{Journals}eufm/7_4/e125/makeup/e125.

3d

European Financial Management, Vol. 7, No. 4, 2001, 481± 496

Why Do Firms Raise Foreign Currency

Denominated Debt? Evidence from
Finland
Matti Keloharju*
Helsinki School of Economics and Business Administration, Runeberginkatu 22 ± 24, FIN-00100
Helsinki, Finland
e-mail: [email protected]

Mervi Niskanen
HaÈme Polytechnic University

Abstract
This study examines the determinants of the decision to raise currency debt. The
results suggest that hedging figures importantly in the currency-of-denomination
decision: firms in which exports constitute a significant fraction of net sales are
more likely to raise currency debt. However, firms also tend to borrow in periods
when the nominal interest rate for the loan currency, relative to other currencies, is
lower than usual. This is consistent with the currency debt issue decision being
affected by speculative motives. Large firms, with a wider access to the international
capital markets, are more likely to borrow in foreign currencies than small firms.

Keywords: currency of denomination; hedging; speculation.

JEL classification: F 23, G32.

1. Introduction

The currency of denomination is one of the most important characteristics in any

corporate debt issue (see, e.g., Levi, 1990; Giddy, 1994). There are three reasons why a
firm might want to raise foreign currency debt.
First, it provides hedging of foreign exchange exposures. The shareholders of a firm
can benefit from hedging at the corporate level because it decreases taxes and the
costs of financial distress (Smith and Stulz, 1985), and helps ensure that the firm has

* We thank Tom Berglund, John Doukas (the Editor), Markku Koskela, Reija Lilja, Pedro
Santa-Clara, Kari Toiviainen, the participants of the joint seminar of Helsinki School of
Economics and Swedish School of Economics, and particularly the referee for constructive
comments. Financial support from the Helsinki Stock Exchange Foundation and the Helsinki
School of Economics and Business Administration Foundation is gratefully acknowledged.

# Blackwell Publishers Ltd 2001, 108 Cowley Road, Oxford OX4 1JF, UK and 350 Main Street, Malden, MA 02148, USA.
{Journals}eufm/7_4/e125/makeup/e125.3d

482 Matti Keloharju and Mervi Niskanen

sufficient internal funds available to take advantage of attractive investment

opportunities (Froot et al., 1993). Occasionally hedging may be of interest for
corporate managers even if it does not benefit shareholders. Hedging increases
managers' expected utility from compensation (Stulz, 1984) and reduces the impact
that foreign exchange gains and losses have on reported earnings (Smith and Stulz,
1985; DeMarzo and Duffie, 1996).
Second, borrowing in foreign currencies may cost less than borrowing in the domestic
currency. Issuing in the Euromarkets may be more economical than domestic
borrowing as it helps to circumvent withholding taxes and capital controls imposed
by many governments. Shapiro (1984) and Rhee et al. (1985) analyse the role of taxes
and flotation costs in the determination of loan currency and show that if the expected
borrowing costs in two currencies are equal before tax, the process of minimising after-
tax borrowing costs generally involves borrowing the weaker currency. Moreover,
Shapiro shows that in some countries, such as Finland, tax laws encourage companies
to borrow the foreign currency, regardless of whether it is weaker or stronger.
Third, speculative reasons may make foreign currency debt an attractive alternative.
A financial manager may choose to deviate from a hedging oriented strategy if she
believes that, adjusting for risk, the differential in interest rates between two currencies
does not accurately reflect the expected exchange rate change. Such occurrences are
not necessarily rare: Hakkarainen et al. (1997) report that 72% of the CFOs of 68
large Finnish corporations believe that they can predict exchange rates at least to some
extent in the short run and 53% in the long run. While the managers' beliefs appear to
be consistent with the well-known findings in the psychology of judgment that people
tend to be overconfident (see, e.g., DeBondt and Thaler, 1995), they may also be
motivated by the fact that the International Fisher Effect does not necessarily hold. In
fact, VireÂn (1988) documents that in 1980 ±87 the uncovered interest rate parity
relationship was very weak or nonexistent in Finland Ð much weaker than for
countries with more developed capital markets such as Switzerland and Sweden. Also
many other researchers, such as Fama (1984), report evidence that is inconsistent with
the International Fisher Effect.
Recent studies by Allayannis and Ofek (forthcoming) and Kedia and Mozumdar
(1999) examine the use of foreign currency debt in large US corporations, while
Doukas et al. (2000) study the determinants of foreign exchange risk exposure of
Japanese companies. The evidence in these papers is consistent with the notion that
hedging demand is an important motivation for issuing currency debt, although the
motivation may differ across currencies.1 To our knowledge, there has not been any
rigorous studies on the use of currency debt in Europe, despite of the importance of
the research question and the fact that the motivations for particularly US firms to
issue currency debt may differ from those of firms in other countries due to differences
in the openness of the economy. The lack of empirical studies on the currency-of-
denomination decision is probably largely due to the fact that banks are the most

1
Geczy et al. (1997) examine the determinants of currency derivative usage. Also their evidence
is consistent with the notion that currency derivatives are primarily used for hedging purposes.
Moreover, they find that firms that use currency swaps or combinations of swaps have relatively
higher levels of currency debt than firms with no currency derivatives. This suggests that
currency derivatives and foreign debt complement each other in reducing a firm's currency risk
exposures.

