Dynamic microsimulation, Taxes and benefits

Research articles · Apr 30, 2024

Keeping the Best of Two Worlds: Linking CGE and Microsimulation Models for Policy Analysis

Konstantins Beņkovskis, Ludmila Fadejeva, Anna Pluta, Anna Zasova

Monetary Policy Department, Latvijas Banka, Riga, Latvia; Baltic International Centre For Economic Policy Studies, Strēlnieku iela, Riga, Latvia

Volume 17(1) Spring 2024

Cite as: K. Beņkovskis, L. Fadejeva, A. Pluta, A. Zasova; 2024; Keeping the Best of Two Worlds: Linking CGE and Microsimulation Models for Policy Analysis; International Journal of Microsimulation; 17(1); 122–144. doi: 10.34196/ijm.00306

Abstract

In this paper, we link a CGE model with the tax-benefit microsimulation model EUROMOD for Latvia. The model linkage is done using an iterative top-down bottom-up approach, ensuring the convergence of changes in disposable income, employment and wage in both models. We also incorporate the unreported wage payments in CGE and EUROMOD to account for the substantial labour tax non-compliance in Latvia and improve the modelling of the fiscal sector. Two simulations demonstrate the advantages of the joint CGE-EUROMOD system over the individual macro and microsimulation models. The lack of income distribution aspect and the scarcity of fiscal instruments in CGE can be overcome by the features of EUROMOD. The CGE model, on the other hand, provides macroeconomic spillovers that are missing in the simulations of EUROMOD.

This article contains 7 figures and 6 tables.

Figures

Figure 1. Latvian CGE model linked with Latvian EUROMOD
Figure 2. Change of economic status in the EUROMOD database. Note: P(U) denotes the probability to be employed, P(I) – probability to be inactive, P(EL) – probability to be employed in low-skilled occupation, P(EM) – probability to be employed in medium-skilled occupation, P(EH) – probability to be employed in high-skilled occupation. P(x|y) corresponds to the conditional probability, e.g. P(U|EL) indicates the probability to become unemployed given that the person is currently employed in low-skilled occupation.
Figure 3. Changes in status and weights
Figure 4. Convergence process of several CGE-EUROMOD simulations. (a) Consistency: root mean square deviation. (b) Consistency: maximum deviation. Note: Own calculations. The "Financial services shock" simulation is described in Section 4.1 and the "Child-related benefits" simulation - in Section 4.2.
Figure 5. Change in the disposable income (left) and share of at risk of poverty (right). (a) Equivalised disposable income by decile (deviation from baseline; %). (b) At risk of poverty (excluding unreported wages) (deviation from baseline; % points). Note: Own calculations. The income cut points for deciles and median income used for estimation of at risk of poverty threshold (60% of median) are based on the baseline scenario for 2018. The figures show the results of the CGE-EUROMOD model.
Figure 6. Changes in equivalised disposable income (deviation from baseline; %). (a) By number of children. (b) By decile of equivalised disposable income. Note: Own calculations. Income cut points for deciles are based on the baseline scenario for 2021 and are constant across models.
Figure A1. Sequence of solutions for the joint CGE-EUROMOD system. Note* denotes that the model ueses a predicted dataset: SUT in the case of the CGE model and EU-SILC in the case of EUROMOD (the year indicates the income data year in EU-SILC).

Tables

Table 1. Aggregate statistics on evaluated unreported wage payments in 2018
The share of employees with unreported wage; %The ratio of unreported wage to legal gross wage for employees with non-zero unreported payments; %
Total19.223.3
1st quintile (based on legal gross wage)42.344.5
2nd quintile37.523.8
3rd quintile15.317.6
4th quintile4.414.9
5th quintile1.515.9

Note: Own calculations.

Table 2. Changes in selected indicators and selected industries in response to the financial services shock simulation (deviation from baseline in 2018; %)
VariableTotalFinancial servicesAuxiliary financial servicesWood productionConstruction
Real output–0.44–13.25–3.060.13–0.23
Nominal gross wage–0.33–4.13–1.26–0.20–0.28
Nominal gross wage (high-skilled)–0.38–5.11–1.31–0.16–0.29
Nominal gross wage (medium-skilled)–0.35–3.51–0.97–0.21–0.29
Nominal gross wage (low-skilled)–0.25–1.60–0.47–0.21–0.25
Employment–0.13–12.21–2.720.15–0.19

Note: Own calculations.

Table 3. Family benefit by number of dependent children ( euro per month)
YearFamily with 1 dependent childFamily with 2 dependent childrenFamily with 3 dependent childrenFamily with 4 dependent children
202111.3844.14122.90229.8
202225.00100.00225.00400.00

Note: Own calculations. We assume that in 2021 the sequential number of children corresponds to the actual number of dependent children in a family.

