3.1 Equilibrium conditions

The equilibrium of the model is defined by a set of prices and an allocation of goods and factors. It involves the simultaneous solution of three sets of equations:

1. –

   Zero-profit conditions.

2. –

   Market clearing in goods and factor markets. One version of the model includes labour unemployment and another version assumes full labour employment.

3. –

   Constraints on disposable income (i. e. total revenue must equal total expenditure), labour market (in the model that includes unemployment), public sector constraints, and macroeconomic closure of the model.

3.2 Firms and production

Production is based on a technology characterized by a nested structure of intermediate inputs, several types of capital and labour. The firms’ decision problem is to maximise profits subject to the technology constraints, obtaining the unit cost functions, which are further used in the zero-profit conditions. In turn, the demands for factors and intermediate inputs are obtained from Shephard’s lemma on cost functions, and then used in the market-clearing equations.

Firms show constant returns to scale in their technologies and fix a competitive pricing rule, with free entry and exit of firms. Figure 3 shows the nested structure of firms’ technology in sector i. This is a three-level technology. The first nest is a Leontief function over intermediate inputs and a composite of labour and capital, where several types of indirect taxes can be levied on intermediate inputs. The second nest is a CES function representing a composite over non-ICT capital and an aggregate of labour and ICT capital. The third nesting level is a CES between ICT capital and labour, where social contributions are levied on labour. The capital belonging to national firms is different from the foreign MNEs’ capital.

Figure 3:

Production function nests for each type of firm.

However, note that within each sector there are two different varieties of the same good: a national variety produced by national firms and a foreign one produced by foreign MNEs. The price of these two varieties can differ because their costs of production vary between national firms and foreign MNEs of the same sector. Thus, we abandon the equal-costs assumption for national firms and foreign MNEs across sectors, which is present in most of the CGEs including MNEs ^[3] (e. g. Petri et al. 2012; Petri/Plummer 2016; Ciuriak et al. 2016). In our case, the intermediate input mix is the same between both types of firms within the same sector, but their labour and capital use is different. Hence, in our model, we split sectors into the two types of firms, using real data on the shares of production, labour and capital displayed at Table 2. As result, the cost structure differs between national firms and foreign MNEs in each sector.

Table 2:

Foreign MNEs’ shares in production, remuneration of employees and rental rate of capital in Spain (2005).

1. Sources: INE (2013b), Eurostat (2012) and OECD (2012).

3.3 Representative household and consumption

There is a representative consumer household that behaves as a rational consumer. The level of consumer’s welfare is determined by the endowments of capital and labour jointly with exogenous net transfers paid by the public sector.

The fixed endowment of labour should be interpreted as a maximum supply of labour since leisure (and unemployment in some versions of the model) is assumed to be endogenous. Hence, labour supply would be elastic up to the endowment constraint. The fixed endowment of capital is supplied to all sectors except to Public services, which only utilizes public fixed capital consumption.

The household’s decision problem consists of choosing an optimal consumption bundle, by maximizing a nested utility function subject to the budget constraint. As shown in Figure 4, preferences are represented by a nested utility function on (consumption of) goods, leisure and savings. Notice that, given our static approach, we consider a unit elasticity of substitution between savings and (consumption of) goods (Howe 1975), so that savings can be interpreted as the purchase of bonds for future consumption. The representative consumer buys all the final consumption goods, but the good from sector Public services. Goods can be subject to several types of indirect taxes.

Figure 4:

Welfare function nests.

The budget constraint includes total factor rents, including all the capital income for MNEs, jointly with exogenous net transfers paid by the public sector. Demand functions for goods, leisure and savings are derived from the first-order conditions and are included in the goods and factor markets equations, as well as in the macroeconomic closure for savings.

3.4 Public sector

The role of the public sector in the model is twofold, i. e. it is an owner of resources (e. g. from its capital endowment and tax revenues), and a purchaser of certain goods. We deal with these two functions in turn.

As an owner of resources, public sector’s wealth includes income from capital rents, tax revenues, and net transfers from the representative household. Capital rents of the public sector include, by definition (see Eurostat 2013), the fixed capital consumption because the net operating surplus is zero for the public sector. The fixed capital consumption has been assigned to sectors Other services and Public services. All capital in Public services is owned by the public sector, whereas in Other services some capital is publicly owned and the rest is private, according to National Accounts data (see Section 4). Taxes consist of social contributions paid by both employers and employees and net indirect taxes. All of them have been modelled using actual ad valorem rates calibrated from benchmark data, with an endogenous revenue level. The rest of taxes have been modelled as exogenous.

The public sector also enters the model as a purchaser of goods. Public sector expenditure includes both market (i. e. output that is disposed of in the market at economically significant prices) and non-market goods (i. e. output that is provided at prices that are not economically significant).

