This Selected Issues paper of Portugal highlights the discussions on the requirement of policies to overcome structural and cyclical impediments to growth, and secure fiscal consolidation. It analyzes the strength of the company balance sheets in supporting the rebound from recession, and also the links between corporate balance sheet strength and investment. It reviews the challenges in the Portuguese economy, the impact of European Union enlargement on Portuguese trade, the pension prospects, and the implications of various policy reform scenarios.

Abstract

This Selected Issues paper of Portugal highlights the discussions on the requirement of policies to overcome structural and cyclical impediments to growth, and secure fiscal consolidation. It analyzes the strength of the company balance sheets in supporting the rebound from recession, and also the links between corporate balance sheet strength and investment. It reviews the challenges in the Portuguese economy, the impact of European Union enlargement on Portuguese trade, the pension prospects, and the implications of various policy reform scenarios.

II. Corporate Balance Sheets and Investment: Empirical Estimates for Portugal and Other EU Countries1

A. Introduction

5. Portugal ranks high among European countries in terms of corporate indebtedness. The debt level of nonfinancial corporations was around 105 percent of GDP in 2002, which compares with an average of 80 percent for a group of six other European countries, including Belgium, Finland, France, Germany, Italy, and Spain (Figure 1). Interest rate convergence after EU accession as well as expectations of large benefits from monetary union have played a key role in fueling domestic demand growth and thus indebtedness. The rise in corporate indebtedness has exceeded substantially the rise in corporate equity, and corporate leverage ratios were also affected negatively by the collapse in equity prices after 2000.

Figure 1:
Figure 1:

Corporate Indebtedness, 1995-2002 1/

(In percent of GDP)

Citation: IMF Staff Country Reports 2004, 081; 10.5089/9781451832181.002.A002

Sources: European Central Bank (ECB), flow-of-funds data; and Fund staff estimates.1/ Total deft includes loans and securities other than shares.

6. This chapter explores to what extent the rise in corporate indebtedness, or other balance sheet indicators, may undermine investment demand in Portugal—a critical issue as the economy emerges from the recession. Real investment is estimated to have declined by almost 15 percent during 2002–03, and this was an important factor behind the recession. While weak prospects for aggregate demand, including the collapse of a housing boom, clearly played a role, there are also concerns that high corporate indebtedness has undermined investment activity—or could do so in the future. Similar concerns about the implications of a possibly overleveraged corporate sector have also been raised for several other euro-area countries.

7. Dynamic panel data techniques are used to investigate the empirical relationship between corporate indebtedness—and other corporate financial health indicators—and investment. The cross-country analysis covers a group of seven European countries (with the country selection determined by data availability). The focus is not only on the general relationship between corporate balance sheet indicators and investment, but also (i) on the presence of asymmetric balance-sheet effects over the cycle (the so-called financial accelerator), and (ii) on the sensitivity of investment above certain threshold values.

8. The econometric results suggest:

  • There is, in general, only weak evidence of links between corporate investment and leverage or other corporate balance sheet indicators.

  • However, investment is significantly affected by these indicators above certain threshold levels of leverage and over the downturn phase of the real business cycle.

  • Relatively weak balance sheets affect corporate investment negatively following real output downturns, but there is little evidence of an impact following an equity market bust.

Based on the empirical results, back of the envelope calculations suggest that balance sheet effects contributed significantly to the investment decline during the present economic downturn in Portugal. However, other factors (as already noted) clearly also played a role, and the observed decline in investment was considerably larger than can be attributed to the estimated balance sheet effects.

9. The remainder of the chapter is organized as follows. Section B reviews the literature on leverage effects and the financial accelerator. Section C describes the data used in the analysis, and the estimation methodology and hypothesis tests are presented in Section D. Following the identification of downturns in Section E, panel estimates of balance-sheet effects are discussed in Section F. Implications for Portugal are presented in Section G and the final section offers concluding remarks.

