Fiscal policies in the Federation of Bosnia and Herzegovina: are spending or revenue measures more effective?

We examine the effectiveness of different fiscal policies in the Federation of Bosnia and Herzegovina (FBiH). For this purpose, we use a structural macroeconomic model for the FBiH. In this model, GDP in the Federation is influenced by world demand and by domestic demand in the Federation. Domestic demand comprises consumption of private households, public consumption, and gross fixed capital formation. Employment depends positively on GDP and negatively on the tax wedge, i.e., the net wage plus social security contribution rates (including the unemployment insurance), and the personal income tax rate in the Federation. The latter allows the analysis of the impact of changes in social security contribution rates or in the income tax rate in the Federation of Bosnia and Herzegovina. The following Federation-specific policy instruments are implemented in the model for the FBiH: Pension funds contribution rate in FBiH; contribution rate for health insurance in FBiH; contribution rate for the unemployment insurance in FBiH; benefits from social security; direct tax rates (income tax rate, corporate tax rate); public consumption in FBiH. Our results show that policy measures that reduce the tax wedge on labour income are highly effective in stimulating employment. Due to the large elasticity of imports with respect to demand, pure demand-side measures have little impact on real variables, indicating that a small open economy like the Federation of Bosnia and Herzegovina has only little scope for influencing macroeconomic developments with pure demand management policies. Our results confirm earlier theoretical and empirical studies showing that the labour market can best be influenced positively by reducing the tax wedge. The multipliers of income tax reductions are larger and oscillate more than the effects of the other fiscal policy measures.


Introduction
The economic and financial crisis of 2007-2009, meanwhile known as the "Great Financial Crisis" or the "Great Recession", the following fiscal consolidation phase, and more recently the sharp recession due to the Coronavirus pandemic, have revived the debate in academia as well as among politicians about the adequacy of active fiscal stabilisation policies. During the "Great Moderation" since the mid-1980s, stabilisation policy had been considered to be of less importance (Lucas 2003). Except for such extreme events as the lockdown measures enacted to contain the spread of the Coronavirus in 2020 and 2021, within academia opinions about the effectiveness of expansionary fiscal policy measures are sharply divided. While some authors (e.g., Taylor 2009) argue against using fiscal policy in a discretionary way, others point towards potentially large multiplier effects of tax reductions or expenditure increases (e.g., Romer and Romer 2010). Although there is a lot of evidence regarding the effects of macroeconomic policies in different countries during the Great Recession, its interpretation still diverges among macroeconomists and politicians. In particular, the role of fiscal policy is subject to ongoing controversies (see, for instance, Coenen et al. 2008Coenen et al. , 2013Cogan et al. 2010). Coenen et al. (2008) analyse macroeconomic effects of fiscal consolidations, while Coenen et al. (2013) and Cogan et al. (2010) analyse expansionary fiscal policies. Coenen et al. (2013) find sizeable effects of higher government consumption and investment, while they conclude that revenue-based fiscal expansions only have small effects. Coenen et al. (2008) find that both expenditure and revenue-based consolidations can have a significant impact on macroeconomic aggregates, at least in a model with an endogenous response of the equilibrium real interest rate. Cogan et al. (2010) conclude that the size of fiscal multipliers depends on the macroeconomic model used and the underlying economic theory. Compared to traditional Keynesian models, fiscal policies are much less effective in new Keynesian models.
In this paper, we aim at contributing to this debate by empirically estimating fiscal policy effects for Bosnia and Herzegovina, or to be precise for its entity named Federation of Bosnia and Herzegovina (FBIH). We are particularly interested in the question whether demand-side (Keynesian) fiscal policies aiming primarily at supporting demand can contribute to stabilising the economy or if some elements of supply-side orientation have to be added to render these policies successful. The debate between Keynesians and supply-siders was a hot topic in the 1980s in the wake of the oil price shocks and (as many macroeconomic policy debates) has not been completely settled since then. The prevailing opinion (though not a consensus) considers demand-side policies to be appropriate when combating an adverse demand-side shock but not necessarily when faced with a supply-side shock (such as stagflation). The Great Recession-as most real-world shocks-contained both demand and supply elements, but most interpretations agree that demand-side elements prevailed. Nevertheless, policies proposed by the European Commission and by the International Monetary Fund (IMF) contain calls for structural reforms to enhance growth and employment both in the short and the long term, which implies for fiscal policy to embed also supply-side measures. By contrast, many politicians and interest group representatives heavily criticize what they call the "austerity regime" of the European Commission, and advocate an expansionary fiscal policy stance in spite of already high public debt.
With the help of an econometric model, we examine the question whether the Federation of Bosnia and Herzegovina would benefit more from demand or from supply-side measures of its fiscal policy. The theoretical and empirical literature finds that in general spending multipliers are larger than tax multipliers. However, measures reducing the tax wedge not only entail demand effects by increasing disposable income, but in addition they bring about supply-side effects by increasing incentives to work or by reducing the upward pressure in wage negotiations.

