices. The money supply falls, and exchange rate appreciates. The reaction of debt to shock of effective tax rate is mixed, which shows that there is no clear pattern how the tax and borrowing policies counteract each other.

Conclusion

The analyses demonstrated effects of different macroeconomic policy shocks on the economy. First, we updated the analysis carried out by other authors regarding empirical analysis of policy effects in Armenia. Second, we added on the existing literature with analyzing also macroeconomic effects of foreign exchange market intervention. The results overall were consistent with existing literature.

Following the results of empirical analysis, monetary policy has substantial impact on prices mainly by exchange rate channel, but we did not find evidence how it affect output, although the contractionary interest rate hike causes short term drop of money supply. Instead, we revealed, that a positive shock of interventions in foreign exchange market (selling foreign currency in the market) leads to short-term slump both in money supply and in output, and has deflationary effect on prices. High transportation costs favour exports of those products with high value relative to weight and discourage sales of the bulky low cost products of light industry, which usually involve unskilled labour-intensive activities [15].

We found out, that the effects of fiscal policy on output is clear: a positive innovation of expenditures leads to short term rise of output, and positive shock of tax rate reduces output more persistently. In both cases, the responses of money supply and prices are consistent with output response: money supply and prices surge after the positive expenditure shock and shrinks after the positive tax rate shock. Probably, tax rate hike also generates cost-push inflation. We also showed, that a shock of expenditures generates increasing government debt to GDP trajectory despite increasing output, which shows that the surprise rise of expenditures was mainly debt financed.

The empirical evidence we gathered speaks in favor of fiscal policy instrument counteracting shocks to output, such as the shock caused by COVID-19 pandemic. Furthermore, we state, that when "dovish" monetary policy (cutting interest rates) is being accompanied by interventions in foreign exchange market to reduce surged volatility of exchange rate, overall actions by Central Bank can have contractionary impact on output. The evidence also advocates that tax rate shocks have more persistent impact on output, whereas effects of spending shocks are short and fade out after one year. Acknowledgments

This work was supported by the Committee of Science of Ministry of Education, Science, Culture and Sport of the Republic of Armenia under Grant Maintaining and Development of Scientific and Scientific-Technical Infrastructure Program within the framework of the topic of "Opportunities for Effective Interaction of the Real and Financial Sectors of the RA Economy".

REFERENCES:

1. IMF World Economic Outlook Update, June

2020

(https://www.imf.org/en/Publications/WE0/Issues/20 20/06/24/WE0UpdateJune2020) and IMF Fiscal Monitor - April 2020 (https://www.imf.org/en/Publications/FM/Issues/2020/ 04/06/fiscal-monitor-april-2020).

2. RA Government Medium Term Expenditures Framework 2021-2023 (Armenian version only -http://www.minfin.am/hy/page/petakan_mijna-zhamket_tsakhseri_tsragre/).

3. RA Central Bank Statistical Overview June

2020

(https://www.cba.am/EN/pperiodicals/Statistical%200 verview%2006.20.pdf).

4. Sims, C. A. (1992). Interpreting the macroeco-nomic time series facts: The effects of monetary policy. European economic review, 36(5), 975-1000.

5. Dabla-Norris, M. E., & Floerkemeier, H. (2006). Transmission mechanisms of monetary policy

in Armenia: evidence from VAR analysis (No. 6-248). International Monetary Fund.

6. Bordon, A. R., & Weber, A. (2010). The transmission mechanism in Armenia: New evidence from a regime switching VAR analysis. IMF Working Papers, 1-31.

7. Rozenov R., and V. Janvelyan (2015). "Fiscal Multipliers in Armenia," Republic of Armenia. Selected Issues, IMF Country Report No.15/66.

8. Lazaryan S. S., and M. A. Elkina (2018), "Estimation of the Government Expenditures Multiplier in the Republic of Armenia," Financial Journal, no. 4, Financial Research Institute of RF.

9. Christopher A. Sims, (1980). Macroeconomics and Reality. Econometrica, Vol. 48, No. 1. pp. 1-48.

10. Lawrence J. Christiano (2012). Christopher A. Sims and Vector Autoregressions, Scand. J. of Economics.

11. Ramey, V. A. (2016). Macroeconomic shocks and their propagation. In Handbook of macroeconomics (Vol. 2, pp. 71-162). Elsevier.

12. Sims, C. A., Stock, J. H., & Watson, M. W. (1990). Inference in linear time series models with some unit roots. Econometrica: Journal of the Econometric Society, 113-144.

13. Armenia on Financial Development Index Database in IMF website: https://data.imf.org/?sk=F8032E80-B36C-43B1-AC26-493C5B1CD33B&sId=1481126573525

14. Macroeconomic Insight: Fear of Floating Hampers Armenia's Economy, OG Research, 2017: https://www.ogresearch.com/news/macroeconomic-in-sight-fear-of-floating-hampers-armenias-econ-omy#:~:text=When%20the%20exter-nal%20shock%20hit,2002%20%2D%20see%20refer-ence%20below.

15. Grigoryan Karen, Study of the Peculiarities of Export Developments in EU Member Countries and in Armenia, Romanian Journal of European Affairs, Vol. 12, No. 3, September 2012, pp. 65-82, Available at SSRN: https://ssrn.com/abstract=2142370 or http://dx.doi.org/10.2139/ssrn.2142370