# Blackwell Publishers Ltd, 2001

{Journals}eufm/7_4/e125/makeup/e125.3d

Foreign Currency Denominated Debt 483

important source of debt financing in many countries (Mayer, 1990), and yet no
sources of company information on new bank loan agreements are readily available
(James, 1987).
This paper examines the role of foreign currency debt by taking advantage of a
unique and comprehensive sample of private and public debt raised by Finnish
corporations. The sample includes virtually all of the non-subsidized long-term
borrowing of 44 listed Finnish corporations between 1985 and 1991, representing
19% of the total long-term corporate borrowing and 30% of foreign currency
denominated long-term corporate borrowing in Finland during the time period.
Finnish data are particularly well suited for examining the role of foreign currency
debt since Finland is a more open economy than most large industrial countries: in
1994 exports constituted 36% of GNP, whereas the corresponding figure was 11% in
the USA and 23% in Germany. Therefore, many Finnish companies Ð including
sample companies which on average generated 35% of their net sales from exports
during the sample period Ð are likely to benefit from raising foreign currency debt
for hedging purposes.
The remainder of the paper is organised as follows. The next section provides a
short description of Finnish financial institutions and the role of currency
denominated loans in Finnish corporate finance. Section three details the data.
The sample consists of 337 new loans with an aggregate nominal amount of FIM30
billion (FIM7  US$1) of which FIM25 billion is foreign currency debt. Section four
documents the empirical findings. The results suggest that hedging induced demand is
an important determinant of the currency-of-denomination decision although we also
find evidence consistent with speculative behaviour affecting the choice of loan
currency. Section five concludes with a summary of the findings.

2. Finnish financial institutions and the role of foreign currency denominated loans in
Finnish corporate finance
Banks play a major role in Finnish corporate finance. As in Japan and Germany, in
particular, a large part of the funds for investments are provided in the form of bank
loans. Mayer (1990) reports sources of average gross financing of non-financial
enterprises in 1970± 85 and finds that in Finland 45% of external funding came from
banks as opposed to 9% from shares and 3% from bonds.
In addition to banks, insurance companies are an important source of institutional
finance for the Finnish corporate sector. For example, in 1991 they provided about
20% of the total borrowing to Finnish industrial companies. Insurance companies
differ markedly from banks in that they lend only in the domestic currency. Moreover,
because of regulation a significant fraction of their loans to the corporate sector is
essentially subsidised.
Another important competitor for banks are state-operated funds that grant
Finnish markka denominated loans with low, often subsidised interest rates. State-
subsidised loans are mostly targeted at firms that operate in rural areas or in areas that
have been hit hardest by the structural changes in the economy. In the late 1980s the
competition between the banks and other suppliers of capital was so fierce that the
average interest rate marginal for fixed rate markka denominated loans for large firms
was on average negative (Niskanen, 1996). This helps to explain why an increasing
fraction of bank loans were denominated in foreign currencies after the liberalisation
of financial markets made that possible.
# Blackwell Publishers Ltd, 2001
{Journals}eufm/7_4/e125/makeup/e125.3d

484 Matti Keloharju and Mervi Niskanen

Foreign currency denominated long-term borrowing and the use of off-balance

sheet hedging instruments were regulated in Finland in the early 1980s. Any
company wishing to borrow long-term in a foreign currency had to obtain
permission from the Bank of Finland. Permission was usually granted provided that
the company seeking it was an industrial company. The only off-balance sheet
hedging instrument available was a forward contract with a bank, allowing firms to
hedge their transaction exposures. In 1985 the Bank of Finland allowed companies
to hedge their transaction and translation exposures without permission and allowed
the use of options contracts. The next year industrial and shipping companies were
allowed to raise currency loans with maturities of at least five years without
permission. In 1987 this ruling was extended to include all companies with the
exception of financial institutions and insurance companies. In 1989 the Bank of
Finland allowed currency swaps and reduced the maturity restrictions for currency
loans to one year.

3. Data
The cornerstone of this study is a database of long-term loans (initial maturity of the
loan at least one year) which was obtained from a private survey. The survey was
targeted at all non-financial Finnish companies that had a listing for at least three
consecutive years on the Helsinki Stock Exchange (first market), in the OTC list
(second market), or the Stockbrokers' list (third market) between the years 1985 and
1991. Of these 67 companies ten refused to participate in the study and three provided
insufficient data. In addition, we exclude ten firms which did not take any long-term,
non-subsidized loans during the sample period. This leaves us with a sample of 44
companies and 337 loans.
The companies were asked to provide the following information on all their long-
term non-subsidized debt financing:
* Lender
* Loan size in Finnish markkas on the day the loan was raised
* Initial loan currency (Only two of the 185 foreign currency loans were swapped
into another currency at the time of raising the debt. Both swaps involved foreign
currencies.)
* Whether the interest rate was initially fixed or floating (ignoring interest rate
swaps), and the reference rate for floating rate loans
* Initial loan maturity (ignoring later prepayments)
* Whether collateral was pledged
* Whether the loan was guaranteed
In addition, we use the following firm specific and economy wide variables:
* Exports-to-net sales
* Log (total book assets)
* Debt-to-book assets = total debt=total book assets
* Multinational dummy. A multinational is defined as a firm with at least one foreign
subsidiary
* Return-on-book assets
P = (EBIT P taxes)=total book assets
* Concentration = niˆ 1 S 2i =( niˆ 1 Si ) 2 , where Si is the net sales of a firm's
business segment i and n is the number of business segments in that firm
# Blackwell Publishers Ltd, 2001
{Journals}eufm/7_4/e125/makeup/e125.3d