Fiscal policies in the Federation of Bosnia and Herzegovina
In this part, we provide a brief introduction to the economic and policy environment in Bosnia and Herzegovina (BIH). BIH was part of Yugoslavia, and it is an independent country since 1992. It has an extremely complicated constitutional structure, and it is composed out of two entities and one district. One of the two entities is the Federation of Bosnia and Herzegovina (FBiH), which is further composed out of ten cantons and municipalities. The entities have constitutional competencies to regulate the area of direct taxes i.e., social security contribution rates, income taxes. The central bank of BiH runs a currency board with the national currency, the Konvertible Mark (KM) pegged to the euro. In 2019 real GDP in Bosnia and Herzegovina rose by 2.8%, followed by a drop of 3.1% in 2020, mainly due to the economic consequences of the Coronavirus pandemic.
Regarding fiscal policy making, the budget preparation at the different levels is usually delayed, and due to institutional deficiencies, the functioning of the country is not efficient. Bosnia's main challenges are low productivity, which results in low salaries, low employment rates, and high emigration. Since the country runs a currency board regime with very underdeveloped financial markets, fiscal policy is the key policy instrument for the government.
In the Federation of Bosnia and Herzegovina, fiscal policy is conducted by the government of FBiH in line with the global fiscal framework for BIH. These policies include measures of public investment, based on an annual and a 3-yerar (t, t + 1, t + 2) expenditure framework document (EFD). In addition, the issue of change in the tax wedge in FBIH is being discussed in the FBIH Parliament. 1 1 3