B. Literature

10. There is no strong theoretical reason in the traditional finance literature to expect the existence of investment-leverage sensitivities, nor to expect that these sensitivities increase monotonically with the level of financing constraints, more specifically with the level of corporate indebtedness, the cash position of the firm, or its liquidity constraints. Modigliani and Miller’s (1958) theory of capital markets predicts that a firm’s financial position or the composition of its liabilities has no effect on its investment decisions. In the Q model of investment under perfect capital markets, Tobin’s Q (i.e., the cost of capital) is all the information needed to guide the firm’s investment decisions. In particular, the magnitude of internal funds and the external financing structure of the firm do not matter and have no additional explanatory power.

11. Some economists and market analysts have challenged this neoclassical prediction and have presented empirical evidence of the sensitivity of investment to different measures of leverage, cash flow, and the “strength” of the balance sheet. These findings are motivated by models of capital market imperfections, including asymmetric information and agency problems (see Bernanke, Gertler, and Gilchrist, 1996). Capital market imperfections might limit the availability of external finance to firms with weak balance sheets. For example, if lenders cannot observe or control the risk involved in the investment project they finance, the stake of a borrower in the project—measured by the share of investment financed by internal funds—might be a signal of the unobserved risk of lending. Consequently, the cost of external funds, and thus its investment spending, might vary with the borrower’s financial health.

12. Fazzari, Hubbard, and Petersen (1988) provided empirical evidence of linkages between cash flow-capital ratios and investment, and they interpreted this as consistent with the presence of important financing constraints on the investment of some firms. To the extent that there is a large differential between internal and external finance and external funds might be scarce at times, investment spending might be sensitive to the availability of internal funds. A large list of empirical studies on the links between investment and liquidity and leverage measures followed Fazzari, Hubbard, and Peterson (1988). For example, Jaeger’s (2003) analysis of time series data for Germany and the United States suggested that leverage effects on corporate investment can be substantial and persistent, particularly if leverage exceeds threshold values.

13. A related literature based on the financial accelerator theory (Bernanke, Gertler, and Gilchrist, 1999) pointed out that adverse shocks to the economy might be amplified by worsening credit-market conditions. An implication of this theory is that borrowers who face significant agency costs of borrowing—for example, small firms and firms with weak balance sheets—are likely to experience larger reductions in their economic activity following a negative shock relative to other borrowers. Vermeulen (2002) tests this theory using data for manufacturing industries in four European countries, and he finds evidence of a financial accelerator with different strength across size classes and asymmetric effects over the cycle in Europe.

14. This chapter builds on Vermeulen’s (2002) work and provides further empirical evidence about investment-balance sheet sensitivities and asymmetric leverage effects over a larger group of countries and activity sectors—including the primary sector, manufacturing, and services—in Europe. Asymmetric effects are examined over real business cycles as well as over stock market price cycles. In addition, leverage threshold effects are analyzed and the cost of high indebtedness in terms of investment is estimated for Portugal.

C. Data

15. The data used in this chapter is drawn from the BACH2 (Bank for the Accounts of Companies Harmonized) database. This database, managed by the European Commission, contains harmonized annual accounts statistics of nonfinancial enterprises for 11 European countries, Japan, and the United States from the early 1980s to 2001. It is constructed through the aggregation of a large number of individual firms’ balance sheet and profit and loss accounts. BACH covers 23 activity sectors, 3 size classes, and 94 harmonized accounting items over different sample periods. These economic and financial data are collected at the national level through the completion of an annual questionnaire submitted by the companies. In the case of Portugal, the questionnaire is sent to 23,000 firms with a compliance rate of close to 70 percent.3

16. This chapter focuses on seven EU countries (Belgium, Finland, France, Germany, Italy, Portugal, and Spain) and 15 activity sectors (covering the primary sector, manufacturing, construction, trade, and service industries). The data availability varies across countries. For instance, for Italy data exist from 1982 to 2001, while, in the case of Portugal, the series used in this chapter start in 1995 (see the Appendix for further detail). The unbalanced panel employed in this chapter contains 102 country-sector pairs4 observed over a different set of years, amounting to 1,392 observations. This means that the sample contains data for an average of 13 years for each county-industry class.

17. As with most empirical work at the level of aggregation used in this chapter, the results reported below need to be interpreted with caution in light of issues related to data quality and consistency. Concerns relate in part to the comparability of the data in the BACH dataset (for example, across different countries) and to potential biases related to the response rates of firms included in the dataset. In addition, data issues in some instances may also arise for some of the macroeconomic data (including those for industrial production) used in the analysis below.