Foreign Currency Denominated Debt 485

* Book-to-market value of assets = total book assets=(total book assets book value
of equity ‡ market value of equity)
* Dividend yield = dividend per share=year-end stock price
* FIM interest rate Ð FIM basket interest rate. During the sample period the Finnish
markka was first pegged to a basket consisting of 12 currencies and, since June
1991, to the ECU. Both the markka and the basket rates are 3-month rates if the
loan is a variable rate loan and 10-year rates if the loan is a fixed rate loan. The
interest rates are calculated at the time the loan is raised.
* Dummies for the years 1985± 90.
Firm specific financial data were collected from Listed Companies in Finland,
PoÈrssitieto, and from annual reports. Interest rate data were provided by courtesy of
the Bank of Finland.

4. Results

4.1. Descriptive statistics

Table 1 shows descriptive statistics of the sample companies. They vary considerably
in size and their usage frequency of the long-term loan markets: for example, some
small companies raised only one loan during the time period whereas some large
companies raised up to 33 loans. Foreign operations play an important role in many
of the sample companies as shown by the fact that on average firms generate 35% of
their net sales from exports.
Table 2 breaks down by year the number and volume of foreign currency
loans, and the difference between the Finnish markka and the basket interest rate.
During the sample period both the short and long-term Finnish markka interest
rates were on average about 3% higher than the index. This created an incentive for
firms to borrow in foreign currencies provided that their financial managers believed
that the Finnish markka would not be devaluated. Both the number and total
volume of foreign currency loans increased towards the end of the period with the
liberation of the financial markets (this pattern is also partly due to the fact
that some old loan documents were missing), only to come down in 1991 marking

Table 1
Descriptive statistics of 44 sample companies between the years 1985 and 1991.

Variable Mean Median Minimum Maximum N

Total book assets (million FIM) 3621 1292 101 21,716 44

Debt-to-book assets 0.55 0.59 0.15 0.92 44
Exports-to-net sales 0.35 0.24 0 0.94 41
Number of loans 7.66 5 1 33 44

Notes: For a given company, the variable total book assets is calculated as the weighted average of total
book assets in the sample years. The weights are determined by the number of loans raised by the company
in each year. The variables debt-to-book assets and exports-to-net sales are calculated in the same way. The
variable exports-to-net sales is not available for three companies. All nominal values are expressed in terms
of the purchasing power of 1991 Finnish markkas (7 FIM  1 US$).

# Blackwell Publishers Ltd, 2001

{Journals}eufm/7_4/e125/makeup/e125.3d

486 Matti Keloharju and Mervi Niskanen

Table 2
The number and volume of foreign currency loans, and the average monthly difference between
the Finnish markka and the basket interest rate, by year.

Foreign currency loans Average FIM ±

FIM basket rate
Proportion of differential,
total (%) % points

No. of Loan volume No. of Loan 10-year 3-month

Year loans (million FIM) loans volume rate rate

1985 5 522 35.7 78.4 2.79 2.99

1986 14 1,490 37.8 66.2 3.57 3.42
1987 18 2,135 51.4 64.7 2.66 2.21
1988 40 5,464 62.5 91.7 2.13 2.14
1989 40 4,610 59.7 88.5 3.49 2.88
1990 40 7,053 63.5 87.5 3.24 3.16
1991 28 4,101 49.1 74.2 2.51 2.99
Totals 185 25,375 Average 51.4 78.7 2.90 2.77

Note: All nominal values are expressed in terms of the purchasing power of 1991 Finnish markkas
(7 FIM  1 US$).

the outbreak of an economic recession and a 12.3% devaluation of the Finnish

markka.
Table 3 compares Finnish markka denominated loans to currency loans. The mean
size of currency loans is more than three times larger than that of markka loans,
suggesting that smaller companies, raising smaller loans, rely on markka debt while
larger corporations, raising larger loans, opt for foreign currencies. For example, 61%
of the loans in the smallest firm size quartile are markka denominated, whereas only

Table 3
Comparison of Finnish markka and foreign currency denominated loans.

FIM denominated Foreign currency

loans denominated loans

Mean loan size (million FIM) 38.8 137.2

Median loan size (million FIM) 12.7 59.0
Mean loan maturity (years) 6.5 6.2
Median loan maturity (years) 6.0 5.0
Proportion of fixed rate loans (%) 77.0 41.6
Proportion of collateralised loans (%) 38.8 30.3
Proportion of guaranteed loans (%) 41.4 4.3
N 152 185

Note: All nominal values are expressed in terms of the purchasing power of 1991 Finnish markkas
(7 FIM  1 US$).