Definitions and determinants of fiscal multipliers
The Great Recession and the following fiscal consolidation revived the discussion of the effectiveness of macroeconomic stabilisation policies. In particular, the size of fiscal multipliers has been hotly debated since then, both in academia and in institutions such as central banks or the IMF. For countries that are members of a currency union such as the Euro area or that are operating under a currency board like Bosnia and Herzegovina, estimates of the effects of fiscal policy are particularly relevant since this is the only remaining macroeconomic policy instrument to deal with adverse shocks. The effectiveness of fiscal policies is usually evaluated via the size of fiscal multipliers, which are usually measured as the ratio of the change in economic output over the change in an exogenous spending or revenue item. Since the size of multipliers is likely to vary over time, various definitions of multipliers may be estimated (see, e.g., Berg 2015). The first one, the impact multiplier, measures the reaction of economic output to a change in a fiscal policy instrument in the same period t, where Y denotes output (usually GDP), INST is the policy instrument, e.g., public consumption or tax revenues, and t is the time period: The multiplier at horizon i is defined at: Since multipliers tend first to rise, reach a peak, and decline again, it is also interesting to derive the maximum response of output to the initial fiscal shock, i.e., the peak multiplier: The determination of fiscal multipliers is wide-spread in academia as well as in policy-oriented papers published by institutions like the IMF and the OECD, since they can be easily communicated and compared across countries and time periods (Čapek and Crespo Cuaresma 2020). Furthermore, the precision of the estimation of fiscal multipliers contributes significantly to the quality of GDP growth predictions (Blanchard and Leigh 2013). In the latter study, the authors find that in advanced economies stronger planned fiscal consolidation in the period after the Great Financial Crisis was associated with lower growth than expected. The likely reason for this is that fiscal multipliers were substantially higher than assumed by forecasters.
Despite their abundance, theoretical and empirical studies have not yet reached a consensus on the size of fiscal multipliers, nor on the question how to design fiscal policies when facing a severe crisis or when consolidating the budget. What has been identified, however, are the factors that determine the effectiveness of fiscal policies. These factors comprise trade openness, labour market rigidities, the size of automatic stabilisers, the exchange rate regime, the share of credit-constrained consumers, the debt level, the effectiveness of fiscal administration, the state of the economy in the business cycle, as well as the stance of monetary policy (see, e.g., Batini et al. 2014). Fiscal policies are less effective ceteris paribus in small open economies than in larger and less open economies, but even for open economies the empirical evidence is mixed. Especially for small open economies, an internationally coordinated fiscal action might be more effective than isolated policies. Furthermore, an already high level of public debt is likely to undermine positive effects of fiscal stimuli. Hence, a clear commitment to fiscal consolidation after overcoming a crisis is required (see, e.g., Spilimbergo et al. 2009;IMF 2008). Fiscal multipliers may also vary with the position in the business cycle. Auerbach and Gorodnichenko (2013) conclude that spending multipliers tend to be larger in recessions than in expansions. Furthermore, strict fiscal consolidation measures in a recession might contribute to a deepening of the recession (Blanchard and Leigh 2013). Labour market rigidities increase fiscal multipliers if these rigidities reduce wage flexibility, since the response of output to a fiscal shock is larger if wages react only little. Lage automatic stabilisers reduce multipliers of active fiscal policies as the automatic response of taxes or spending offsets part of the initial fiscal policy action (Batini et al. 2014).
Regarding the design of fiscal policies, it is still an open issue whether taxes or public spending are more effective at stimulating output and employment (see, for example, Erceg and Lindé, 2013;Bouakez et al. 2014;Dufrénot et al. 2016). The side effects on government debt, which are a relevant problem in view of the relatively high debt to GDP ratios in many industrialised countries, may also be affected differently by revenue and expenditure side measures. Using a similar approach as in the present paper, Weyerstrass et al. (2020) find that those fiscal policy measures that entail both demand and supply-side effects are more effective at stimulating output than pure demand-side measures. Employment can be most effectively stimulated by cutting the income tax rate and the social security contribution rate, i.e., by reducing the tax wedge on labour income and positively affecting the country's international competitiveness. Except for higher spending on research and development, all fiscal policy measures lead to higher public debt.