D. Estimation Methodology and Hypothesis Tests

18. The chapter performs three hypothesis tests. First, the perfect capital market prediction is tested, that is, the hypothesis that investment is not affected by the composition of a firm’s balance sheet (for example, its debt-equity ratio). The following indicators, which provide information about a firm’s financial position, are analyzed: debt-to-asset ratio (DA), debt-to-equity ratio (DE), debt-to-internal funds ratio (DIF), short-term debt-to-assets ratio (SDCA), coverage ratio, that is, cash flow to interest payments ratio (COV), short-term debt as a fraction of total debt (FS), and cash flow-to-asset ratio (CA). These are typical measures of leverage, liquidity, and corporate financial health.

19. The following investment equation is estimated:

IKi,t-αi+δt+β1IKi,t-1+β2IKi,t-2+β3SKi,t-1+β4Bi,t-1+ϵi,t(1)

where: IKit is the investment–to-capital ratio of the country-industry pair i at time t, SK is the sales-to-capital ratio, B is a corporate balance sheet measure (i.e., DA, DE, DIF, SDCA, COV, FS, or CA) δt is a time fixed effect, and αt is an unobserved individual fixed effect. The individual fixed effect is intended to capture all time-invariant factors, including unobservable characteristics, associated with a country-sector pair that have affected investment to capital ratios historically. The time effects capture common time developments across all country-industry pairs in the panel. Because of lack of data and other potential problems associated with measuring Tobin’s Q (i.e., the cost of capital), a sales variable is introduced in the specification to control for underlying investment opportunities; and robustness checks reported later also include an interest-rate proxy for the cost of capital.

20. Given the presence of lagged dependent variable terms on the right-hand side of equation (1), the standard fixed effect estimator is inconsistent (Nickell, 1981), so the Arellano and Bond (1991) estimator is employed. First differencing equation (1) removes the αt and produces an equation that is estimable by instrumental variables. Arellano and Bond derived a generalized method of moments estimator (GMM) using lagged levels of the dependent variable and the predetermined variables. This methodology assumes that there is no second-order autocorrelation in the first-differenced idiosyncratic errors. In order to minimize this potential problem, a second lag of the investment-capital ratio is included in the regression. Regressions with only one lagged dependent variable were also estimated and yielded similar results.

21. The key parameter of interest in equation (1) is β4. The null hypothesis of the first test is that β4 = 0. Failing to reject this hypothesis would suggest that investment decisions are not affected by the quality or strength of balance sheet indicators, as predicted by the perfect capital market theory.

22. The second exercise seeks to test the presence of asymmetric balance sheet effects over the cycle. Using the same dynamic panel data methodology, the following specification is estimated:

IKi,t=αi+δt+β1IKi,t-1+β2IKi,t-2+β3SKi,t-1+(β4+β5Rjt)Bi,t-1+ϵi,t(2)

where Rjt is a dummy variable defined to be one if country j is in a downturn at time t and zero otherwise. The chapter examines two types of downturns, namely real sector downturns—characterized in terms of industrial production or the output gap—and stock market busts (as described in more detail in Section F). The objective of this analysis is twofold. First, the exercise attempts to determine the existence of a financial accelerator in Europe and the presence of binding external finance constraints during downturn phases of the cycle. The sign and statistical significance of β5 will shed some light on this question. Second, if asymmetries over the cycle are relevant, the analysis attempts to identify what type of cycles—real sector, equity price, or both—contribute to amplify the effects of weak balance sheets on corporate investment.

23. A third set of tests investigates threshold effects. Some market analysts as well as Jaeger (2003) have argued that an increase in leverage will have adverse effects on corporate investment only if debt or other balance sheet indicators exceed some threshold value. Similarly, firms with relatively low value of total assets are thought to be more vulnerable to high indebtedness. In order to examine these effects, equations (1) and (2) are estimated for two different samples. This includes a sample split for country-sector classes with high debt-asset ratios versus country-sector pairs with low debt-asset ratios. Likewise, the sample is divided according to the following financial constraints: high versus low debt-equity ratios and high versus low value of assets. Threshold levels are characterized by summary statistics of these indicators (average, median, and various percentiles).