# Blackwell Publishers Ltd, 2001

{Journals}eufm/7_4/e125/makeup/e125.3d

Foreign Currency Denominated Debt 487

20% of the loans in the largest firm size quartile are markka loans. This finding is
probably due to the fact that foreign lenders are mostly interested in doing business
with the largest companies: their debt issue volumes are the largest, and the
information disadvantage of foreign lenders relative to domestic lenders is at its
smallest. Since a financial institution is likely to lend its local currency, borrowing
from foreign financial institutions probably provides foreign currency denominated
financing.
Fixed rate loans account for 77% of markka loans but only 42% of currency loans.
The dominant role of fixed rate markka loans can be explained by the fact that
floating markka denominated Helibor interest rates were introduced in 1987, and only
gradually gained popularity.2 The proportion of collateralised and, in particular,
guaranteed markka loans is larger than that of currency loans. This is mostly because
markka loans are the key source of financing for small companies, which are most
likely to be asked to pledge collateral or provide additional security to the lender by
purchasing a bank or insurance company guarantee.
Table 4 shows the relationship between the currency of denomination and the
source of the loan. The majority of the loans from domestic banks are foreign
currency denominated even though markka loans constitute the largest single group.
Other important lending currencies for domestic banks are the US dollar and the
German mark.
Consistent with Mayer (1990), public bonds are a relatively unimportant source of
financing: only one of the 152 Finnish markka denominated loans or debt issues is a
public bond and ten are privately placed. Similarly, only eight of the 185 foreign
currency denominated loans or debt issues are foreign bonds, and all except one are
privately placed. There are no Eurobonds in the sample.
The currency distribution of the sample loans is in line with the overall distribution
of foreign currency denominated billing of the Finnish corporate sector. In 1991 the
most important currency by far was the US dollar, representing 45% of total foreign
currency denominated billing. The German mark came next with 8% and the French
franc was the third with 5%. Since the US dollar and the German mark are also the
most frequently used foreign loan currencies Ð29% and 16% of the number of loans
and 46% and 15% of the volume of loans Ðthe results are consistent with the notion
that the sample firms use foreign currency denominated borrowing to hedge against
receivables in the same currencies.

4.2. The determinants of the decision to raise foreign currency denominated debt

Table 5 examines the determinants of the decision to raise foreign currency debt in a
probit regression framework. The dependent variable in the regression is a dummy
variable which takes the value of one if a loan is denominated in any foreign currency
and the value of zero if the loan is Finnish markka denominated.
The independent variables are as follows. First, companies are expected to use
foreign currency loans to hedge their currency receivables. Therefore, the coefficient
of the variable measuring the share of exports of net sales is expected to have a

2
In addition to 17 Helibor loans, floating rate loans include 18 loans which use the Bank of
Finland's base rate as the reference interest rate. This is an administrative interest rate mostly
used for mortgages and small firm loans.

# Blackwell Publishers Ltd, 2001

 {Journals}eufm/7_4/e125/makeup/e125.3d
# Blackwell Publishers Ltd, 2001

488
Table 4
The number of loans from each loan source and their currency of denomination.

Currency

FIM Basket ECU US$ DEM JP¥ CHF BEF GB£ SEK Other Totals

Matti Keloharju and Mervi Niskanen

Domestic sources of financing:
Banks 85 33 6 26 19 3 5 3 5 3 188
Insurance companies 56 56
Other institutions 10 3 2 5 20
Public bonds 1 1
Foreign sources of financing:
American banks 5 2 7
German banks 1 2 4 7
Japanese banks 9 2 6 1 18
Swiss banks 2 1 9 12
Luxemburgian banks 2 5 7
British banks 2 1 3
Swedish banks 3 1 4
Syndicates and other banks 3 1 1 2 7
Public bonds 1 1
Japanese insurance companies 1 5 6
Totals 152 38 9 53 30 14 14 9 7 5 6 337
{Journals}eufm/7_4/e125/makeup/e125.3d

Foreign Currency Denominated Debt 489

Table 5
The determinants of the decision to raise foreign currency denominated debt.

Dependent variable:
Expected Foreign currency denomination
Independent variables sign dummy

Constant ? 2.73 3.96

(0.0001) * * * (0.0000) * * *
Exports-to-net sales ‡ 1.67 1.35
(0.0000) * * * (0.0007) * * *
Multinational dummy ‡ 0.26 0.11
(0.2940) (0.697)
Log (Total book assets) ‡ 0.25 0.24
(0.0004) * * * (0.0040) * * *
Debt-to-book assets ‡ 1.20 1.22
(0.0095) * * * (0.0263) * *
Industrial concentration ‡ 0.30 0.46
(0.3741) (0.2544)
Book-to-market value of assets 1.29
(0.0883) *
Dividend yield ‡ 5.36
(0.2778)
Return-on-book assets ? 1.59 2.36
(0.3005) (0.2890)
FIM interest rateÐ FIM basket 0.29 0.21
interest rate (0.0024) * * * (0.0555) *
Year 1985 dummy ? 0.17 0.53
(0.6969) (0.3647)
Year 1986 dummy ? 0.30 0.17
(0.3868) (0.6845)
Year 1987 dummy ? 0.50 1.32
(0.1357) (0.0069) * * *
Year 1988 dummy ? 0.70 1.03
(0.0199) * * (0.0068) * * *
Year 1989 dummy ? 0.66 0.87
(0.0278) * * (0.0095) * * *
Year 1990 dummy ? 0.57 0.37
(0.0568) * (0.2430)
Pseudo R-squared 0.224 0.254
N 321 269

Notes: The results are obtained from a probit regression where the dependent variable obtains the value of
one if a loan is denominated in a foreign currency and zero otherwise. The number of observations differs
from full sample size because the variable exports-to-net sales is not available for all companies and because
some companies were not listed the whole sample period. Two-tailed p-values are in parentheses. *, * *, and
* * * denote significance at the 10%, 5%, and 1% level, respectively. Pseudo R2 is computed as
1 ln L(
)=ln L(!), where ln L(
) is the value of the likelihood function evaluated at the maximum
likelihood estimates and ln L(!) is the maximum value of the likelihood function under the hypothesis that
all independent variables equal zero.

