An econometric analysis of foreign direct investment in Spain, 1964-89.

I. Introduction

Foreign direct investment (FDI hereafter) has been deemed as one of the main factors underlying the strong growth rates experienced by the Spanish economy for the last thirty years. In fact, the rising inflows of foreign capital made possible by the liberalizing measures introduced since 1959 allowed the deep transformation of an economy until then characterized by its relative backwardness and an almost total lack of openness, being one of the most important channels leading to the integration of the Spanish economy in world markets. In addition, the Spanish accession to the European Community (EC) in 1986 has involved a new boost in FDI, that would reflect the favorable prospects for the country's economic future faced with the challenges of the Single European Market.(1)

Despite the crucial role played by FDI in the Spanish economy, the available empirical evidence is rather scant, being in general of a descriptive nature, and (specially those studies with an inter-industry scope) sometimes partial, in the sense of examining only one aspect (such as trade or profitability) in its relation with foreign investment. In Bajo-Rubio [4] a first attempt was made in order to test in a systematic way the main hypotheses proposed in the literature to explain the determinants of FDI, both from a macroeconomic and sectorial point of view. The aim of this paper is to provide some more robust evidence on the hypotheses tested in the first part of that paper (i.e., the allocation over time of gross aggregate FDI inflows in the Spanish economy for the 1964-89 period) using cointegration analysis.

Before proceeding further, it could be of interest to describe the FD1 series used in the paper. Hence, in the first column of Table I we show the balance of payments data on total gross FDI inflows received by the Spanish economy for the 1961-70, 1971-73, 1974-77, 1978-85 and 1986-89 periods. This table also reports the Gross Domestic Product (GDP) and Gross Capital Formation (GCF), as well as the ratios of total FDI inflows to the latter variables. As can be seen, there is a growing trend in both shares (broken only during the 1974-77 period, in which political instability was dominant, coinciding with the transition to a democratic regime). The shares are even more impressive after the Spanish integration into the EC in 1986.

Table I. Total FDI Inflows in Spain, 1961-89 (Gross Investment, Billion
Pesetas)

             FDI        GDP      FDI/GDP x 100      GCF      FDI/GCF x 100

1961-70      82.5     16008.8         0.46         4220.2         1.95
1971-73      65.3     10627.8         0.61         2769.4         2.35
1974-77     104.8     27597.1         0.38         7512.9         1.39
1978-85     979.6    151408.5         0.65        31285.5         3.13
1986-89    2263.3    151889.1         1.49        35105.0         6.45
1961-89    3495.5    341682.6         1.02        80893.0         4.32

Source: Spain's Balance of Payments, Corrales and Taguas [17] and own
elaboration.
Table II. Total FDI Inflows in Spain, 1961-89: Manufacturing and
Non-manufacturing Investment (Gross Investment, Billion Pesetas)

             FDI      MFDI     MFDI/FDI x 100     NFDI   NFDI/FDI x 100

1961-70      82.5      58.6        71.0            23.9       28.9
1971-73      65.3      53.1        81.3            12.2       18.7
1974-77     104.8      78.7        75.1            26.1       24.9
1978-85     979.6     622.9        63.6           356.7       36.4
1986-89    2263.3    1022.5        45.2          1240.8       54.8
1961-89    3495.5    1835.8        52.5          1659.7       47.5

Note: MFDI and NFDI denote manufacturing and non-manufacturing FDI inflows,
respectively.

Source: Spain's Balance of Payments and own elaboration.

Next, the allocation of total FDI between manufacturing and non-manufacturing activities is shown in Table II. As can be seen, until recently manufacturing has been the major recipient of FDI inflows (accounting for around 70 per cent of total FDI), mainly in the chemical, motor vehicles and machinery sectors. However, its share has decreased steadily over the last decade, especially after 1986, until it stood around 45 per cent on average during the 1986-89 period. This decline has been explained by the rise of FDI in non-manufacturing activities, such as real estate, finance and insurance.

Finally, from Table III we can see the geographical distribution of total FDI between its two main sources: the EC and the United States. It is apparent that there has been a redistribution in the geographical origin of FDI in recent years: even though the shares of the EC and the US in total FDI were similar during the 1960s and the 1970s, the former has increased steadily over the 1980s and in particular after the Spanish integration into the EC, so in the last few years more than a half of total FDI has come from the EC. The US share, on the contrary, has fallen remarkably, now accounting on average for hardly 5 per cent of total FDI.

The rest of the paper is organized as follows. In section II we outline a simple theoretical model of plant and equipment expenditure by a multinational enterprise (MNE). Section III presents the empirical results. Finally, section IV offers the main conclusions and policy implications derived from the paper.

Table III. Total FDI Inflows in Spain, 1961-89: Investment from the EC and the
US (Gross Investment, Billion Pesetas)

             FDI      EFDI    EFDI/FDI x 100     USFDI    USFDI/FDI x 100

1961-70      82.5      27.5        33.4           27.3         33.1
1971-73      65.3      26.3        40.3           22.9         35.0
1974-77     104.8      42.7        40.8           31.7         30.3
1978-85     979.6     401.7        41.0          195.3         19.9
1986-89    2263.3    1309.8        57.9          123.8          5.5
1961-89    3495.5    1808.0        51.7          401.0         11.5

Note: EFDI and USFDI denote FDI inflows coming from the EC and the US,
respectively.

Source: Spain's Balance of Payments and own elaboration.

II. Theoretical Model

The aim of this section is to briefly outline a simple theoretical model of investment expenditures by a MNE, which will be used as a benchmark for the empirical exercise performed in section III.(2)

The structure of the model is similar to those developed, among others, by Ray [46], Goldsbrough [31], Barrell and Pain [10] and Stevens and Lipsey [52], and will try to stress the role played by demand and relative costs when determining production and capital allocation by a MNE.