A macroeconometric model of the Federation of Bosnia and Herzegovina
For the simulations, we use a macroeconometric model of the Federation of Bosnia and Herzegovina. The model was estimated with annual data for the period 2000-2018. However, for many variables the period for which data are available is much shorter. In principle, the model is comparable to a model for the entire country Bosnia and Herzegovina as described in Weyerstrass (2009). One crucial difference is that the latter model also includes a supply block determining potential GDP via a production function. Due to the lack of capital stock data on the entity level, the model for the FBiH cannot contain such a supply part. Since most time series have a unit root, i.e., they are non-stationary in levels, but stationary in first differences, almost all behavioural equations were estimated in error-correction form. The model equations are shown together with a variable list in the appendix. In the model, GDP in the FBiH economy is influenced by international demand and by domestic demand in the Federation. Domestic demand comprises consumption of private households, public consumption, and gross fixed capital formation. Private consumption is determined on the level of the Federation. It is influenced by disposable income, which in turn depends on taxes and social security contributions as well as transfers in the FBiH. Investment is also determined on the level of the Federation. It is influenced by GDP as the overall activity variable and by the real interest rate. Public consumption is determined in a Taylor type fiscal rule. In this rule, public consumption is positively linked to the past development of the fiscal balance. The notion behind this modelling is that an increase in a fiscal deficit should be counteracted in the following period by a public spending restraint, while an improvement in the fiscal balance allows the government to spend more. In the simulations, we turn this equation off and instead treat government consumption as an exogenous policy variable, the variation of which determines one of the fiscal multipliers. Exports of the FBiH are influenced by international demand, which is approximated by aggregated GDP in the Federation's ten most important trading partners, and by the real effective exchange rate of the Bosnian currency, the Konvertible Mark (KM). Imports depend on final demand, while an influence of the exchange rate was not supported by the data. Employment, i.e., labour demand by companies, depends positively on GDP and negatively on the net wage plus the tax wedge. For policy simulations different from those described in this paper, we differentiated employment by three qualification levels. The influence of the tax wedge on labour demand varies with the qualification level. The tax wedge, i.e., the ratio between the net and the gross wage, where the gross wage is computed by summing net wage and the contributions to the various parts of social security: the health, pension, unemployment, and accident insurance. Labour supply is determined via the participation rate. It positively depends on the gross wage. In the price block, consumer prices and the GDP deflator are determined. The consumer price index (CPI) depends on the gross wage as the most important domestic cost factor, and on world prices. The latter are approximated by the HWWI index of raw material prices. The GDP deflator is explained by the CPI. In the public sector block, collections from social security contributions, value added tax (VAT) revenues, income tax revenues, other tax revenues, other revenues, social benefits, subsidies, public consumption, other expenditures, and the budget balance are determined. The revenues are determined by multiplying the relevant tax base by the respective tax or contribution rate. VAT revenues are determined in an estimated equation with total private consumption, multiplied by the normal VAT rate, as determinant. The use of a behavioural instead of a definition equation is necessary to account for tax exemptions. Social benefits, which are an important part of disposable income, are explained by health care expenditures and by pensions. Health care expenditures are determined by the revenues of the health fund.
Pensions are calculated by the number of pensioners, multiplied by the average pension. As in the case of government consumption, for the determination of fiscal multipliers the equation determining social benefits is turned off, thus making social benefits exogenous in the model, which allows treating them as a policy instrument. Figure 1 shows an outline of the FBiH model. The following FBiH-specific policy instruments are implemented in the model for the FBiH.
• Pension funds contribution rate in FBiH • Contribution rate for Health Insurance, FBiH • Contribution rate for Unemployment Insurance, FBiH • Income tax rates (personal and corporate) • Value added tax (VAT) rate • Benefits from social security • Public consumption in FBiH Although our macroeconometric FBiH model is used for forecasting and policy simulations, it should be noted that the model-like every structural econometric model-may be subject to the famous Lucas critique. Lucas (1976) argued that the relations between macroeconomic aggregates in an econometric model change when the macroeconomic policy regime changes. In this case, the effects of a new policy regime cannot be predicted using an empirical model based on data from previous periods when that policy regime was not in place. As Sargent (1981) argues, the Lucas critique is partly based on the notion that the parameters of an observed policy rule should not be viewed as structural. Instead, structural parameters in Sargent's conception are "deep parameters" such as preferences and technologies. These parameters would be invariant, even under changing policy regimes. Providing for such "deep parameters" requires a different class of macroeconomic models, namely computable general equilibrium (CGE) or dynamic stochastic general equilibrium (DSGE) models. An approach taking the Lucas critique into account in structural models like ours emerged in London School of Economics' tradition initiated by Sargan (1964). According to this approach, economic theory guides the determination of the underlying long-run specification, while the dynamic adjustment process is derived from an analysis of the time series properties of the data series. Error correction models involving cointegrated variables combine the long-run equilibrium and the short-run adjustment mechanism.