# Blackwell Publishers Ltd, 2001

{Journals}eufm/7_4/e125/makeup/e125.3d

490 Matti Keloharju and Mervi Niskanen

positive sign.3 Similarly, the coefficient for the multinational dummy (indicating
whether the firm has foreign subsidiaries or not) is expected to be positive.
Second, we include a measure for firm size: the log of total book assets. Large
companies are likely to have a larger international reputation and a more direct access
to international capital markets than small companies. Since borrowing from foreign
sources is likely to be denominated in foreign currencies, the coefficient for the
variable log of total assets is expected to be positive.
Third, the expected costs from financial distress are positively related to leverage
and negatively related to profitability. Therefore, hedging by using currency debt (as
opposed to using markka debt) is more valuable for companies which have high
leverage ratios or low profitability. Thus the expected coefficient for the ratio of debt-
to-total book assets is positive.4 The expected coefficient for profitability, proxied by
return-on assets, is less clear-cut. While the financial distress argument implies that the
expected coefficient for ROA should be negative, reputational effects suggest a
converse relation. High ROA companies are more reputable and as a result are
expected to have an easier access to foreign debt.5
Fourth, firms that are industrially diversified are less likely to benefit from hedging.
Following Berger and Ofek (1995) our measure of industrial concentration is the
revenue-based Hefindahl index, computed here from two-digit level SIC code data.
We expect the coefficient for the industrial concentration variable to be positive.
Fifth, firms with more growth options in their investment opportunity set are more
likely to undertake a hedging program aimed at reducing the variance in firm value
(Froot et al., 1993; Nance et al., 1993). Since firms with promising growth
opportunities are likely to have small book-to-market ratios, the coefficient for the
book-to-market ratio is expected to be negative.
Sixth, as pointed out by Nance et al. (1993), a firm's hedging policy is also
influenced by its other financial policies. For example, a firm could substitute a
hedging policy and decrease the risk of default by restricting its dividends. Therefore,
we expect that the dividend yield is positively related to the probability that a firm
hedges by raising foreign currency denominated debt.
Seventh, the expected costs of finance are likely to affect the choice between foreign
and domestic currency denominated debt. The Finnish government permits unrealized

3
Ideally one would like to examine how foreign currency loans are used for hedging the net
exposure in each currency. Unfortunately data on the sample companies' net exposures or sales
distribution by currency were not available. Therefore, we rely on this relatively crude measure
of currency exposure.
4
Note that the decision to use foreign debt (and debt in general) is endogenous. For example,
the issue to use foreign debt may be related to past debt, both domestic and foreign. Therefore,
it is possible that firms that hedge actually have ex post higher leverage, as hedging has allowed
them to take on additional debt. Given that our data lists debt transactions from the sample
period, but not the amount of currency or markka debt outstanding at any given point of time,
it would be very difficult to meaningfully relate new loans to existing debt, particularly in the
early years of our sample period when the loan stock for each company is largely unknown.
Therefore, we do not make any attempt to control for the effect of endogeneity on the results.
5
In principle this reputation effect could be controlled for by including variables measuring the
credit rating in the regression. Unfortunately this is not possible in our study since none of the
44 sample companies had a long-term credit rating during the sample period.

# Blackwell Publishers Ltd, 2001

{Journals}eufm/7_4/e125/makeup/e125.3d

Foreign Currency Denominated Debt 491

exchange losses on foreign currency loans to be tax deductible immediately, while