We begin by assuming that the first choice a multinational producer must face is whether or not to undertake FDI, which would involve the choice of an output level in the foreign country. The cost function faced by the firm has two components, associated with producing in the domestic and foreign plants, respectively:

C = [c.sub.d]([Q.sub.d])[Q.sub.d] + [c.sub.f]([Q.sub.f])[Q.sub.f], (1)

where C and c denote total and unit costs, Q is the level of output, and the subscripts d and f refer to the domestic and foreign variables, respectively. Then, the firm would minimize (1) subject to the constraint that output should equal total demand, [Mathematical Expression Omitted]:

[Mathematical Expression Omitted].

Defining the Lagrangean function

[Mathematical Expression Omitted],

and differentiating it with respect to [Q.sub.d], [Q.sub.f] and [Lambda] we obtain the following necessary conditions for the solution of the constrained optimization problem:

[Delta]L/[Delta][Q.sub.d] = [c[prime].sub.d] + [c.sub.d]([Q.sub.d]) - [Lambda] = 0, (4)

[Delta]L/[Delta][Q.sub.f] = [c[prime].sub.f][Q.sub.f] + [c.sub.f]([Q.sub.f]) - [Lambda] = 0, (5)

[Mathematical Expression Omitted],

where [c[prime].sub.d] = d[c.sub.d]/d[Q.sub.d] and [c[prime].sub.f] = d[c.sub.f]/d[Q.sub.f]. From (4) and (5) we have

[c[prime].sub.d][Q.sub.d] + [c.sub.d] = [c[prime].sub.f][Q.sub.f] + [c.sub.f],

which is the familiar condition, for a producer distributing his output between two plants (domestic and foreign, in this case), of equalization of marginal costs in both plants.

Solving for [Q.sub.f], after replacing [Q.sub.d] from (6), it follows that the equilibrium level of production in the foreign plant is given by

[Mathematical Expression Omitted],

where [[Gamma].sub.1] = [c[prime].sub.d]/([c[prime].sub.d] + [c[prime].sub.f]) and [[Gamma].sub.2] = 1/([c[prime].sub.d] + [c[prime].sub.f]) are both assumed to be positive.(3) That is, the output to be produced at the foreign plant is positively related to total demand and negatively related to their unit costs relative to those of the home country plant.

However, this is not the whole story. Once the decision of producing abroad has been made, the MNE must face a second choice involving input substitution within the foreign plant. Assuming for simplicity that production in the foreign plant takes place using two inputs, labor (L) and capital (K), by means of a Cobb-Douglas technology, the foreign subsidiary would minimize its total costs

[C.sub.f] = [w.sub.f][L.sub.f] + [q.sub.f][K.sub.f], (8)

(where w and q denote, respectively, the wage rate and the user cost of capital, both in real terms), subject to the constraint given by the production function, in Cobb-Douglas form

[Mathematical Expression Omitted].

The Lagrangean

[Mathematical Expression Omitted],

gives first-order conditions

[Delta]L/[Delta][L.sub.f] = [w.sub.f] - [Lambda][Alpha]([Q.sub.f]/[L.sub.f]) = 0, (11)

[Delta]L/[Delta][K.sub.f] = [q.sub.f] - [Lambda][Beta]([Q.sub.f]/[K.sub.f]) = 0, (12)

[Mathematical Expression Omitted].

Again, from (11) and (12) we can obtain

([w.sub.f][L.sub.f])/([Alpha][Q.sub.f]) = ([q.sub.f][K.sub.f])/([Beta][Q.sub.f])

so that, replacing [L.sub.f] from (13) and solving for [K.sub.f] yields

[Mathematical Expression Omitted].

Replacing further [Q.sub.f] from (7), we obtain the final expression for the subsidiary's desired capital stock when producing in the foreign plant:

[Mathematical Expression Omitted].

As can be seen from (15), this desired capital stock appears positively related to total demand, as an indicator of the profitability of FDI, and negatively related to the host country's unit costs, relative to those of the home country. Notice, however, that now the effect of relative unit costs is not unambiguous in the case of labor, when substitution between inputs is taken into account: an increase in wages in the host country could lead to a higher (instead of lower) capital stock if a strong substitution effect between labor and capital is in effect in that country (see Cushman [19] for a more detailed account).(4)

Equation (15) might be augmented by introducing the effect of trade barriers in the host country by means of an additional term in the cost function (1), that would lead to a positive relationship with [Q.sub.f] (and hence with [K.sub.f]): high tariff barriers in the host country would mean an incentive for the firms wishing to gain access to that market, which would settle there by means of FDI in order to overcome such barriers.(5)

From the previous discussion, omitting subscripts and aggregating across foreign subsidiaries, we can write an expression for the desired stock of foreign capital in the host country, denoted by K*, as

[Mathematical Expression Omitted]

where K* would depend positively on the level of aggregate demand (AD); negatively on the relative (host country vis-a-vis home country) unit costs (RUC) (unless there were a strong substitution effect between capital and labor, as previously mentioned--see note 4); and positively on the level of trade barriers (T).

Notice finally that the level of FDI in a period would be the result of accumulated past decisions to modify the desired capital stock, so that, due to adjustment costs and operating lags, we can write

[FDI.sub.t] = [summation of][[Lambda].sub.i]([K*.sub.t-i] - [K*.sub.t-i-1]) where i=0 to n + [Delta][K.sub.t-1] (17)

where

[summation of][[Lambda].sub.i] = 1 where i=0 to n.