E. Identifying Downturns

24. This section presents alternative definitions of downturns and how the dummy variable Rjt in equation (2) is constructed in each case. In the first definition, a downturn is identified in country j in year t if the country’s industrial production growth at time t is negative. Table 1 contains data on industrial production for the seven countries in the study. The shaded areas on the table identify the downturn years. For instance, Portugal experienced industrial production slowdowns in 1985, from 1992 to 1994 and in 2000. Likewise, in France a downturn took place between 1991 and 1993.

Table 1.

Industrial Production 1/

article image
Source: IMF, World Economic Outlook database.Note: Shaded area indicate downturns.

Excludes construction.

25. The chapter also examines downturns characterized in terms of the output gap. Country j in year t is in a downturn if its output gap is widening.5 Table 2 presents estimates of the output gap in percent of GDP, including indicators of widening output gap downturns. This represents a broader definition of downturns, with Portugal satisfying this condition for 7 years in the period 1982-2001 (compared to 5 years implied by the definition in Table 1).

Table 2.

Output Gap

article image
Source: IMF, World Economic Outlook database.Note: Numbers I identify downturns.

26. In addition to identifying real sector downturns, this chapter also identifies equity price “busts” and tests the financial accelerator theory for periods where stock market valuations are falling. Recent papers (Jaeger, 2003; and IMF, 2003) suggest that leverage effects might be asymmetric over equity price cycles. Using data on European stock market indices—BEL 20 (Brussels), HEX 20 (Helsinki), CAC 40 (Paris), DAX (Frankfurt), MIB 30 (Milano), BVL 30 (Lisbon), and IBEX 35 (Madrid)—a dummy variable is defined to take on the value of “1” during stock price declines (periods of negative growth of the stock market index) and “0” otherwise. In those cases where index series do not extend back to the starting sample year, additional information is taken from IMF (2003), which describes a procedure to identify asset price booms and busts.

F. Empirical Results

Summary statistics and empirical trends

27. Table 4 presents summary statistics of the variables used in the econometric analysis over the seven countries and 15 activities sectors in the study. As already mentioned, time periods vary from country to country. Median values of the balance-sheet indicators by sectors of activity (Figure 2) indicate that the buildings and civil engineering sector and the trade sector rank at the top of the different leverage measures (debt-to-asset, debt-to-equity, and debt-to-internal funds). Firms in the transport and communication sector have the lowest liquidity position, in terms of short-term debt-to-asset ratio and short-term debt as a fraction of total debt, followed by the energy and water sector. With regards the coverage ratio, the best-positioned industries are chemicals and extraction of non-metallic minerals.

Figure 2.
Figure 2.

Portugal: Median Values by Sectors 1/

Citation: IMF Staff Country Reports 2004, 081; 10.5089/9781451832181.002.A002

1/ Sector 1 is energy and water; sectors 2-11, manufacturing industries; sector 12 is buildings and civil engineering; sector 13 is trade; sector 14 is transport communication; and sector is 15 other service.
Table 3.

European Stock Market Indices

article image
Source: Bloomberg.
Table 4:

Summary Statistics of the Variables Used

article image
Source: BACH database.

28. Figure 3 displays investment capital ratios for the seven European countries in the study. These ratios are median values over the fifteen industries under consideration. Some downturn episodes are reflected extremely well in these investment trends. Most countries experienced downturns in the early 1990s and resumed growth after 1993, and the latter year typically coincides with a trough in the investment-to-capital ratio.

Figure 3.
Figure 3.

Investment-to-Capital Ratios, 1982-2000

(In percent of GDP)

Citation: IMF Staff Country Reports 2004, 081; 10.5089/9781451832181.002.A002

Source: BACH database.

Investment-balance sheet sensitivities

29. This section discusses the estimation results of the dynamic investment model described in equation (1). Table 5 reports the robust one-step Arellano and Bond estimators for different balance sheet indicators.6 The estimated coefficients suggest that there is only weak evidence that balance sheet positions affect investment decisions.

Table 5.