taxes on exchange gains are deferred until realized. Shapiro (1984) shows that such tax
laws encourage companies operating in Finland to prefer (on an expected value basis)
to borrow the foreign currency, regardless of whether it is weaker or stronger,
provided that the International Fisher Effect holds before tax. Moreover, Shapiro
shows that the weaker the markka is, the smaller are the tax savings from borrowing
the foreign currency. In other words, the larger the interest rate differential between
the Finnish markka and its basket, the smaller the incentives for firms to raise foreign
currency denominated debt. We thus expect the coefficient for the interest rate
differential to be negative.
Eighth and finally, we include year dummies to control for the increasing demand
for foreign currency debt after the restrictions on that particular loan type were lifted.
The variable exports-to-net sales has the expected sign and is highly significant (p-
value from a two-tailed test less than 0.07%). This is consistent with the hypothesis
that companies raise currency debt to hedge their foreign exchange exposures.
However, as pointed out by the referee, this result also is consistent with the following
interpretation. Firms with extensive exports are likely to have a better international
reputation than companies that export less. This is likely to improve their ability to
raise foreign debt.
The variable log of book assets is also positive as expected and highly significant.
Thus the results support the notion that large firms with a generally better
international reputation and a wider access to the international capital markets are
more likely to raise foreign currency debt.6
The results for the other variables are less clear-cut. The coefficient for the
differential between the FIM and FIM basket interest rates is in both specifications
significantly or almost significantly positive (in the first specification the p-value
is 0.002, in the second 0.06), the opposite of what was predicted by the tax
hypothesis.
Our explanation to this finding is that the interest rate differential variable also
captures the effect of speculation. The results are a likely indication that the financial
managers of the sample firms did not believe that the International Fisher Effect held
at the time of borrowing; instead, they may have believed that they could reduce the
costs of finance by raising foreign currency denominated loans (which on average have
a lower nominal interest rate than the Finnish markka), and hedging their position if
the devaluation of the Finnish markka would become an issue. This is understandable
given VireÂn's (1988) evidence that the uncovered interest rate parity relationship was
very weak or nonexistent for the Finnish markka in 1980 ± 87.
An investigation of the sample firms' choice of foreign currency and the timing of
their borrowing gives further support for the interpretation that speculative motives

6
Large companies typically export relatively more than small companies: the Pearson
correlation coefficient between the variables exports-to-net sales and log (total book assets) is
0.44. The resulting collinearity problem can be assessed by checking whether the coefficients on
the two variables and the interpretation of the results are affected if either of the two variables is
dropped from the model. When we do this, the significance level for the variable left in the
model increases. The interpretation of the results does not change, however, suggesting that the
collinearity effect is not serious.

# Blackwell Publishers Ltd, 2001

{Journals}eufm/7_4/e125/makeup/e125.3d

492 Matti Keloharju and Mervi Niskanen

have an effect on financial managers' decision making.7 Despite of Shapiro's (1984)

tax argument that Finnish firms should have preferred currencies with high nominal
interest rates, the sample firms tended to borrow in periods when the nominal interest
rate for the loan currency, relative to other currencies, was lower than during the
sample period in general. In other words, the sample firms did not appear to believe
that the International Fisher Effect held for foreign currencies. The sample firms'
preference for borrowing in periods when the nominal interest rate for the loan
currency is lower than usual is significant at the 10% level for variable rate loans and
at the 35% level for fixed rate loans (one-tailed test).
In the probit estimations, the coefficients for industrial concentration, dividend
yield and return-on-book assets are of their expected sign but are far from significant.
The signs for the multinational dummy differ across specifications. Moreover, the
variables debt-to-book assets and book-to-market value of assets have the wrong sign
but are significant at the 10% level or lower. In other words, the variables that are
related to the hedging policies of the firm in general but have no direct link to hedging
foreign exchange exposures have much less to do with the probability of raising
foreign currency debt than the variables that are directly linked to foreign exchange
exposures (exports-to-net sales) or to the firm's access to foreign sources of finance
(log of total book assets). Although not reported here, this result does not change
qualitatively if we use the current ratio instead of the dividend yield or the interest
coverage ratio instead of the ratio of debt to book assets.

4.3. Additional tests

The above probit estimations do not attempt to take advantage of one potentially
important observable variable: the firm that raised the loan. On average, the sample
firms raised eight loans during the sample period, suggesting that there is substantial
correlation among observations and identical observations (except for the interest rate
differential) for all the loans in each year for a given firm. Therefore, a firm-specific

7
We use bootstrapping to document the sample firms' preference for borrowing in periods when
the nominal interest rate for the loan currency is lower than usual. The analysis consists of six
stages. We first rank in each month the nominal interest rate of the 12 foreign loan currencies
used by the sample firms. Second, we register the rank of each loan currency in the month the
loan was raised. This gives us 68 ranks for fixed rate loans and 79 ranks for variable rate loans
(the total number of ranks is smaller than the total number of foreign currency loans because
the sample includes 38 foreign currency basket loans that are not considered in this analysis).
Third, we determine the benchmark against which we can compare the sample firms' interest
rate preferences. We do this by pairing the nominal interest rate rank of each loan with a
randomly selected interest rate rank in the loan currency (the benchmark rank is selected from
among 84 ranks in the loan currency, one rank for each month within the sample period). This
makes sure that the sample loans and the benchmark have an identical currency distribution,
and that the results are not affected by the sample firms' potential hedging-driven preference for
currencies that have lower-than-average nominal interest rates. Fourth, we perform a non-
parametric rank sum test which compares the interest rate ranks for the loans to those for the
benchmark. Fifth, we repeat the random selection of the benchmark vector and the rank sum
test 999 times. Sixth and finally, we analyse the distribution of the results from the rank sum
tests. The results suggest that in 90% of the cases, variable rate loans have a lower interest rate
than the benchmark. The corresponding number for fixed rate loans is 65%.