3.5 Foreign sector

The model incorporates the small open economy assumption, meaning that the country faces a perfectly elastic export supply function. There is also a constant elasticity of transformation function between national and foreign sales. Regarding imports, we assume that goods are differentiated according to their origin (i. e. national or foreign), following Armington’s assumption (Armington 1969), which allows for the possibility of intra-industry trade despite the assumption of exogenous world prices.

The foreign sector is closed by assuming that the difference between receipts and payments from the rest of the world is exogenous. This constraint would avoid, e. g. a permanent increase in exports with no change in imports, an unlikely scenario since it would involve an unlimited capital inflow to the country. However, this requires a matching movement in trade flows.

3.6 Factor markets

Two factors enter into the model: capital and labour. Regarding capital, both the private and the public sector own fixed endowments. Capital rents adjust to clear the national capital market, under the assumptions of capital international immobility, and no mobility across national sectors nor types of firms. Capital is specific in three levels, i. e. each sector employs different specific capital, and capital is also differentiated according to ownership (i. e. national and foreign) and type of assets (ICT and non-ICT).

The only owner of labour is the representative household. The demand for leisure is derived from the household’s optimization problem. Hence, labour supply (i. e. the labour endowment less the demand for leisure) would be elastic up to the fixed amount of labour. Labour is assumed to be internationally immobile, but mobile across national sectors.

There are two versions of the model with respect to labour. The first one is the standard perfect competition labour market. In addition, we assume a second version based on the wage equation (Blanchflower/Oswald 1990, 1994). It has been used in CGE models; see, e. g. Rutherford et al. (2002). Accordingly, the model includes the following constraint:

where W/CPI denotes the real wage, U is the unemployment rate, U0‾ is the unemployment rate in the benchmark, and β is a nonnegative parameter that measures the sensitivity of real wages to the rate of unemployment. Thus, as β approaches infinity, the real wage approaches its benchmark value (which is 1 according to the calibration process explained below): this is the case of rigid real wages when wages do not change when unemployment does. At the other extreme, as β approaches zero, the unemployment rate approaches its benchmark value, with real wages being flexible. Other intermediate values for β would mean different flexibility levels of real wages to the unemployment rate.

3.7 Macroeconomic closure

Total investment is split into sectoral gross capital formation using a fixed-coefficients Leontief structure (Dervis et al. 1981). Notice that, in our static framework, total gross capital formation affects the economy as a component of final demand. The model embodies a macroeconomic closure equation stating that investment and savings (private, public, and foreign) are equal.

Finally, the model is solved as explained in Rutherford (1999), with the general equilibrium model defined as a mixed complementarity problem (see Mathiesen 1985). The software used in the empirical application is GAMS/MPSGE.

5.1 Definition of scenarios

Simulations have been performed under six labour market scenarios, where the first five involve a labour market with unemployment, and the last one displays a perfect competition labour market. Specifically, these scenarios are:

1. A very rigid real wage with respect to unemployment rate (β=20).

2. A rigid real wage with respect to unemployment rate (β=3).

3. A Reference scenario with a plausible sensibility of real wage with respect to unemployment rate (β=1.5).

4. A flexible real wage with respect to unemployment rate (β=0.1).

5. A very flexible real wage with respect to unemployment rate (β=0.001).

6. No unemployment.

6.1 Macroeconomics results

The impact on the main macroeconomic variables appears in Table 3. The variables displayed are Gross Domestic Product (GDP), real wages, employment, unemployment rate and welfare (the Hicksian Equivalent Variations index). Panel A offers the impact of the actual levels of ICT Capital Stock growth for the period 2005–2014. Panel B displays the impact of a 1 % change in the ICT Capital Stock. Both cases are expansionary in macroeconomic terms, motivated by the increase in capital endowment. Hence, we can forecast ex-ante that new capital can substitute labour and/or complement it.

The effects on GDP, employment and welfare are going to be positive and unemployment would decrease. The magnitude of the changes is higher for the national firms’ investment in ICT assets than for foreign MNEs investment. Also, the higher the wage rigidity, the larger the positive effects in employment, GDP and welfare. The change in GDP is addressed by changes in factor employment (i. e. capital and employed labour) and in factor prices. With respect to capital, its contribution to GDP growth through the expansion in capital endowment was compensated with the fall in the real capital rent. In fact, this price fall took place in Spain for ICT assets with respect to other components of GFCF (see Mas et al. 2015: p. 12). This is a worldwide effect due to the decline of the price for computer and digital equipment. With respect to labour, the GDP increases due to both the increase in employment (i. e. lower unemployment and leisure) and the increase in real wages. This overall positive effect in GDP is especially relevant in the scenarios with more rigid wages where the real wage moderation generates larger employment growth than with more flexible wages. Notably, the increase in labour productivity due to the new ICT capital always generates wage increases, although when wages are more flexible, the workers benefit more from wage gains, than from new jobs.