Balance Sheet Indicators and the Investment-Capital Ratio

(Dependent variable is IKt)

article image
Source: BACH database.Notes: Robust standard errors in parentheses.* significant at 10%; ** significant at 5%; *** significant at 1%.Ind-Ctry is the number of industry-country pairs (i.e., cross-section dimension of the panel). All regressions include time dummies. AR1 and AR2 are first order and second order serial correlation tests. Lagged values of the variables used as intruments. All regressions pass the Sargan test of overidentifying restrictions.

30. Higher leverage ratios—as measured by the debt-to-assets and the debt-to-equity ratios—tend to lead to lower investment rates, as suggested by some agency models, but the coefficients are not statistically significant (see columns (1) and (3)). Only the degree to which a firm is able to finance its investment using internal funds has a statistically significant, though small, impact on investment. A one standard deviation increase in the debt-to-internal funds ratio leads to a drop in the investment-capital ratio close to 0.2 percentage points.7 This result suggests that the debt-to-internal funds ratio has a role in signaling the unobserved risk of investment projects to lenders. On the other hand, the impact of the firm’s liquidity position is mixed. More specifically, short-term debt as a fraction of total debt is not statistically significant (column (5)) while a higher short-term debt-to-asset ratio tends to reduce investment, but only at the 10 percent significance level.

31. These estimation results provide some evidence of investment-cash flow sensitivities. To the extent that firms are constrained in their ability to raise funds externally or if internal finance has cost advantages over external finance, investment decisions may be sensitive to the availability of internal funds. In addition, cash flows may contain information about investment opportunities. In this sense, higher availability of internal funds may affect current investment. The regression results suggest that a one standard deviation increase in the cash flow-to-asset ratio leads to a rise in the investment-to-capital ratio by close to 1 percentage point.8

32. The tests of autocorrelations in the first-differenced residuals (see last rows in Table 5) indicate the presence of first-order autocorrelation but no second-order autocorrelation, implying that the coefficient estimates are consistent. Moreover, the Sargan test (not reported) accepts the null hypothesis that the over-identifying restrictions are valid.9

Asymmetric balance sheet effects

33. The results concerning asymmetric balance sheet effects—the estimation of equation (2) above—depend crucially on the variable used to identify the periods of economic or financial market weakness. First, the results suggest asymmetric balance sheet effects over the real business cycle. That is, firms with weak balance sheets are more likely to experience larger contractions in investment following a real downturn than financially healthy firms. Second, there is no evidence of a financial accelerator operating over the equity price cycles. After a stock market bust, financially more constrained firms are not more likely to experience investment reductions.

34. Tables 6 and 7 indicate that, in general, leverage and liquidity indicators have no impact on investment outside real sector downturns but they are important in determining investment spending during real sector slowdowns—defined as periods of negative industrial production growth or widening output gaps. Specifically, a one standard deviation increase of the debt-to-asset ratio (DA) leads to a drop in the investment-to-capital ratio of 0.3 percentage point following a widening-output-gap type of downturn.10 Likewise, a liquidity shortfall, defined as a one standard deviation increase of short-term debt-to-total debt, leads to an investment contraction of 0.4 percentage point during economic downturns and has no effect outside the downturn periods. The exception is the debt-to-equity indicator, for which the coefficient estimates are negative but not statistically significant, even during periods of a real sector downturn.

Table 6.

Asymmetric Effects over Real Sector Downturns I 1/

(Dependent variable is IKf)

article image
Source: BACH database.Notes: Robust standard errors in parentheses.* significant at 10%; **significant at 5%;*** significant at 1%.Ind-Ctry is the number of industry-country pairs (i.e., cross-section dimension of the panel). All regressions include time dummies. AR1 and AR2 are first order and second order serial correlation tests. Lagged values of the variables used as intruments. All regressions pass the Sargan test of over identifying restrictions.R=1 if industrial production growth is negative.

Downturns are characterized by negative industrial production growth.

Table 7.