Equation (17) may be written in a form suitable for estimation by using the so-called Koyck transformation (see, e.g., Coen [15]), assuming that the lags decline at a constant geometric rate:

[[Lambda].sub.i] = [Lambda][(1 - [Lambda]).sup.i].

so that

[FDI.sub.t] = [Lambda]([K*.sub.t] - [K.sub.t-1]) + [Delta][K.sub.t-1]

or, alternatively,

[FDI.sub.t] = [Lambda][K.sub.t]* + ([Delta] - [Lambda])[K.sub.t-1]. (18)

Therefore, FDI would be a function of the determinants of [K*.sub.t], as well as of the stock of foreign capital at the beginning of the period ([K.sub.t-1]).

The basis for our empirical analysis will be equations (16) and (18), which have been extended in order to take into account three additional effects. First, it should be noted that macroeconomic instability, reflecting the presence of internal economic pressures and unability to balance the budget or restrict money supply, might affect profitability expectations of international investors. We have used the inflation rate as a proxy for the degree of macroeconomic instability, and the uncertainty associated with it. This relationship between inflation and uncertainty has been stressed in the literature on the costs of inflation: so, by changing the pattern of asset accumulation from financial to real assets, and by shortening the effective maturity of the contracts, uncertainty associated with inflation would tend to reduce the rate of investment (see Fischer and Modigliani [27]). Then, the Spanish rate of inflation has been used as an indicator of macroeconomic instability and uncertainty, in the sense that a lower inflation rate should mean a better climate for foreign investment, so favoring FDI inflows.

Second, the integration of the Spanish economy into the EC since January 1st 1986 should have meant a change in the expectations held by foreign investors in Spain: even though the lowering in trade barriers might lead to prefer exports instead of FDI, it can also be argued that foreign investors might expect better prospects for an economy definitively integrated with the other European countries, specially in view to the 1993 Single European Market and maybe using Spain as an exporting platform for supplying other European markets. We have proxied the effects from the Spanish integration into the EC by a dummy variable for the years of the Spanish membership into the EC.

Third, we have also examined the possible influence of the exchange rate (necessary to translate relative costs into a common currency) on FDI. This is a recurrent issue in the literature on FDI since the seminal contribution by Aliber [2]. According to this author, the pattern of FDI would reflect the fact that the MNE capitalizes the same flow of expected returns to a higher rate than the host country's firms, which would be explained by the relative weakness of those countries' currencies. Therefore, their firms would be in disadvantage as compared to those from the investor countries, since they should borrow at an interest rate which includes a risk premium for a possible depreciation of their currencies; hence, the MNEs would face their investments with a lower financial cost because they would have not to pay that premium.

Later on, the role played by the exchange rate has been embodied in different contributions, both theoretical and empirical, having taken a renewed interest with the more elaborated theory recently proposed by Froot and Stein [28]. According to this theory, and due to the existence of informational imperfections in the capital market, a low real value of the domestic currency would be associated with FDI inflows, since this depreciation would increase the relative wealth position of foreigners and hence might lead them to acquire certain home assets.(6)

III. Empirical Results

This section will present the results from estimating a modified version of equations (16) and (18). From the discussion above, the model developed in section II can be rewritten in log-linear form as follows:

[Mathematical Expression Omitted],

where L denotes natural logarithm, and the expected signs of the coefficients are

[[Theta].sub.1] [is greater than] 0, [[Theta].sub.2] [is less than] 0, [Mathematical Expression Omitted], [[Theta].sub.4] [is less than] 0, [[Theta].sub.5] [is greater than] 0, [[Theta].sub.6] [is less than] 0, [[Theta].sub.7] [is greater than] 0, [Mathematical Expression Omitted].

From expression (19), FDI inflows in Spain would be related to:

a) The Spanish Gross Domestic Product (GDP), proxying the size of the Spanish market, positively. This would reflect the so-called "market size hypothesis" (quite common in empirical studies of FDI), which assumes a positive association between FDI and the expected sales of foreign subsidiaries in the host country, as an indicator of the profitability of FDI. On the other hand, a non-significant relationship between FDI and market size would mean that foreign firms would be more concerned with exports than with supplying domestic markets.

b) The Spanish inflation rate (1NF), as a proxy for macroeconomic instability, negatively, as discussed above.

c) The Spanish unit labor costs (ULC), which would be in principle negatively related to FDI inflows, unless a strong substitution effect between labor and capital were in operation, in which case the effect would be reversed as was argued in section II.(7)

d) The Spanish user cost of capital (UCC) which should be negatively related to FDI inflows.(8)

e) A measure of trade barriers (PROT), positively (see the discussion in section II).

f) The real effective exchange rate of the Peseta against the industrialized countries (REER), negatively. Notice that this index of effective exchange rate is built so that a decrease would mean a depreciation, which would translate into increased FDI inflows, as was shown in section II.

g) A dummy variable for the years after the Spanish integration into the EC (DEC), which would reflect the expectations associated with EC membership and would be expected to be positively related to FDI inflows.

h) The lagged value of the foreign capital stock ([K.sub.-1]). Since there are no direct estimations of this variable in the Spanish economy, it has been proxied by the accumulated sum of gross FDI inflows since 1964, net of disinvestment payments.(9)

Regarding the dependent variable, we used the gross FDI inflows received by the Spanish economy, according to balance of payments figures, in real terms.(10) In addition, we have analyzed manufacturing and non-manufacturing FDI, and that coming from the EC and the US, also in real terms (denoted by MFDI, NFDI, EFDI and USFDI, respectively).