In order to disentangle the role of firms’ ownership in the effects, the 1 % shock clarifies that the positive effect of ICT investment is always higher if it is a national firm’s investment. In all scenarios, with flexible or rigid wages, with unemployment or with full employment, the positive effects on GDP, real wages, employment, unemployment rate and welfare are larger if ICT investment is accomplished by national firms. The main characteristics of the labour market (i. e. wage rigidity or flexibility) are relevant for the size of the effect, but not for the prevalence of national firms’ ICT investment effects. The panel A showing the real 2005–2014 shock also displays a much higher effect of national firms versus foreign MNEs, but the asymmetric sectoral shock (i. e. the size of the shock is presented in column 7 at Table 1) could interfere in this result. From the comparison of the two panels another conclusion can be inferred: if real 2005–2014 ICT investment flows would be done by national firms, the gains for the Spanish economy as a whole (in terms of GDP, employment, real wages, unemployment and welfare) should be around twice the gains if the ICT investment is performed by foreign MNEs. The rationale is that foreign MNEs are more capital intensive, and the labour productivity gains derived from the shock are lower.

If we proxy the changes in employment for the Reference scenario with the data for Spain from the Labour Force Survey and National Accounts, we can conclude that a 1 % growth in ICT investment in all sectors executed by national firms would increase employment by 3649 jobs (0.019 %). The change would be 3457 jobs (0.018 %) when foreign MNEs are the investors. The change in GDP would be 316.4 millions of euros and 307.1 millions, respectively.

Those results, even generated by simulation, support that the assertion “MNEs are not only outgrowing firms in all other industries, but also disrupting traditional patterns of job creation” (UNCTAD 2017, p. 159) would be true for the Spanish case, assuming that national firms are performing closer to those traditional patterns. Moreover, “although Tech MNEs are creating more employment as they grow, sources of corporate value are shifting from labour to capital” and “this [MNEs’] employment creation is (…) significantly lower than the increase in total assets” (ibid., p. 162) are compatible with our results. Even, the role of intangibles (see Section 2) is more prominent in large MNEs and “the focus is moving toward capital components such as intangibles and cash, which generate relatively little employment” (ibid., p. 162).

6.2 Microeconomic results

Next, we present in Table 4 he results across sectors for the two most relevant variables, namely, employment and output (in physical units) and we also analyse the goods demand side through final consumption (in real value). In order to interpret these sectoral results, it is necessary to take into account the constraints the model imposes on productive factors, i. e. labour is modelled under an unemployment rule in five scenarios and leisure can take place in all scenarios, but capital is assumed fully employed and specific at three levels: sector, ownership (i. e. national/foreign) and type of asset (i. e. ICT/Non-ICT). Notice also that in a general equilibrium framework the results can be driven by several forces which, in some cases, move in opposite directions.

We provide in Table 4 an overview of the evolution of both production and employment across sectors, with the Reference scenario, which uses a central value for the elasticity of wage adjustment (β=1.5). The size of the real shock 2005–2014 is not positive for all the sectors, as seen in Table 1. Electronics, Construction, Financial intermediation, and Activities auxiliary to financial intermediation share a negative ICT capital flow, although we have modelled for them a zero change to avoid the potential negative use of capital as input. Nevertheless, we do not observe a large positive nor negative correlation between change in capital (column 7 in Table 1) and change in labour at sectoral level (first two columns in Table 4). The non-weighted correlations are −0.14 for national investment and −0.15 for the foreign one. The overall effect is job creating (see Table 3) showing complementarity, but also those small negative correlations show that substitution is taking place.

There are several sectors with negative effect on employment: Telecommunications, Renting of machinery & equipment and Computer and related activities. Moreover, Other business activities also has a negative effect on employment when the investment is performed by national firms. All those negatively affected sectors are related with Services. Those changes are explained by changes in productivity at the sectoral level due to the entrance of capital. The case of the homogeneous shock of 1 % change in ICT Capital Stock for all sectors also generates some employment negative effects more extended to Services sectors: Besides Telecommunications, Renting of machinery & equipment and Computer and related activities, also Financial intermediation and Activities auxiliary to financial intermediation have a negative employment change. In the same way, Autor et al. (2015) show for the US case that non-manufactures are more negatively affected from the incorporation of new technologies than manufactures since the 1990s. With the 1 % shock, sectors as Real estate activities, Energy, Beverages & Tobacco, Textile products and Air and water transport are the most outstanding as employment generators in percentage.