Asymmetric Effects over Real Sector Downturns II 1/

(Dependent variable is IKt)

article image
Source: BACH database.Notes: Robust standard errors in parentheses.* significant at 10%; ** significant at 5%; *** significant at 1%.Ind-Ctry is the number of industry-country pairs (i.e., cross-section dimension of the panel). All regressions include time dummies. AR1 and AR2 are first order and second order serial correlation tests. Lagged values of the variables used as intruments. All regressions pass the Sargan test of overidentifying restrictions.R=l if output gap is widening.

Downturns are characterized by widening output gap.

35. Overall, the results are consistent with the financial accelerator theory: negative shocks reduce the borrower’s net worth and increase the agency costs of lending, amplifying the investment effects of the initial shock. Furthermore, financing constraints might become binding in periods of high uncertainty or less availability of funds, such as recessions. There is indeed evidence that banks change their lending standards—from more laxity to tightness—systematically over the real business cycle (Asea and Blomberg, 1997).

36. While some balance sheet effects on investment tend to be more pronounced during real sector downturns, this is not the case for the impact of cash flow indicators on investment. This is evidenced by the negative coefficient corresponding to the interaction term in the coverage ratio and cash-to-assets regressions (columns (6) and (7) in Tables 6 and7). Theoretically, there are at least two effects working in opposite directions. On the one hand, higher relative costs of external finance during recession tend to favor the use of internal funds and hence increase the investment cash flow sensitivity. On the other hand, lower marginal profitability of capital leads firms to postpone investment projects and wait for higher returns, independently of the size of internal cash flows. This second effect dominates empirically, making investment less sensitive to cash flow during recession periods. In particular, while a one standard deviation increase in the cash flow-to-asset ratio leads to an increase of the investment-capital ratio by 0.9 percentage point outside downturns, the effect is 0.4 percentage point lower during slowdowns (although this still leaves, on net, an overall positive effect of cash flow on investment even during downturns).

37. This finding is similar to results in Kaplan and Zingales (1997). Their results suggest that the sensitivity of investment to cash flow is not necessarily increasing in the degree of financing constraints. Using a sample of manufacturing firms in the United States, they find that those firms classified as less financially constrained exhibit significantly greater investment cash flow sensitivity than those firms classified as more financially constrained. Vermeulen (2001) also finds this pattern and argues that it is consistent with the informational interpretation of cash flow. Cash flow is likely to provide less information about future investment opportunities during downturns.

38. Contrary to the findings for a real sector downturn reported in Tables 6 and 7, the results reported in Table 8 suggest that firms’ investment is not sensitive to more leveraged or less liquid balance sheets during stock market busts. There is, therefore, no evidence of asymmetric balance sheet effects over equity price cycles. One possible interpretation relates to banks’ lending cycles. While bank lending tightening takes place at the trough of output recessions, there is no clear correlation between financial market cycles and bank lending cycles.11 Furthermore, there is a large body of evidence that suggests that cyclical changes in firm financing are dominated by changes in bank lending, especially at the peak and during the downward phase of the cycle.12

Table 8.

Asymmetric Effects over Equity Price Cycles

(Dependent variable is IK/)

article image
Source: BACH database.Notes: Robust standard errors in parentheses.* significant at 10%; ** significant at 5%;*** significant at 1%.Ind-Ctry is the number of industry-country pairs (i.e., cross-section dimension of the panel). All regressions include time dummies. AR1 and AR2 are first order and second order serial correlation tests. Lagged values of the variables used as intruments. All regressions pass the Sargan test of overidentifying restrictions.R=l if equity market bust (negative growth of the stock market index).

39. With banks changing their lending standards systematically over the real business cycle and with firms depending on banks for external financing, it is not surprising that leverage positions affect investment negatively during output slowdowns. However, since stock market busts do not correlate with credit tightening by banks, small changes in balance sheet indicators do not affect these firms’ financial constraints substantially, and therefore their investment decisions are not substantially changed. In other words, even those firms with weaker balance sheet continue to have access to bank lending during equity price troughs and do not seem to change their investment behavior.