Simulation design
We estimate the multipliers, i.e., the effects of changes in the various fiscal policy instrument, by performing different simulations with our macroeconometric model. To this end, we first run a baseline simulation over the period 2019-2030, assuming plausible paths of the exogenous variables like world demand and international raw material prices as well as the fiscal policy instruments. Then, we implement changes to the policy instruments, one instrument at a time, and perform new simulations. We then calculate differences in important macroeconomic aggregates (nominal and real GDP, inflation, employment, unemployment rate, net exports, and the budget balance in relation to GDP) between the various policy simulations and the baseline simulation. For the baseline simulation, we assume that, due to the Coronavirus pandemic, demand from the Federation's most important trading partners decreases by 3.5% in 2020 and rises by 3.5% in 2021 and by 3% p.a. from 2022 onwards. World raw material prices are assumed to collapse by 30% in 2020 (after having decreased by 10% in 2019), before rising by 30% in 2021 and b 5% annually from 2022 onwards. For the FBiH's working-age population we assume a continuation of the declining trend, while the number of pensioners should continue to rise. Regarding the fiscal policy variables, we hold the tax and social security contribution rates constant at their latest observed values, varying between 2018 and 2019. For government consumption, we take the results from a first baseline simulation with government consumption being endogenously determined via a fiscal rule. This results in quite low growth rates of 1.9% in 2020, 0.3% in 2021, and 0.5% in 2022. Then the growth rate of public consumption gradually picks up to 3.3% in 2030. Similarly, the path of social benefits is firstly determined endogenously in an estimated equation. This simulation gives a negative change of social benefits by 4.3% in 2020. Afterwards, these transfers rise, with the growth rate increasing from 3.2% in 2021 to 6% in 2030. In the simulations determining the fiscal multipliers, we generally take these paths of public consumption and social benefits, only changing them in the respective policy simulations.
These assumptions and settings result in paths of the macroeconomic aggregates of the FBiH during the period 2019 to 2030, forming the basis on which to determine the fiscal multipliers. Due to the Coronavirus pandemic, the model predicts real GDP in the Federation to decline by 4.0% in 2020 and to grow by 7.3% in 2021. Afterwards, the real growth rate gradually moderates from 5.7% in 2022 to 4.7% in 2030. This brings about a gradual reduction in unemployment. The unemployment rate first rises from 38.4% in 2019 to 41.1% in 2020, but then decreases to just 19% in 2030. Of course, this favourable development is not caused by the real GDP growth alone, but also by the decline of the population in working age. Due to the Coronavirus pandemic, employment decreases by 4.7% in 2020, followed by an increase of 3.4% in 2021. Afterwards, the annual growth rate of the number of employees settles at around 3.5% until 2030. The consumer price index is forecast to decrease by 2.3% in 2020, which is due to the collapse in international raw material prices because of the Coronavirus crisis. Then, the inflation rate becomes positive in 2021, reaching 1.3% at the end of the simulation period.
For determining the macroeconomic impacts of different fiscal policies, we perform various policy experiments in which we change, one in each simulation, the fiscal policy instruments. For each instrument, we distinguish two simulations, one with permanent and one with temporary fiscal policy measures. For the permanent fiscal policy measures, we implement changes to the instruments from the year 2020 onwards, while in the scenarios of temporary measures we implement these changes in the year 2020 only. We assume an increase in government consumption and social benefits by 100 million KM each. The magnitude of the simulation experiment is taken arbitrarily but it is based on the fiscal size of these components in the FBIH budget, and it is comparable to the revenue measures i.e., the initial revenue loss is approximately 100 million KM. Also, the magnitude was specified in such a way to avoid potential distortions in the market mechanism, monetary policy reactions or large budget deficits that could lead to violations of fiscal rules. Specifically, we perform the experiments summarized in Table 1.
The policy instruments work through different channels. The spending measures entail a demand-side effect only, either directly (government consumption) or indirectly by raising disposable income of private households (social benefits). Also, reductions in the VAT rate are primarily directed towards demand. To the contrary, reductions in the income tax rate as well as in the social security contribution rate decrease the tax burden on labour income. By reducing the tax wedge between gross and net wages, they reduce the pressure on gross wages in wage negotiations. Thus, these measures decrease labour costs, thereby increasing incentives for companies to employ more workers. Regarding the scenario with the reduction in the social security contribution rate, we chose health care as the representative social security branch. In addition to health care, the social security system comprises the unemployment insurance, the pension insurance, and the accident insurance. Contributions to the social security system are in general divided between employers and employees. Hence, we implemented the reduction in the social security contribution rate by cutting both the employers' and the employees' rates by 2 percentage points each. Which side bears the social security contributions depends on their incidence, which in turn is determined by the relative power of trade unions and employer associations in the wage negotiations.