# Blackwell Publishers Ltd, 2001

{Journals}eufm/7_4/e125/makeup/e125.3d

Foreign Currency Denominated Debt 493

variable might capture some of the unexplained variation in the probability of raising
currency debt. As a robustness test, we estimate a random effects probit model that
takes into account this panel feature in the data. The random effects model assumes
that, in addition to the intercept term and the regressors in the equation, the model
includes a random disturbance component ui , which characterises the ith observation
and is constant through time when the firm is used as a grouping variable. This
random disturbance component can be viewed as a collection of factors not in the
regression that are specific to each firm (Greene, 1997, p. 623 ±624).
Estimations of the random effects probit model (not reported here) suggest that the
firm-specific effects are generally not significant; a likelihood ratio test suggests that
the more complicated random effects model performs no better than the simple probit
model. The results from the tests of the hypotheses are qualitatively similar for both
models.
As an additional robustness test, and analogously to Allayannis and Ofek
(forthcoming), we run a probit regression where the dependent variable is a dummy
for whether a company raised at least one currency loan in a given year, conditional
on that it raised at least one loan of any type (currency or Finnish markka loan) in
that year. The results (not formally reported in the paper) suggest that the variable
exports-to-net sales is highly significantly positively related to the probability of
issuing currency debt. Moreover, the results from the second stage of a Cragg two-
stage regression suggest that exports-to-net sales is highly significantly positively
related to the fraction of currency loans raised by a given firm in a given year. The
coefficient for interest rate differential is positive but generally not significant in both
the probit model and the second stage of the Cragg model. The decrease in the
significance level for the interest rate differential variable is expected, since
aggregation of the data from an individual loan level to a level where the firm-year
intersection is the unit of observation throws away data. Moreover, aggregation
eliminates all within-year variation in the interest rate differential variable, decreasing
the accuracy with which speculative demand for currency loans can be linked to
relative interest rate changes.
To check for robustness of the results across time periods, we estimate the probit
regression for subperiods 1985± 87, 1988± 89, and 1990± 91. Although not formally
reported in the paper, the variable exports-to-net sales is positive and highly
significant in all subperiods. The interest rate differential is highly significantly
positive in the 1990 ±91 subperiod but insignificant in other periods. This is consistent
with the notion that the speculative demand for currency loans was particularly
evident in the eve of the devaluation which occurred in November 1991.
Kedia and Mozumdar (1999) document that firms' rationale in borrowing in Swiss
francs and Japanese yen is different from the rationale for the other currencies. We
investigate this issue by running separate probit regressions for the currencies with the
largest number of observations (US$, DEM, JP¥, CHF). Our results suggest that the
key variables Ð the exports-to-net sales and the interest rate differential Ð are positive
and highly significant for the US dollar and the German mark, the currencies with the
largest number of observations. Consistent with Kedia and Mozumdar (1999), the
Swiss franc has the lowest coefficient for the exports-to-net sales variable Ða possible
indication of other than hedging driven demand for Swiss franc loans. However, one
must note that the small number of Swiss franc and Japanese yen loans (14 loans of
both currencies) makes it difficult to make firm conclusions about these currencies. In
fact, the yen displays the largest coefficient for the exports-to-net sales variable and
# Blackwell Publishers Ltd, 2001
{Journals}eufm/7_4/e125/makeup/e125.3d

494 Matti Keloharju and Mervi Niskanen

yet it is insignificant at conventional levels, a manifestation of the low power of the

test when there are few observations.
Stulz (1996) suggests that firms may speculate by creating an exposure through
foreign exchange derivatives. Although our sample does not include data on the use
of foreign exchange derivatives, it is occasionally possible to make inferences of
speculative behaviour from the use of currency debt. For example, more than a
quarter of all loans (12 of 44) raised by the companies with no exports were
denominated in foreign currency. The paucity of a natural exposure for these
companies, and the fact that currency swaps were very rare during the sample period,
hints that these loans may have been raised for speculative purposes. Similarly, some
of the Swiss franc and yen denominated loans have probably not been used to hedge
Swiss franc or yen denominated receivables. For example, the chief financial officer of
one major sample company told us in an interview that the company raised a yen loan
in early 1991 to take advantage of the favourable interest rate Ð the company did not
have any natural exposure in yen and did not hedge its yen loan. The FIM value of the
loan more than doubled by the time it was repaid after five years.
So far we have shown evidence from firm level behaviour that is consistent with
hedging and with speculation. To complement this evidence with the market's reaction
to the use of foreign debt, and to make an attempt to better disentangle the hedging
and speculative motives from each other, we apply the two-stage method developed by
Jorion (1990) and others, as specified in Allayannis and Ofek (1999). At the first stage
we regress the monthly returns for a stock from the sample period 1985± 91 against the
Helsinki Stock Exchange general index and exchange rate index for the Finnish
markka. At the second stage, we regress the estimated (firm level) coefficient for the
exchange rate index against average exports-to-net sales (across the sample years) and
the fraction of the number of currency loans to the number of all loans for that firm.
The results suggest that none of the variables at the second stage are significant: the
t-value for average exports-to-net sales is 0.55, whereas the fraction of currency
loans has a t-value of 0.02. The results are similar if the fraction of currency loan
volume (in terms of FIM) is used as an independent variable, or if the exchange rate
beta is computed with respect to an individual currency index (we try US$=FIM,
DEM=FIM, JP¥=FIM, CHF=FIM, and GB£=FIM indices). The insignificance of the
results probably reflects the fact that the exchange rate betas are relatively noisy
estimates of exchange rate exposures, a fact that is exacerbated by thin trading on the
Helsinki Stock Exchange.8

5. Conclusions

This study examines the determinants of the decision to raise foreign currency debt by
taking advantage of a unique and comprehensive sample of private and public debt
raised by 44 Finnish corporations. The results suggest that hedging is an important

8
As an additional check, we investigate whether the market beta is associated with a company's
use of foreign debt. We do not find evidence of such an association. The Pearson (Spearman
rank) correlation between the foreign debt ratio and the market beta is 0.06 (0.27); neither of
these correlations is significant at conventional levels. Similarly, the market betas for companies
that raised currency debt are not significantly different from the betas for companies that did
not raise currency debt.