Even though FDI data are available since 1961, both the reliability of these data for the first years and the lack of data before 1964 for several explanatory variables led us to use as our period of analysis the years 1964 through 1989.(11) On the other hand, since data on PROT are not available for 1989, estimated equations including that variable refer to the 1964-88 period. The exact definition of the variables as well as the data sources are detailed in the Appendix.

The econometric methodology used in this paper is based on the so-called "cointegration analysis" [25], that has provided further support for the error correction model (ECM) and has greatly enhanced the approach to non-stationary time series. The literature on cointegration and unit roots is surveyed in Dolado, Jenkinson and Sosvilla-Rivero [23] or Campbell and Perron [12].

As it is well known, a previous step in cointegration analysis consists of testing the order of integration of the variables. Several statistical tests for unit roots have been developed to test for stationarity in time series. In this paper we made use of the Dickey and Fuller [21] tests, and the results are shown in Table IV for the first differences and levels of the variables. As can be seen in part B of the table, for the level variables the null hypothesis that a given series contains a unit root was rejected only for the FDI coming from the US, and for our proxies for the capital stock in non-manufacturing and US FDI.(12) On the other hand, as seen in part A of the table, the first differences of most of the variables were stationary.

Regarding the variable LGDP, the Dickey-Fuller tests would not reject the null hypothesis of a second unit root. However, as has been argued elsewhere [3], this variable would have been affected by several exogenous shocks during the period of analysis, which would not be realizations of the underlying data-generating mechanism. To test for this hypothesis, we made use of the test proposed by Perron [41], which allows for a shift in the intercept and/or slope of the trend function. In our case, the relevant tests correspond to Perron's model B, (i.e., his "changing growth" model), but with two changes in the slope of the trend of the series, occurring in the TABULAR DATA OMITTED years 1974 and 1985. The results, shown in Table V, allow us now clearly to reject the existence of a second unit root, so that we could conclude that LGDP would be also a I(1) variable.(13)

After determining the order of integration of the variables, we followed the two-step estimation procedure for dynamic modeling suggested by Engle and Granger [25]. So, in a first step the so-called "cointegrating regression", in which all the variables would be in levels and no dynamics included, would be estimated by ordinary least squares (OLS), and the residuals from this regression tested for the presence of a unit root. If the residuals were found to be stationary, the cointegrating regression might be taken as a long-run relationship and we could then proceed to the second step, where an ECM including those lagged residuals as an error-correction term would be postulated in order to model the short-run dynamics.

Table V. Perron Unit Tests for LGDP

A) First Difference

[Mathematical Expression Omitted]

[Mathematical Expression Omitted]

B) Level

[Mathematical Expression Omitted]

[Mathematical Expression Omitted]

Notes: t-statistics in parenthesis 

The additional variables in the equations are defined as follows:

TIME = time trend DU1 = 1 for 1975-89, 0 otherwise DU2 = 1 for 1986-89, 0 otherwise DT1 = TIME for 1975-89, 0 otherwise DT2 = TIME for 1986-89, 0 otherwise

Notice that even though the estimation by OLS of the cointegrating regression yields "super-consistent" estimates [53], the joint dependence of most aggregate time series and their non-stationarity invalidate the routine application of many standard statistical procedures. To overcome this problem, in the first step of the Engle-Granger procedure we alternatively applied the estimation method proposed by Phillips and Hansen [42]. This single-equation semiparametric method allows the direct estimation of the long-run relationship in a two-step procedure, filtering the data in the first step using a nonparametric correction.(14) These authors presented a class of Wald tests which are modified by semiparametric corrections for serial correlation and second-order endogeneity bias (see Banerjee, Dolado, Hendry and Smith [8]). The resulting test statistics (termed fully-modified Wald tests) have limiting [[Chi].sup.2] distributions and therefore allow inference to proceed in a conventional way.

The results of applying the Phillips-Hansen procedure to equation (19) for total FDI were as follows, where (since the model was estimated in logs) the estimated coefficients denote elasticities:

[Mathematical Expression Omitted]

[Mathematical Expression Omitted]

Notice that the figures in brackets below each coefficient are not the standard t-statistics but, as mentioned above, the Phillips and Hansen's fully-modified Wald statistics, distributed as a [[Chi].sup.2] with one degree of freedom.(15)

As can be seen in equation (20), a first variable appearing as highly influential is the size of the Spanish market, proxied by the level of GDP at constant prices. In fact, the market size hypothesis has proved to play a highly successful role in other previous studies, both in terms of levels, as in this paper [7; 49; 50; 31; 38; 39; 48; 18; 10] and in terms of growth [30; 47; 48; 18]. This result would be also in line with the recent models on economic geography, which tend to consider as a basic factor when locating economic activities the proximity to large markets, where consumption takes place [36].

Other variable showing a highly significant effect was inflation, suggesting that the uncertainty implied by a high rate of inflation would have hindered FDI.

The estimated coefficient on trade barriers showed the expected positive sign, which could be interpreted as foreign investors coming to the Spanish economy to overcome relatively high trade barriers, so giving further support to our former result on the importance of the Spanish market for FDI. Similar results for the case of the US FDI in the EC were obtained in Scaperlanda and Balough [48].

Finally, a negative and significant coefficient was also obtained for the lagged foreign capital stock.

With regard to unit labor costs, they appeared with a negative and non-significant coefficient so they were not included in the cointegrating regression. When relative labor costs between Spain and the rest of the world were jointly included in the regression, the coefficient turned out to be also negative, but non-significant. In fact, the evidence in the literature is not crystal clear, including both negative and significant effects from relative (host vis-a-vis home country) labor costs on FDI inflows (see, e.g., Goldsbrough [31], Cushman [19] or Barrell and Pain [10]), as well as inconclusive results (Culem [18]; Cushman [20]). Regarding the user costs of capital, their coefficients were positive and not significant either separately or jointly introduced.

