Political Risk and Irreversible Investment
*Koç University, Istanbul, Turkey and Centre for Economic Policy Research, London, UK, e-mail: saltug{at}koc.edu.tr
Carleton University, Canada, e-mail: fdemers{at}connect.carleton.ca and mdemers{at}connect.carleton.ca
 |
Abstract |
|---|
 Top Abstract 1 Introduction2 A case study of...3 A theoretical framework4 Numerical results4.1 The long-run impact5 ConclusionsAppendixA Data AppendixB The regime shift...C Parameterizing the modelReferences |
|---|
The objective of this article is two-fold. First, we develop a theoretical model to investigate the impact of political risk on irreversible investment. Second, we apply our model to an analysis of the effects of risk of separation of the province of Quebec from the Canadian federation. We model the probability of a regime switch using the properties of the electoral process and examine the response of investment to changes in the risk of separation. We consider the impact of investors' perception of the risk of separation and financial market volatility separately. We show that political risk has a depressing impact on investment even if the "bad" regime has never been observed in the sample. (JEL Codes: E22, D92, O16, O11)
Key Words: Irreversible investment • political risk • regime shifts • Quebec investment • Canada.
|
1 Introduction |
|---|
|
TopAbstract1 Introduction2 A case study of...3 A theoretical framework4 Numerical results4.1 The long-run impact5 ConclusionsAppendixA Data AppendixB The regime shift...C Parameterizing the modelReferences |
|---|
In emerging and developing economies investors are often faced with political risk due to the risk of expropriation, disruptions in market access, unfavorable government regulations, currency, debt and fiscal crises, policy reversals, political upheaval, etc... Investors in developed economies are also exposed as the debates on outsourcing reveal. Yet there has been little work in quantifying the effect of political risk on real investment decisions.
A literature has examined the impact of political risk on foreign direct investment (FDI) flows and capital flight. In his provocative article, Lucas (1990) lists political risk as one of the factors behind the puzzling lack of capital flows from rich to poor countries. Svensson (1998) argues that the lack of political stability leads to lower investment because political instability impedes the incentive to create an efficient system of property rights. Lensink, Hermes and Murinde (2000) find that after controlling for macroeconomic and policy variables, political risk stimulates the magnitude of capital flight. Janeba (2002) considers the investment decisions of a multinational firm in a politically risky world and emphasizes the trade-off between lack of credibility and low cost in the determination of FDI decisions. In a different vein, Kim and Mei (2001) study the impact of political news on stock market volatility for the Hong Kong stock market while Lobo (1999) finds that US stock returns responds differently to the type of election and to the party of the administration in power. [See also Mei and Guo (2004).] Political risk regarding the future of European Monetary Union is also widely perceived as affecting the behavior of financial markets. Remolona, Wickens and Gong (1998) find that the real risk premium on UK bonds fluctuated considerably since the exit from the ERM, reflecting uncertainty about the real economy.
There is also a growing literature that studies the economic effects of political integration and separation. According to one view, as economic integration increases, the political costs of separation may decrease. This view is expounded by Alesina, Spolaore and Wacziarg (2000, 2005), who examine the relation between openness in trade and the equilibrium number and size of countries. They find that the economic benefits of country size are mitigated by the degree of openness to trade, and also that the history of the creation of nation-states depends on the trade regime. An alternative view suggests that political separation would lead to a change of political constituency. Hence, the policy of a seceding country would be determined by the citizens of that country alone. Braun, Hausmann and Pritchett (2004) show that political disintegration has ambiguous effects on the quality of policies and negative effects on economic growth because it reduces economic market size. The role of political stability and economic integration in fostering trade has been stressed by McCallum (1995), Helliwell (1996) and Anderson and van Wincoop (2001). Using a theoretically well grounded gravity model Anderson and van Wincoop (2001) finds that trade between Canada and the US is lowered by 45 percent due to the modest border between the two countries. Trade among Canadian provinces is six times larger, while trade among US states is 25 percent larger thus indicating that for a small country the border effect is quantitatively more important.
The objective of this article is two-fold. First, we develop a theoretical model to investigate the impact of political risk on irreversible investment. Second, we apply our model to an analysis of the risk of separation of the province of Quebec from the Canadian federation on Quebec investment. The case study of Quebec provides for a unique "natural experiment" of the impact of political risk in a developed economy. The issue has existed for almost 50 years and the data are of high quality. Two episodes of political risk are clearly identifiable: the 1970's and the 1990's. In this article, we seek to determine the quantitative impact of the risk of separation on Quebec investment for the second episode of political risk, namely the 1990's.
As a way of modeling the presence of political risk, we allow for regime switches between a "good" regime and a "bad" regime (separation), where the former is characterized by more "favorable" distributions for the state of demand or productivity. While the firm knows which regime it is currently residing in, each period it must assess the transition probabilities on the basis of a vector of economic and political variables. We introduce an electoral process whereby a federalist party (L) and a separatist party (PQ) compete for power. When the latter is elected it may hold a referendum on separation and a switch to the "bad" regime becomes possible.