3 6 Simulation results-interpretation of fiscal multipliers
In this section, we show the results of the previously described policy scenarios. In each of the following figures, deviations of the respective macroeconomic aggregate from the baseline simulation are depicted. Since we focus not only on nominal GDP but also on other important macroeconomic aggregates (inflation, employment, unemployment, net exports, public budget), not all results are fiscal multipliers in the strict sense. However, we normalised the results by dividing the change in the respective variables by the change in total expenditures or total revenues, respectively. Exceptions are the results for the inflation rate and for the unemployment rate, since these impacts would be extremely small in the case of the mentioned normalisation. For these two variables, the absolute deviations between the baseline and the alternative scenarios as described in Table 1 are shown in the figures. The effects are shown in Figs. 2 , 3, 4, 5, 6, 7 and 8. In addition, for the temporary measures, Table 2 summarises the impact multipliers, the multipliers after 5 years, and the maximum effects.
As one would expect, the GDP effects of increases of public consumption are a bit larger than those of increases in social benefits of the same magnitude (Figs. 2,  3). Regarding the tax measures, reductions of the income tax rate have the smallest effect, and the effects become negative over time in the case of a temporary tax reduction. Although our model focuses on the demand-side, it generates typical supply-side effects of some fiscal policies. In particular, employment can be very effectively increased, and hence unemployment reduced, by measures that reduce the tax wedge or labour costs (Figs. 4, 5). Interestingly, these positive labour market effects are rather short-lived. This result is in line with Weyerstrass et al. (2020) for Slovenia.
The inflationary impacts of the fiscal policy measures are small (Fig. 6). Interestingly, the inflation effects of the temporary measures even become negative over time, in particular regarding the temporary demand-side effects. A negative effect means that inflation with the implementation of the measure is lower than inflation in the baseline simulation. All fiscal policy measures, except for the income Health care contribution rate − 4 pp from 2020 onwards tax reduction in some years, decrease net exports (Fig. 7). This is to be expected since the measure (at least temporarily) raise domestic demand and hence imports. Furthermore, the (although only slightly) higher inflation leads to a real appreciation of the Bosnian currency which is also detrimental to net exports. Finally, of course all expansionary fiscal policy measures deteriorate public finances (Fig. 8). Striking is the profile of the reaction of all macroeconomic variables to changes in the income tax rate. This fiscal policy measure entails much larger effects than the other measures in the first years after implementation, and the multipliers oscillate over time. In the model, the reaction of employment to changes in the personal income tax rate is substantial, making this instrument highly effective in influencing the macroeconomy.
In line with the definition of multipliers is Sect. 3, Table 2 summarises the impact multipliers, the multipliers after 5 years, and the maximum multipliers for the temporary policy measures. For this analysis, we focus on the temporary measures, since regarding the sustainability of public finances, temporary measures are more appropriate than permanent expansionary measures. This is also in line with recommendations of institutions like the IMF or the OECD that fiscal stimuli should by timely, targeted, and temporary. The impact multiplier of the income tax reduction is the largest, followed by the other tax measures. The applies to GDP and employment. Also, the maximum multipliers of the tax reduction are larger than the effects of the other measures. However, after 5 years, most multipliers are zero, i.e., the macroeconomic variables have returned to their baseline values, but the multipliers of the tax reduction are negative after 5 years. It would be premature to suggest strong and precise recommendations for the current macroeconomic situation of the Bosnian or the Federation's economy based on just one model specification. It should be noted, however, that in a recent study with a New Keynesian DSGE model, Sims and Wolff (2018) show that countercyclical fiscal policies are more successful when using productive public investment instead of public consumption. Since due to lack of data our model does not incorporate potential GDP, we cannot explore these effects in our model. Our results do confirm earlier theoretical and empirical studies showing that the labour market can best be influenced positively by reducing the tax wedge.