We did not find a significant long-run relationship between total FDI inflows and our variable proxying integration into the EC, even though its coefficient showed the expected sign. As far as the real effective exchange rate, its coefficient was negative but not significant.

To conclude, note that in the last line of equation (20) we report, together with the coefficient of determination, several cointegration tests: the cointegrating regression Durbin-Watson and (augmented) Dickey-Fuller test statistics, as well as those proposed by Phillips and Ouliaris [43] (CRDW, CR(A)DF, and [Mathematical Expression Omitted], respectively). The results from these tests suggest that we can reject the null hypothesis of no-cointegration at least at the 5% level of significance, so equation (20) can be tentatively thought as representing a long-run relationship.(16)

Turning now to the components of total FDI inflows, namely manufacturing and non-manufacturing FDI, the results were now as follows:

[Mathematical Expression Omitted]

[Mathematical Expression Omitted].

As can be seen in equations (21) and (22), the results from manufacturing and non-manufacturing FDI are roughly similar than those for total FDI regarding the variables real GDP and inflation. However, we have found now a significant and negative effect for the user cost of capital in equation (21) and for the unit labor cost in equation (22). So, this fact might provide a rationale for the lack of significance of factor costs on total FDI, since the (significant) effects of capital and labor costs on manufacturing and non-manufacturing FDI could cancel on average. Notice that the lagged capital stock was not included in the equation for non-manufacturing FDI since it proved to be a stationary variable, although when introduced it also showed a negative and significant coefficient.

Finally, the former analysis was also applied to the FDI inflows coming from the EC which, as shown in Section I, accounted on average for more than a half of total FDI during the period considered.(17) The results in this case were the following:

[Mathematical Expression Omitted]

where the strongest effect is again that of real GDP; also significant are the coefficients on the inflation rate, the lagged capital stock, and the dummy variable proxying Spanish integration into the EC. The latter result is similar to that of Schmitz and Bieri [51], who found an increased FDI from the US to the EC at the time of its constitution. We were not able to find, however, a significant relationship with unit labor costs, the user cost of capital, trade barriers, or the real effective exchange rate. As in equations (21) and (22), the null hypothesis of no-cointegration was again rejected, suggesting that these equations may represent long-run relationships.(18)

Regarding the short run, and following the "general-to-specific" modeling methodology [34], an initially over-parameterized ECM with two lags on the dependent as well as the explanatory variables was continuously simplified and re-parameterized until a parsimonious representation of the data generation process was obtained. The results, obtained by OLS, are shown in Table VI. Note that the figures in brackets below each coefficient are now the standard t-statistics (since all the variables are stationary), and that the terms [Mathematical Expression Omitted] (i = 1, . . ., 4) denote the residuals from equations (20) to (23), respectively.

As can be seen in the table, the null hypothesis of no error correction was rejected in all cases, giving further support to the cointegrating equations (20) to (23) as long-run relationships [35]. The estimated coefficients show the expected signs, being also statistically significant. It should be noticed that we have also found a significant role for the real effective exchange rate and for the user cost of capital in the short-run equation for total FDI, as well as for trade barriers in the equation for manufacturing FDI.

Finally, we also report some diagnostic tests for normality, second-order autoregressive conditional heteroscedasticity, and second-order residual autocorrelation (N, ARCH, and LM, respectively), which do not show any sign of misspecification in those equations.

TABULAR DATA OMITTED

IV. Conclusions

We have examined throughout this paper the empirical relevance of several hypotheses put forward in the literature in order to explain the evolution of the gross aggregate FDI inflows received by the Spanish economy during the 1964-89 period. To this end, we made use of cointegration techniques, which allowed us to obtain robust and reliable estimates of the parameters in the empirical relationships.

After testing the order of integration of the variables, we found a long-run relationship between total gross FDI inflows and several macroeconomic variables: the level of real GDP, the rate of inflation, the level of trade barriers, and the lagged foreign capital stock.

When splitting total FDI into its two components of manufacturing and non-manufacturing, we found that their main determinants were roughly similar than those for total FDI, especially regarding the variables real GDP and inflation. However, we found significant and negative effects for the user cost of capital and the unit labor cost only for manufacturing and non-manufacturing FDI, respectively, and not for the total.

On the other hand, when examining FDI inflows coming from the EC, we found that they were also related to the level of real GDP and the inflation rate, as well as to the expectations associated with Spanish integration into the EC.

Finally, we also found evidence that the data could identify an appropriate error correction model for the short-run dynamics, hence giving further support to the cointegrating equations as long-run relationships.

The results of this paper allow us to draw some policy implications. First of all, it is confirmed that a stable and growing economy, in addition to the expectations created by the Single European Market for 1993, remain as necessary conditions to keep a high level of FDI inflows in Spain over the next years. In this sense, the recent deceleration in FDI would provide additional support to this hypothesis, given the slower rate of growth registered by the Spanish economy relative to that previously experienced.

On the other hand, unit labor costs proved to be significant only for non-manufacturing FDI, which would suggest that higher labor costs in Spain should not be taken as a serious threat to FDI growth, at least regarding manufacturing FDI. In addition to the substitution effect between labor and capital discussed in section II, this fact might also reveal a preference of foreign investors in the Spanish manufacturing industry for a skilled labor force, according to the often quoted argument [44] about the tendency for MNEs, when choosing a country to locate their investments, to attach a lower value to the availability of cheap unskilled labor and natural resources, so giving more weight to some other factors such as skilled scientific and technical personnel, as well as advanced infrastructure.(19) Then, it would seem that MNEs in Spain would have been more concerned with an expanding domestic market and a skilled labor force, than would have tended to prevail over the relatively advantageous Spanish labor costs. Furthermore, as Bentolila and Blanchard [9] argue, an excessive concern for reducing unit labor costs could worsen the socio-political climate, affecting the prospects for MNEs and making them reluctant to invest in Spain.