The arguments in the recent literature provide some justification for modeling political risk in this way. Despite the countervailing arguments regarding the role of increased economic integration on the impact of separation or secession, one of the major impacts of separation appears to lie in the reduction in trade and market access due to the erection of borders between Quebec and the rest of Canada. More specifically, in the event of separation, firms in Quebec would expect: (1) reduced market access to the rest of Canada due to the passage from a highly integrated economic union in the Canadian federation to a looser form of economic union such as a customs union; (2) reduced short-run and possibly long-run access to NAFTA partners’ markets; (3) loss of the monetary union with the rest of Canada; (4) a high debt to GDP ratio as Quebec would have to pay its share of the federal debt. Firms would also expect a higher long-term risk premium. Therefore, even in the absence of a crisis, at least a recession would be expected.
Separation could also lead to the expectation of a crisis, or "sudden stop" phenomenon whereby much higher interest rates and a more severe loss of output and productivity would be expected following separation. Such a situation might occur if the rest of Canada were to refuse to negotiate or that negotiations were to prove arduous. In 1995, the Quebec referendum question stipulated that failing an agreement with the rest of Canada, Quebec would proceed with a unilateral declaration of independence (UDI). However, a UDI would provoke a conflict of legitimacy (since the electorate is highly divided). There would be massive capital flight leading to a banking crisis as individuals would expect that a separate Quebec would attempt to create a Quebec dollar in order to devalue. The analogy to the Mexican (1994), Argentinian (2001) and the East Asian crises is immediate. Since the Quebec economy is highly sophisticated, financial capital is highly mobile, and the amplitude of the crisis could be even greater.
Our analysis is based on the profit-maximization model of a risk neutral representative monopolistically competitive firm under uncertainty and irreversibility. The firm-specific or industry-specific nature of most investment goods implies that investment decisions are, at least largely, irreversible, and therefore more sensitive to uncertainty due to an option value of waiting, or to an endogenous risk premium.1 The experience in Quebec features declines in investment together with increases in financial volatility. We argue that these events are consistent with the presence of political risk as being responsible for the decline in investment in Quebec. Bittlingmayer (1998) presents a similar argument for Germany in its transition from Empire to the Weimar Republic. As in our analysis, he attributes this phenomenon to the "bad news" principle and the option value of waiting in models with investment irreversibility under uncertainty. In our framework, the perceived threat of separation can affect investment in two ways. The first is what we call the political risk channel. According to this channel, it is the perceived likelihood of transiting to a regime with less favorable fundamentals that is responsible for the decline in investment. The second is directly through increases in real interest rates or the stochastic discount factor firms use to discount future cash flows. We call this the financial market effect of political risk.
Our quantitative analysis proceeds in several parts. First, we analyze the public's perceptions of political risk using data on poll/election/referendum results and financial market variables. Second, we use annual sectoral data on investment and capital stocks to provide a simple test of the impact on political risk on investment behavior. For this purpose, we take Ontario as a point of comparison, and show that the investment-capital stock ratios in machinery and equipment investment for all broadly defined sectors of the economy were significantly lower in Quebec during the second period of political risk. Third, we conduct a simulation exercise to examine the impact of political risk on investment behavior. Our approach to simulating the model is to ask what must investors’ perceptions for fundamentals in a likely separation state be so that the model can generate the observed declines in investment. This is similar to the reverse calibration that Danthine and Donaldson (1999) implement in their application.2
The article is organized as follows. Section 2 discusses political risk in the context of Quebec. Section 3 describes the model. Section 4 describes the quantitative impact of regime shifts and an absorbing state on investment, and discusses the role of stochastic interest rates. Some concluding remarks are in Section 5.
|
2 A case study of Quebec |
|---|
|
TopAbstract1 Introduction2 A case study of...3 A theoretical framework4 Numerical results4.1 The long-run impact5 ConclusionsAppendixA Data AppendixB The regime shift...C Parameterizing the modelReferences |
|---|
2.1 A brief historical perspective
Two major episodes of political risk have had an important impact on the Quebec economy, and particularly, on Quebec investment. The first is that of the 1970s, marked by the provincial election in 1976 of a separatist party, the Parti Québécois (PQ). When the PQ finally held its referendum in 1980, it was defeated, with 60 percent of the voters being opposed to giving a mandate to the Quebec government to negotiate "sovereignty-association"3 with the rest of Canada. The defeat of the PQ in the referendum led to a period of political calm in the 1980's. The PQ was re-elected in 1981, but only after shelving its sovereignty plans. Subsequently, it was defeated in December 1985 by its federalist opponent.