# Blackwell Publishers Ltd, 2001

{Journals}eufm/7_4/e125/makeup/e125.3d

Foreign Currency Denominated Debt 495

determinant of the currency-of-denomination decision: firms in which exports

constitute a significant fraction of net sales are most likely to raise foreign currency
debt. Large firms, with a wider access to the international capital markets, tend to
borrow in foreign currencies relatively more often than small firms. We also find
evidence that speculative reasons have an effect on the choice of loan currency.
Despite of a conflicting tax motive, the sample firms tended to borrow in periods when
the nominal interest rate for the loan currency, relative to other currencies, was lower
than during the sample period in general.

References
Allayannis, G. and Ofek, E., `Exchange rate exposure, hedging, and the use of foreign currency
derivatives', Journal of International Money and Finance, forthcoming.
Berger, P. and Ofek, E., `Diversification's effect on firm value', Journal of Financial Economics,
Vol. 37, 1995, pp. 39 ±65.
DeBondt, W. and Thaler, R., `Financial decision-making in markets and firms: a behavioral
perspective', in: R. Jarrow, V. Maksimovic and W. Ziemba (eds), Handbooks in Operations
Research and Management Science: Finance (Amsterdam: North-Holland, 1995).
DeMarzo, P. and Duffie, D., `Corporate incentives for hedging and hedge accounting', Review
of Financial Studies, Vol. 8, 1996, pp. 743± 771.
Doukas, J., Hall, P. and Lang, L., `Exchange rate exposure at the firm and industry level',
Working Paper (Old Dominion University, 2000).
Fama, E., `Forward and spot exchange rates', Journal of Monetary Economics, Vol. 14, 1984,
pp. 319± 338.
Froot, K., Scharfstein, D. and Stein, J., `Risk management: Coordinating corporate investment
and financing policies', Journal of Finance, Vol. 48, 1993, pp. 1629± 1658.
Geczy, C., Minton, B. and Schrand, C., `Why firms use currency derivatives', Journal of
Finance, Vol. 52, 1997, pp. 1323± 1354.
Giddy, I., Global Financial Markets (D.C. Heath, 1994).
Greene, W., Econometric Analysis (New Jersey: Prentice Hall, 1997).
Hakkarainen, A., Joseph, N., Kasanen, E. and Puttonen, V., `Foreign exchange exposure
management among Finnish firms', European Financial Management, Vol. 3, 1997, pp. 255±268.
James, C., `Some evidence on the uniqueness of bank loans', Journal of Financial Economics,
Vol. 19, 1987, pp. 217±235.
Jorion, P., `The exchange-rate exposure of U. S. multinationals', Journal of Business, Vol. 63,
1990, pp. 331±345.
Kedia, S. and Mozumdar, A., `Is foreign currency denominated debt a hedging instrument?',
Working Paper (Harvard University, 1999).
Levi, M., International Finance (Singapore: McGraw-Hill, 1990).
Mayer, C., `Financial systems, corporate finance, and economic development', in: G. Hubbard
(ed.), Asymmetric Information, Corporate Finance, and Investment (Chicago: University of
Chicago Press, 1990).
Nance, D., Smith, C. and Smithson, C., `On the determinants of corporate hedging', Journal of
Finance, Vol. 48, 1993, pp. 267± 284.
Niskanen, M., `Corporate borrowing costs and the benefits of lending relationships when banks
hold equity claims in the borrowing firms', Licentiate thesis, Helsinki School of Economics
and Business Administration, 1996.
Rhee, G., Chang, R. and Koveos, P., `The currency-of-denomination decision for debt
financing', Journal of International Business Studies, Vol. 16, 1985, pp. 143± 150.
Shapiro, A., `The impact of taxation on the currency-of-denomination decision for long-term
foreign borrowing and lending', Journal of International Business Studies, Vol. 15, 1984,
pp. 15 ± 25.

# Blackwell Publishers Ltd, 2001

{Journals}eufm/7_4/e125/makeup/e125.3d

496 Matti Keloharju and Mervi Niskanen

Smith, C. and Stulz, R., `The determinants of firms' hedging policies', Journal of Financial and
Quantitative Analysis, Vol. 20, 1985, pp. 391± 405.
Stulz, R., `Optimal hedging policies', Journal of Financial and Quantitative Analysis, Vol. 19,
1984, pp. 127± 140.
Stulz, R., `Rethinking risk management', Journal of Applied Corporate Finance, Vol. 9 (Fall),
1996, pp. 8 ± 24.
VireÂn, M., `Determination of interest rates in small open economies: a review of recent
evidence', Discussion Paper No. 16=88 (Bank of Finland, 1988).

# Blackwell Publishers Ltd, 2001