In any case, although the volume of factors, labour and capital, has decreased in several sectors, all the sectors present an increase in physical output production. This increase is larger in the sectors where employment has lowered (i. e. Telecommunications, Renting of machinery & equipment and Computer and related activities for the 2005–2014 shock, and the same sectors plus Financial intermediation and Activities auxiliary to financial intermediation for the 1 % shock). This means that the entrance of digital capital here would boost output with the collateral effect of reducing employment.

Finally, we provide some comments on final consumption, which is also reflected in Table 4. There is a product demand effect. The lower prices involve a raise in real income, so there is going to be an income effect with new product demand which is going to generate additional labour demand. This increase in real income would be spent in all kind of goods: form low-tech non-tradables to the rest of goods. This effect is captured in our general equilibrium model. The gain in real income (as reflected in the gain in GDP) is going to expand final consumption in all the goods, both in the real 2005–2014 shock and in the 1 % shock. Gregory et al. (2016) also note that this effect depends on where the income is spent. The effect would be lower, for example, if foreign MNEs repatriate capital income to their home economies (see Latorre (2009) for an example of this type of effect of profit repatriation in a CGE context). We have assumed that all the income generated in Spain is spent following the Spanish representative household pattern of consumption, so most of the income is spent in national goods. Nevertheless, we can add a new insight: it is also relevant for final consumption what type of firm is generating the change in income. Results in final consumption for the 1 % shock display that the effect in final consumption is larger when the ICT investment is performed by national firms. The reason is that the change in goods prices benefits more the consumers when the investors are national firms.

Next, we check the role of wage flexibility in the sectoral employment creation according to the six scenarios specified, from the most rigid framework to the most flexible wages, including a full employment scenario. Tables 5 and 6 reflect the sectoral employment effects when the ICT investment is performed by national firms (Panel A) or foreign MNEs (Panel B). Table 5 groups the effects for the 2005–2014 change in ICT Capital Stock and Table 6 for the 1 % change in ICT Capital Stock.

The sectoral employment changes reported in Tables 5 and 6 convey a multifaceted picture. There are wide sectoral asymmetric effects, with sectors positively and negatively affected. With respect to Panel A in Table 5, firstly, the more flexible the wages, the more sectors are negatively affected. The higher increase in wages (see Table 3) for flexible wages scenarios is behind this negative effect. It is noteworthy that the negative effect is highly concentrated in non-manufactures. For example, the negative employment effects for the full employment scenario are in nine sectors out of 25 (from the most to the least negative effect): Computer and related activities, Telecommunications, Renting of machinery & equipment, Other business activities, Energy, Wholesale and retail trade, Primary goods, Other transport and Food products. Note that most of them are related to Services. Secondly, not only the number of negatively affected sectors increases with the flexibility of wages, also the size of the negative effect rises (and it is reduced for the sectors positively affected). Thirdly, the sectoral range of the changes also decreases with the flexibility of wages, with a lower variance of the results. Finally, Panel B in Table 5 (i. e. ICT investment by foreign MNEs) shows a similar pattern with two differences: (i) The positively affected sectors have a smaller positive effect while negatively affected sectors have a larger negative effect; (ii) the eight negatively affected sectors are not exactly the same for the most flexible wages scenarios. Again, there is an employment decrease for seven sectors: Computer and related activities, Telecommunications, Renting of machinery & equipment, Energy, Wholesale and retail trade, Primary goods and Food products. But now Sale & repair of motor vehicles exhibits employment decreases, while Other business activities and Other transport now manifest employment growth.

Table 6 (i. e., 1 % change in ICT Capital Stock) allows to analyse the effects isolated from the sectoral asymmetric shock from the 2005–2014 change in ICT Capital Stock. Some of the characteristics of these results are common to the ones explained in the previous paragraph for the 2005–2014 shock:

1. –

   Sectoral asymmetric effects.

2. –

   The more flexible the wages, the more sectors are negatively affected.

3. –

   The increase in the size of the negative effect with more wage flexibility and, at the same time, the decrease in the size of the effect for the positively affected sectors.

4. –

   Most of the negatively affected sectors are Services while manufactures exhibit a pattern of job creation with ICT investment.

But the comparison between Panel A and B in Table 6 provides new insights on the relevance of firms’ ownership. Firstly, the sectoral variance of the results is wider when the ICT investment is performed by MNEs with larger negative and positive sectoral effects. And secondly, the number of negatively affected sectors is smaller with foreign MNEs’ ICT investment with flexible wages and tends to be larger with rigid wages. The scenario with very rigid wages shows 4 sectors with negative change in employment, while the very flexible wages scenario (and the full employment scenario) has 7 sectors with lower employment. In Panel A, with national firms’ ICT investment, the scenario with very rigid wages shows 3 sectors with negative change in employment, while the very flexible wages scenario has 9 sectors.