Threshold effects

40. The presence of threshold effects is examined by estimating equations (1) and (2) in two different samples: high- and low-leverage country-sector classes. The estimation results in Table 9 indicate that additional increases in indebtedness have a negative impact on investment spending when the firm’s debt-to-asset and debt-to-equity ratios are above median levels. More specifically, the debt-to-asset ratio does not affect the investment decisions of those firms in a low leverage regime. However, for highly indebted firms, an additional 10 percent increase in this ratio leads to a drop in investment by 2.5 percent. Likewise, a 10 percent increase in the debt-to-equity ratio reduces investment by 2 percent in the high-leverage firms and has no impact on the low-leverage firms. Furthermore, asymmetric effects over the real cycle are amplified in the high-leverage sample. Finally, (and not reported in Table 9) firms with relatively low value of total assets are also more vulnerable to high indebtedness.

Table 9.

Leverage Threshold Effects

(Dependent variable is IKt)

article image
Source: BACH database.Notes: Robust standard errors in parentheses.* significant at 10%; ** significant at 5%; *** significant at 1%.Ind-Ctry is the number of industry-country pairs (i.e., cross-section dimension of the panel). All regressions include time dummies. AR1 and AR2 are first order and second order serial correlation tests. Lagged values of the variables used as intruments. All regressions pass the Sargan test of overidentifying restrictions.R=l if industrial production growth is negative.(1) and (3) DAt-1 larger than median value(2) and (4) DAt-1 smaller than median value.(5) and (7) DEt-1 larger than 60th percentile.(6) and (8) DEt-1 smaller than 60th percentile.

Robustness check: controlling for the cost of capital

41. This section examines whether the implicit interest rate assumption in equations (1) and (2)—namely, that interest rates have evolved similarly across countries over time—has implications for the results reported above. The inclusion of time and country-industry dummies in the previous estimates controls for the omitted cost of capital variable when the time-series variability of the cost of capital is the same across countries. However, the level and volatility of interest rates was to some extent different across countries, in particular prior to monetary union period (Figure 4). In order to account for country-specific interest rates, all regressions were reestimated including the lending rate as an explanatory variable.

Figure 4.
Figure 4.

Lending Rates, 1982-2002

(In percent)

Citation: IMF Staff Country Reports 2004, 081; 10.5089/9781451832181.002.A002

Source: IMF, International Financial Statistics.

42. The sign of the potential omitted variable bias on the financial sector variables, such as leverage, is likely to be positive.13 This implies, for example, that the true effect of leverage on investment would become more negative, once the interest rate variables are included in the regression.

43. The estimation results including the cost of capital variable suggest, however, that the earlier estimation results are all robust to the inclusion of a cost of capital variable. Although the lending rate has a negative and statistically significant effect on investment—a 1 percent increase in lending rates leads to a drop of about 0.2 percent in the investment-to-capital ratio—the coefficients of all balance sheet indicators remain basically unchanged. Tables 10 through 12 replicate the results in Table 5-7 and reconfirm the previous results. First, there is at best weak evidence of investment-leverage sensitivities in general. However, weak balance sheets affect investment negatively following real sector downturns and above certain threshold levels of leverage. In addition, there are not asymmetric effects working over equity price cycles. Overall, these results suggest that while country-specific interest rate effects are important to characterize investment, the omission of this variable is not central to the main questions addressed in this chapter—the effects of different corporate balance sheet indicators and investment.

Table 10.

Investment-Balance Sheet Sensitivities Controlling for the Cost of Capital

(Dependent variable is IKt)

article image
Source: BACH database.Notes: Robust standard errors in parentheses.* significant at 10%; ** significant at 5%; *** significant at 1%.Ind-Ctry is the number of industry-country pairs (i.e., cross-section dimension of the panel). All regressions include time dummies. AR1 and AR2 are first order and second order serial correlation tests. Lagged values of the variables used as instruments. All regressions pass the Sargan test of overidentifying restrictions.
Table 11.

Asymmetric Effects over Real Sector Downturns Controlling for the Cost of Capital

(Dependent variable is IKt)

article image
Source: BACH database.Notes: Robust standard errors in parentheses.* significant at 10%; ** significant at 5%; *** significant at 1%.Ind-Ctry is the number of industry-country pairs (i.e., cross-section dimension of the panel). All regressions include time dummies. AR1 and AR2 are first order and second order serial correlation tests. Lagged values of the variables used as instruments. All regressions pass the Sargan test of overidentifying restrictions.R=l if industrial production growth is negative.