Further support to our hypothesis about the attractiveness of the Spanish economy for foreign investors would be provided by comparing the Spanish experience with that of Portugal after the accession of both economies to the EC. In this sense Portugal is a country with a smaller domestic market and lower labor skills, one which, despite its relatively lower wages, has attracted a much lower amount of FDI inflows than Spain since 1986 (and that directed mainly to traditional sectors) [16]. In contrast, the boom in FDI inflows to Spain after 1986 (mainly directed to more dynamic sectors) can be explained by the wider Spanish domestic market, together with a combination of relatively high skills and (still) low wages.

Appendix: Definition of the Variables and Data Sources

Dependent variables:

FDI, MFDI, NFDI, EFDI and USFDI are, respectively, the total gross FDI inflows received by the Spanish economy, those directed to manufacturing and non-manufacturing activities, and those coming from the EC and the US, expressed in real terms using the GDP deflator. Sources: Spain's Balance of Payments and Corrales and Taguas [17].

Explanatory variables:

GDP = Gross Domestic Product, at market prices, in billion Pesetas of 1980. Source: Corrales and Taguas [17].

INF = rate of change in GDP deflator, at market prices. Source: Corrales and Taguas [17].

ULC = index of Spanish real unit labor costs. Source: Eurostat.

UCC = Spanish user cost of capital in real terms, computed by the authors (following the procedure employed in Andres, Escribano, Molinas and Taguas [3]) as ([P.sub.l]/P)(i + 0.1 - [[Pi].sub.I]), where [P.sub.I], P, i, and [[Pi].sub.I] denote the deflator of private productive investment, the GDP deflator, the long-run nominal interest rate, and the rate of change in [P.sub.I] one period ahead, with 0.1 standing for the rate of depreciation. Source: Corrales and Taguas [17].

PROT = annual average of tariff duties plus border compensation of indirect taxes (the "compensating tax for internal imposition" or ICGI until 1985, and the "value added tax" or IVA since 1986), weighted by the value of imports. Source: Gamir [29].

REER = index of the real effective exchange rate of the Peseta against the industrialized countries. Source: Ministry of Economy and Finance's data bank.

DEC = dummy variable taking the value one for the year 1986 through 1989, and zero otherwise.

K, KM, KN, KE, and KUS = foreign capital stock for total, manufacturing, non-manufacturing, EC and US FDI, expressed in real terms using the GDP deflator. Sources: Spain's Balance of Payments and Corrales and Taguas [17].

1. See Bajo-Rubio and Tortes [6] for an evaluation of FDI trends after the Spanish integration into the EC.

2. See Agarwal [1] and Lizondo [37] for surveys of the literature on the determinants of FDI.

3. Omitting subscripts, we have c[prime] = (MC - c)/Q, where MC denotes the marginal cost. Provided that there are no increasing returns, MC [is greater than] c, and therefore [[Gamma].sub.i] [is greater than] 0, for i = 1, 2.

4. It can be shown that if an increase in unit labor costs in the host country (denoted by [Mathematical Expression Omitted]) is going to affect negatively the desired level of capital in that country, the condition required from equation (15) is [Mathematical Expression Omitted].

5. Formally, (1) should be augmented with the term t([Q.sub.d] - [D.sub.d]), where [Q.sub.d] - [D.sub.d] are the parent company's exports to the host country of the FDI (i.e., the difference between output and demand in the subsidiary's home country), and t is the tariff rate levied on those exports in that country. Then, (4) should be amended to [Mathematical Expression Omitted], which would yield the modified version of (7):

[Mathematical Expression Omitted].

Substitution of (7[prime]) into (14) would give us a new version of (15) in which t would appear as an additional determinant of [K.sub.f].

6. For a more detailed explanation, see Froot and Stein [28]. A recent survey on the models embodying the impact of the exchange rate on FDI can be found in Capel [13].

7. Even though, according to the theoretical model the relevant variable should be the relative (Spanish vis-a-vis foreign) unit labor costs, we report only the results obtained using Spanish costs since it could be argued the existence of correlation between the index of foreign unit labor costs and the real effective exchange rate, since the former are expressed in a common currency in relation to the Spanish ones. In any case, the econometric results when using both indices (Spanish and relative) proved to be quite similar in all cases.

8. In a similar way as in the labor cost case, we report only the results for the Spanish user cost of capital, given the lack of any comparable measure for the rest of the world. However, we have used some estimates of the net return on the net capital stock for the non-agricultural sector, computed by the EC and kindly provided by Samuel Bentolila, but its contribution never proved to be significant.

9. We are grateful to Antonio Carrascosa and an anonymous referee for suggesting us this proxy for the foreign capital stock. Note, on the other hand, that, given the policy of autarky followed by the Spanish authorities during the 1939-59 period, FDI inflows (and therefore the foreign capital stock) would have been negligible until the early 1960s.

10. Notice that, even though this variable is not strictly equivalent to foreign capital formation, the use of FDI figures from the balance of payments as a proxy variable for capital formation by foreign subsidiaries is a customary practice in the literature; see Hartman [33, 486] for a justification of this practice. For a detailed discussion of the difference between FDI and GCF in the Spanish case, see Carrascosa [14].