Appendix: The macroeconometric model of the Federation of Bosnia and Herzegovina
This appendix provides a list of all behavioural equations and identities of the model used for the simulation. A list of the variables is provided below the equations. The equations can be grouped into different blocks: • GDP and its components: nominal and real GDP, gross fixed capital formation, private consumption, public consumption, exports, imports. Public consumption can also be grouped into the public sector part of the model (see below). • Prices and wages: gross and net wage, consumer price index (CPI), GDP deflator. • Labour market: employment, total as well as divided into three qualification levels (low, medium, high), unemployment, number of pensioners. • Public sector: collections from social security contributions, VAT revenues, income tax revenues, other tax revenues, other revenues, social benefits, subsidies, (public consumption), other expenditures, budget balance.

Imports of goods and services
Imports of goods and services (MP) are influenced by domestic demand, approximated by GDP. In addition, dummies for 2009 and 2010 had to be included.

Public consumption (fiscal statistics)
In a Taylor type fiscal rule, government consumption (GOVCONS) depends on the lagged budget balance (BALGDP). An improvement in the budget balance allows the government to spend more in the following year, while a deterioration of the fiscal balance has to be counteracted by a spending restraint. Furthermore, lagged public consumption influences current public consumption since public expenditures cannot be adjusted quickly.

Nominal GDP
Nominal GDP could theoretically be determined by adding up the expenditure components. However, data on the level of the entities is less accurate than data for the state level. In particular, it is difficult to measure exports and imports for the entities. Furthermore, capital formation does not include changes in inventories. In addition, public consumption refers to fiscal statistics, not the national accounts, and there are some discrepancies between these two concepts. Hence, GDP is determined in a behavioural equation by the expenditure components (HCP, GFCFP, XP, MP). A dummy for 2013 is also considered.

Gross wage rate
Wages depend positively on prices and negatively on the unemployment rate. The elasticity of gross nominal wages (GWAGE) with respect to consumer prices (CPI) have been restricted to 1 so as to ensure that in the long run real wages do not decline. The negative influence of the unemployment rate follows the notion that an improvement in the labour market situation gives employees or trade unions, respectively, more power in wage negotiations, while an increase of the unemployment rate negatively affects the employees' negotiation power.

Consumer price index (CPI)
Consumer prices (CPI) are influenced by domestic costs, approximated by labour costs, which are in turn represented by the average gross wage (GWAGE), and by international raw material prices (HWWI).

Total employment
Employment depends (EMP) positively on GDP (GDPP) and negatively on the gross wage (GWAGE) as well as the relation between social security contribution rates in the Federation and in the Republika Srpska (RS). The latter allows the analysis of the impact of changes in social security contribution rates in the Federation relative to those in the RS. Furthermore, lagged employment plays a role since employment is typically adjusted smoothly.

Employment-low qualification
Also the share of persons with low qualification in total employment (EMP_ LOW/EMP) depends negatively on the relation between social security contribution rates in the Federation and in the RS. Dummies for 2009 and 2010 had also to be included.

Employment-medium qualification
The share of persons with medium qualification in total employment (EMP_ MED/EMP) depends positively on the relation between social security contribution rates in the Federation and in the RS. The notion is that non-wage labour costs are particularly detrimental to employment with low qualification. A dummy for 2008 had also to be included. Employment with high qualification is calculated as the residual of total employment minus employment with low and medium qualification.

Employment (LFS)
Employment according to the labour force survey (EMPLFS) is determined by registered employment (EMP).