Table A1. Multinomial logistic regression
(1)(2)(3)(4)(5)(6)(7)(8)
MenMenMenMenWomenWomenWomenWomen
VariablesLow-skilledMedium-skilledUnemployedInactiveLow-skilledMedium-skilledUnemployedInactive
Secondary education–1.612***–1.669***–2.251***–2.339***–1.098*–0.207–2.155***–2.619***
Tertiary education–4.405***–4.264***–4.769***–4.967***–4.268***–2.471***–4.603***–5.142***
Age–0.068***–0.068***–0.124***–0.266***–0.049***–0.057***–0.105***–0.212***
Age20.001***0.001***0.002***0.003***0.001***0.001***0.001***0.002***
Cities other than Riga0.441***0.410***0.472***0.489***0.097**0.0280.456***0.141***
Thinly populated area0.636***0.506***0.772***0.855***–0.070*–0.194***0.405***0.320***
Married–0.262***–0.153***–0.701***–0.831***–0.142***–0.078*0.0060.241***
Household non-employment income, per household member, in logs0.028***0.034***0.028**0.059***–0.007–0.010–0.032***0.003
Earnings of other household members, per household member, in logs–0.034***–0.066***–0.171***–0.158***–0.019***–0.003–0.126***–0.065***
Number of children under 3 years of age–0.112*–0.067–0.788***–0.656***–0.015–0.199**1.087***1.107***
Disabled0.917***0.1703.153***4.945***0.839***0.749**3.313***4.879***
Paying mortgage–0.512***–0.560***–0.869***–0.819***–0.478***–0.458***–0.614***–0.478***
Constant3.363***3.514***5.567***7.843***3.264***1.945**5.276***7.517***
Observations33’05533’05533’05533’05536’22036’22036’22036’220
YearsYesYesYesYesYesYesYesYes

*** p<0.01, ** p<0.05, * p<0.1.

Note: Own calculations. Data used is EU-SILC 2011-EU-SILC 2019, Latvia. High-skilled is base outcome of the multinomial logistic regression.

Table A2. Heckman selection model
(1)(2)(3)(4)
MenMenWomenWomen
VariablesRelative wage in logsSelection modelRelative wage in logsSelection model
Secondary education0.348***0.387***0.443***0.495***
Tertiary education0.521***0.669***0.566***0.806***
Medium-skilled0.173***–0.0360.124***–0.021
High-skilled0.355***0.241***0.289***0.314***
Age0.041***0.0030.014***0.027***
Age2 (*1000)–0.525***–0.227**–0.177***–0.417***
Cities other than Riga–0.088***–0.051–0.159***–0.079***
Thinly populated area–0.068***–0.227***–0.148***–0.198***
Married0.089***0.227***0.008–0.217***
Household non-employment income, per household member, in logs––0.010––0.010*
Earnings of other household members, per household member, in logs–0.101***–0.062***
Paying mortgage–0.472***–0.134***
Number of children under 3 years of age–0.253***––0.911***
Disabled––1.003***––1.044***
Number of months spent in full-time work0.085***–0.098***–
Number of months spent in part-time work0.015***–0.034***–
Constant–2.223***0.500**–2.033***–0.024
atanh(ρ)–0.411***––0.474***–
ln(σ)–0.486***––0.392***–
Observations24’47724’47728’38428’384
YearsYesYesYesYes

*** p<0.01, ** p<0.05, * p<0.1.

Note: Own calculations. Data used is EU-SILC 2011–EU-SILC 2019, Latvia. Dependent variable in wage equation is employment income relative to the industry average (in logs).

Table A3. Changes in selected macroeconomic, fiscal and income distribution variables in response to two scenarios (deviation from baseline; %).
VariableScenario 1: Shock to financial servicesScenario 2: Changing family benefits
Macroeconomic variables
Real GDP–0.440.28
Real private consumption–0.170.81
Real investments–0.420.26
Real exports–0.70–0.13
Real imports–0.250.30
GDP deflator–0.220.15
Consumption deflator–0.160.11
Export deflator–0.120.08
Import deflator0.000.00
Unemployment (percentage points)0.09–0.07
Employment–0.130.10
Employment (high-skilled)–0.170.09
Employment (medium-skilled)–0.200.13
Employment (low-skilled)–0.040.10
Gross real wage–0.170.09
Gross nominal wage–0.330.20
Gross nominal wage (high-skilled)–0.380.19
Gross nominal wage (medium-skilled)–0.350.21
Gross nominal wage (low-skilled)–0.250.20
Gross nominal reported wage–0.440.26
Fiscal variables
Budget balance (percentage points to GDP)–0.28–0.06
Budget expenditures0.100.65
Social expenditures0.312.58
Old-age pensions–0.020.02
Disability pensions0.000.00
Sickness benefits0.000.00
Unemployment benefits5.720.07
Family-related benefits0.0014.39
Other benefits0.04–0.15
Budget revenues–0.650.50
Social security contribution revenues–0.640.30
Personal income tax revenues–0.970.37
Value added tax revenues–0.540.96
Excise tax revenues–0.390.55
Income distribution (EUROMOD-CGE)
Gini coefficient (equivalised total disposable income)–0.25–0.44
Gini coefficient (equivalised reported disposable income)–0.32–0.48
S80/S20 (equivalised total disposable income)–0.01–0.65
S80/S20 (equivalised reported disposable income)–0.11–0.85

Notes: Own calculations. Baseline for Scenario 1 is the year 2018, for Scenario 2 - the year 2021.