11. Notice that, even though our sample is not too long, simulations performed by Hakkio and Rush [32] suggest that, if we are interested in finding long-run relationships among a set of variables, there is not much difference between the information contained in annual data and that contained in quarterly or monthly data (if available).

12. The conclusion obtained for these three variables was confirmed by using an alternative non-parametric test for a unit root proposed by Boero and Burridge [11]. The results from this test are available from the authors upon request.

13. The relevant statistics would be the t-ratios on [Mathematical Expression Omitted] (i.e., -6.54 and -1.53 for the first difference and the level of LGDP, respectively). The critical values for two breaks in the slope of the trend are -5.45 and -4.76, at the 1% and 5% levels of significance, respectively [45].

14. A detailed explanation of the Phillips-Hansen estimation procedure can be found in the previous version of this paper [5].

15. The critical values for these tests are 2.71, 3.84 and 6.63 for the 10%, 5% and 1% significance levels, respectively.

16. The 5% critical values for the CRDW, CRADF and [Mathematical Expression Omitted] tests, for five variables in the cointegrating regression and 50 observations, are 1.10, -4.15 and -4.40, respectively [22; 26; 43].

17. Notice that we do not test for cointegration between US FDI and its determinants, since LUSFDI proved to be a stationary variable.

18. The 5% critical values for the CRDW, CR(A)DF and [Mathematical Expression Omitted] tests are, respectively: 1.10, -4.76 and -4.40 for equations (21) and (23); and 0.89, -3.75 and -3.74 for equation (22) (see references in note 16).

19. This would be supported by the highly significant effect found for the skill variables on the inter-industry allocation of manufacturing FDI in Bajo-Rubio [4]. On the other hand, labor costs did not seem to play any significant role in a similar study (making use of cluster analysis) by Egea-Roman and Lopez-Pueyo [24].

References

1. Agarwal, Jamuna P., "Determinants of Foreign Direct Investment: A Survey." Weltwirtschaftliches Archiv, No. 4, 1980, 739-73.

2. Aliber, Robert Z. "A Theory of Direct Foreign Investment," in The International Corporation, edited by Charles P. Kindleberger. Cambridge, Mass.: The MIT Press, 1970, pp. 17-34.

3. Andres, Javier, Alvaro Escribano, Cesar Molinas and David Taguas. La Inversion en Espana. Econometria con Restricciones de Equilibrio. Barcelona: Antoni Bosch, 1990.

4. Bajo-Rubio, Oscar. "Macroeconomic and Sectoral Determinants of Foreign Direct Investment in Spain: An Empirical Analysis." Paper presented at the V Congress of the European Economic Association, Lisbon, 1990.

5. ----- and Simon Sosvilla-Rivero. "Modelling Foreign Direct Investment in Spain, 1961-1988: An Application of Cointegration Analysis." Paper presented at the European Meeting of the Econometric Society, Cambridge, 1991.

6. ----- and Angel Torres. "External Trade and Foreign Direct Investment after the Spanish Integration into the EC (1986-90)," in The Spanish Economy before the European Single Market, edited by Jose Vinals. Oxford: Oxford University Press, forthcoming 1994.

7. Bandera, V. N. and J. T. White, "US Direct Investments and Domestic Markets in Europe." Economia Internazionale, February 1968, 117-33.

8. Banerjee, Anindya, Juan J. Dolado, David F. Hendry, and Gregor W. Smith, "Exploring Equilibrium Relationships in Econometrics through Static Models: Some Monte Carlo Evidence." Oxford Bulletin of Economics and Statistics, August 1986, 253-77.

9. Bentolila, Samuel and Olivier Blanchard, "Spanish Unemployment." Economic Policy, April 1990, 234-81.

10. Barrell, Ray and Nigel Pain. "An Econometric Analysis of US Foreign Direct Investment." Paper presented at the European Meeting of the Econometric Society, Cambridge, 1991.

11. Boero, Gianna and Peter Burridge. "A Simple Non-Parametric Test for a Unit Root." Discussion Paper 91-26, Department of Economics, The University of Birmingham, Birmingham, 1991.

12. Campbell, John Y. and Pierre Perron, "Pitfalls and Opportunities: What Macroeconomists Should Know about Unit Roots." NBER Macroeconomics Annual, 1992, pp. 141-201.

13. Capel, Jeannette, "The Exchange Rate Impact on Foreign Direct Investment. A Classification and Assessment of the Literature." Tinbergen Institute Research Bulletin, No. 2, 1990, 47-59.

14. Carrascosa, Antonio, "Inversion Extranjera e Inversion Productiva." Economia Industrial, July-August 1990, 173-79.

15. Coen, Robert M. "The Effect of Cash Flow on the Speed of Adjustment," in Tax Incentives and Capital Spending, edited by Gary Fromm. Washington, D.C.: The Brookings Institution, 1971, pp. 131-96.

16. Corado, Cristina and Antonio Leite. "Industrial Location and Foreign Investment in the Iberian Peninsula." Paper presented at the V Congress of the European Economic Association, Lisbon, 1990.

17. Corrales, Adolfo and David Taguas. "Series Macroeconomicas para el Periodo 1954-88: Un Intento de Homogeneizacion," in La Economia Espanola. Una Perspectiva Macroeconomica, edited by Cesar Molinas, Miguel Sebastian, and Antonio Zabalza. Barcelona: Antoni Bosch, 1991, pp. 583-646.

18. Culem, Claudy, "The Locational Determinants of Direct Investments among Industrialized Countries." European Economic Review, April 1988, 885-904.

19. Cushman, David O., "The Effects of Real Wages and Labor Productivity on Foreign Direct Investment." Southern Economic Journal, July 1987, 174-85.

