Financial Cycle, Stress, and Policy Roles in Small Open Economy: Spillover Index Approach

Abstract

This chapter analyzes the dynamic spillover of shocks between selected macroeconomic and financial variables in the case of a small open economy. To evaluate the effects of macroprudential policy, alongside its interaction with the fiscal and monetary policy, a spillover methodology approach was made as an extension to the VAR (vector autoregression) modeling. Such methodology enables the researcher and policymakers a dynamic approach of modeling, with detailed insights into the sources of time-varying variability of variable interconnection. Based on quarterly data from 4Q 1995 to 4Q 2021 for the case of Croatia, this chapter analyzes the interactions and shock spillovers between the financial stress, credit, real activity, inflation, macroprudential and fiscal policy, with exogenous variables of the Eurozone monetary policy and its real activity. The approach of this study enables macroprudential policymakers to make decisions promptly, and the decisions could be adjusted based on the interactions between the financial system and the real economy. Furthermore, as findings indicate that macroprudential, monetary, and fiscal policies are not isolated, their interactions and transmission channels affect individual policy effectiveness and spillover from one goal to the other. Therefore, future tailoring of all policies needs to be based on such findings and trying to achieve complementarity with them. Although this is a difficult task, just admitting that such interactions exist eases the later modeling and fine-tuning of all instruments.

Appendices

Appendix 1

Fig. 5

Static generalized impulse response functions. (Source: Author’s calculation)

Table 5 Spillover table, full sample, one lag for monetary policy, h = 24

Table 6 Spillover table, full sample, two lags for monetary policy, h = 12

Table 7 Spillover table, full sample, two lags for monetary policy, h = 18

Table 8 Spillover table, full sample, two lags for monetary policy, h = 24

Table 9 Spillover table, full sample, no lags for monetary policy, h = 12

Table 10 Spillover table, full sample, no lags for monetary policy, h = 18

Table 11 Spillover table, full sample, no lags for monetary policy, h = 24

Appendix 2: Software to Estimate Spillover Indices and Codes

VAR models and spillover indices can be estimated in R, as was done in this research. The following packages are used: vars, frequencyConnectednes and for the dynamic part of the analysis, pbapply. First, VAR model was estimated with the following commands:

• round((VARselect(matrix,lag.max=12,type="const",exo=exo))$selection,5) – where the lag selection was made for 12 VAR models in total

• est<-VAR(matrix,p=3,type="const",exo=exo) – where the model is estimated for the static case

• arch.test(est, lags.single = 24, lags.multi = 24, multivariate.only = T); testing for heteroskedasticity in the estimated model

• serial.test(est,type="PT.asymptotic",lags.pt = 24); testing for autocorrelation in the estimated model

Then, the spillover tables were estimated:

• spilloverDY12(est, n.ahead = h, no.corr=F); for the static case

• and for the dynamic case:

  • window<-28; length of the window

  • caller <- function(j){

      var.2c<-VAR(matrix[(1:window)+j,],p=3,type="const",exo=exo

    [(1:window)+j,])

    sp1<-spilloverDY12(var.2c, n.ahead = h, no.corr=F)

    total<-overall(sp1)

      list(total)

    }; the rolling approach estimation of VAR, and spillover table, with the total spillover

  • out.sd <- pbapply::pblapply(0:(nrow(matrix)-window), caller); extracting the values

  • spillover<-matrix(unlist(out.sd), nrow = length(out.sd), byrow = TRUE); extracting the values

Net spillovers over time were estimated with the net(…) function instead of overall(…), and the pairwise indices were estimated over time with the pairwise(…) function.

Some other possible software to use in order to estimate spillover tables and indices are Eviews, with the Add-in “dyindex”, or Matlab with the functions at the following link: https://www.mathworks.com/matlabcentral/fileexchange/74563-diebold-and-yilmaz-2009-2012-2014-spillover-index.