20. -----, "Exchange-rate Uncertainty and Foreign Direct Investment in the United States." Weltwirtschaftliches Archiv, No. 2, 1988, 322-36.

21. Dickey, David and Wayne Fuller, "Distribution of the Estimators for Autoregressive Time Series with a Unit Root," Journal of the American Statistical Association, June 1979, 427-31.

22. Dolado, Juan J., "Cointegracion: Una Panoramica." Documento de Trabajo 8902, Banco de Espana, Madrid, 1989.

23. -----, Tim Jenkinson, and Simon Sosvilla-Rivero, "Cointegration and Unit Roots." Journal of Economic Surveys, No. 3, 1990, 249-73.

24. Egea-Roman, M. Pilar and Carmen Lopez-Pueyo, "Comportamiento Sectorial de la Inversion Extranjera Directa en Espana (1986-1989)." Informacion Comercial Espanola, August-September 1991, 75-92.

25. Engle, Robert F. and Clive W. J. Granger, "Cointegration and Error Correction: Representation, Estimation, and Testing." Econometrica, March 1987, 251-76.

26. ----- and Sam Yoo, "Forecasting and Testing in Cointegrated Systems." Journal of Econometrics, May 1987, 143-59.

27. Fischer, Stanley and Franco Modigliani, "Towards an Understanding of the Real Effects and Costs of Inflation." Weltwirtschaftliches Archiv, No. 4, 1978, 810-33.

28. Froot, Kenneth and Jeremy Stein, "Exchange Rates and Foreign Direct Investment: An Imperfect Capital Markets Approach." Quarterly Journal of Economics, November 1991, 1191-217.

29. Gamir, Luis, "Politica Arancelaria." Informacion Comercial Espanola, January 1990, 19-50.

30. Goldberg, Murray A., "The Determinants of US Direct Investment in the EEC: Comment." American Economic Review, September 1972, 692-99.

31. Goldsbrough, David, "The Role of Foreign Direct Investment in the External Adjustment Process." IMF Staff Papers, December 1979, 725-54.

32. Hakkio, Craig and Mark Rush, "Cointegration: How Short is the Long Run?" Journal of International Money and Finance, December 1991, 571-81.

33. Hartman, David G., "Tax Policy and Foreign Direct Investment in the United States." National Tax Journal, December 1984, 475-87.

34. Hendry, David, Adrian Pagan and Denis Sargan. "Dynamic Specification," in Handbook of Econometrics, Vol. 2, edited by Zvi Griliches and Michael Intriligator. Amsterdam: North-Holland, 1984, pp. 1023-1100.

35. Kremers, Jeroen, Neil Ericsson and Juan J. Dolado, "The Power of Cointegration Tests." Oxford Bulletin of Economics and Statistics, August 1992, 325-48.

36. Krugman, Paul, "Increasing Returns and Economic Geography." Journal of Political Economy, June 1991, 483-99.

37. Lizondo, J. Saul. "Foreign Direct Investment," in Determinants and Systemic Consequences of International Capital Flows, Occasional Paper No. 77, Washington, D.C.: International Monetary Fund, 1991, pp. 68-82.

38. Lunn, John, "Determinants of US Direct Investment in the EEC: Further Evidence." European Economic Review, January 1980, 93-101.

39. -----, "Determinants of US Direct Investment in the EEC: Revisited Again." European Economic Review, May 1983, 391-93.

40. MacKinnon, James G. "Critical Values for Cointegration Tests," in Long-Run Economic Relationships: Readings in Cointegration, edited by Robert F. Engle and Clive W. J. Granger. Oxford: Oxford University Press, 1991, pp. 267-76.

41. Perron, Pierre, "The Great Crash, the Oil Price Shock, and the Unit Root Hypothesis." Econometrica, November 1989, 1361-401.

42. Phillips, Peter C. B. and Bruce E. Hansen, "Statistical Inference in Instrumental Variables Regression with I(1) Processes." Review of Economic Studies, January 1990, 99-125.

43. ----- and Sam Ouliaris, "Asymptotic Properties of Residual Based Tests for Cointegration." Econometrica, January 1990, 165-93.

44. Porter, Michael E. "Competition in Global Industries: A Conceptual Framework," in Competition in Global Industries, edited by Michael E. Porter. Boston: Harvard Business School Press, 1986, pp. 15-60.

45. Rappoport, Peter and Lucrezia Reichlin, "Segmented Trends and Non-Stationary Time Series." Economic Journal, Supplement 1989, 168-77.

46. Ray, Edward John, "Foreign Direct Investment in Manufacturing." Journal of Political Economy, April 1977, 283-97.

47. Root, Franklin R. and Ahmed A. Ahmed, "Empirical Determinants of Manufacturing Direct Foreign Investment in Developing Countries." Economic Development and Cultural Change, July 1979, 751-67.

48. Scaperlanda, Anthony and Robert Balough, "Determinants of US Direct Investment in the EEC: Revisited." European Economic Review, May 1983, 381-90.

49. ----- and Laurence J. Mauer, "The Determinants of US Direct Investment in the EEC." American Economic Review, September 1969, 558-68.

50. ----- and -----, "The Determinants of US Direct Investment in the EEC: Reply." American Economic Review, September 1972, 700-704.

51. Schmitz, Andrew and Jurg Bieri, "EEC Tariffs and US Direct Investment." European Economic Review, November 1972, 259-70.

52. Stevens, Guy and Robert Lipsey, "Interactions between Domestic and Foreign Investment." Journal of International Money and Finance, February 1992, 40-62.

53. Stock, James, "Asymptotic Properties of Least Squares Estimators of Cointegrating Vectors." Econometrica, September 1987, 381-86.