The second episode of political risk is that of the 1990s, marked by the federal government's failed attempts at constitutional reform (the "Meech Lake Accord"), which led to a rise in popularity of the separatist option in Quebec (see Table 1), the subsequent election of the PQ in 1994, and another referendum on sovereignty with a "European style" partnership very narrowly lost by the PQ in 1995.4 The margin of 0.8 percent encouraged separatist leaders to promise another referendum which maintained uncertainty until the 1998 provincial elections when the federalists succeeded in obtaining more votes than the PQ although the latter managed to obtain more seats in the legislature and to form the government. As a result, the PQ temporarily abandoned its plans for another referendum.5
|
2.2 Perceptions of political risk: poll results and financial market data
What do Quebecers and financial markets think about the likelihood of separation of Quebec from the Canadian federation? While there is no direct observation of this perception, we look at some indicators such as opinion polls, election and referendum results as well as financial market data such as bond spreads.
In Table 1, we establish what we will refer to as the "raw" poll data (given in the second column of Table 1). For years during which there was an election or a referendum, we give preference to these results over poll results. For years during which no survey of opinion poll was conducted, we use an approximate figure in view of the political events of the time.6
This table clearly reveals that there were two major periods of political risk (1976–80 and 1990–98) separated by a period of stability. It is worth noting that opinion polls report the voting " intentions" of Quebecers in a referendum on independence. Yet some of these intentions do not actually materialize into actual "Yes" votes in a referendum, and hence overestimate the actual support for sovereignty.7 Hence, the raw poll data in Table 1 may overestimate the perceived support of Quebec sovereignty. In Table 1, we also report the predicted values from a 3-year moving average applied to the Yes votes in the raw poll data.8
To obtain an alternative indicator of expectations, we also consider data on financial market variables. Johnson and McIlrath (1998) report that opinion poll announcements during the 1995 Quebec referendum that increased the likelihood of Quebec sovereignty markedly increased spreads of Quebec bond yields over those of both Canada and Ontario yields. Figure 1 displays the difference between the spreads on 10-year bonds for Quebec and Ontario and Canada, respectively.9 The first panel of Figure 1 shows that spreads between 10-year bonds for Quebec and Ontario increased in 1990 and early 1991 following the failure of the Meech Lake Accord. However, they declined for the remainder of 1991, and for 1992 and 1993 as investors became convinced that fundamental support for separation was lower than 50 percent and that the (federalist) Premier of Quebec would work to diffuse tensions. The difference between the bond spreads rose again in 1994 and 1995 with the provincial election and the referendum but eventually fell after the referendum.
|
The perception of political risk can also be observed by analyzing the movements in the Canada–US bond spreads. Focusing on the 1990–91 episode, the high long-term bond spreads are partly ascribable to the Bank of Canada's concern about preventing the Canada–US exchange rate from slipping, and partly (from 1991 on) to the radical inflation reduction strategy adopted by the Bank of Canada.10 However, an additional effect came into play in 1990: the spread was fairly high during the entire Meech Lake episode (June 1990), that is, both before the rejection and after. The high spreads in 1995 can be directly ascribed to the effects of political risk.
2.3 Investment in Quebec
In this article, we analyze the impact of political risk on the investment decisions of an average (or representative) Quebec firm. First, we examine the relative investment performance in two Canadian provinces, Quebec and Ontario.11 We consider the behavior of investment in machinery and equipment (M&E) and structures for three broadly defined sectors, namely, manufacturing industries, a broadly defined business sector (which includes manufacturing industries plus construction, transportation, trade, and other services industries) and total industries.12 Sectoral data on investment and capital stocks include both public and private investment expenditures. To the extent that political risk had an additional effect of reducing public investment in Quebec relative to Ontario, this will be reflected in our data. Figures 2 and 3 display the investment-capital stock ratios for Quebec and Ontario (measured on the right axis) versus the smoothed poll results (shown on the left axis). The smoothed poll results,13 clearly indicate the rise in separatist sentiment and political uncertainty from 1990 to 1998 preceded by political stability from 1981 to 1989.
|
|
Table 2 provides a simple test of the impact of political risk on investment in Quebec by testing for the equality of the average investment-capital stock ratio for Quebec versus Ontario across periods of political risk and political stability. To net out the effect of Canadian factors such as the recessions of 1980–82 and 1990–91, the second episode of political risk (1990–98) is compared with the second period of political stability (1981–89). Define the random variable djt
[(I/K)Que/(I/K)Ont]jt,where j = 0 refers to the episode of political stability (1981–89) and j = 1 refers to the episode of political risk (1990–98), respectively. Let µ0 and µ1 denote the respective population means. Under the null hypothesis, the expected value of djt should be equal across the periods of political stability and political risk. From Table 2, the null hypothesis that µ0 = µ1 versus the alternative that µ0 > µ1 can be rejected at the 5 percent for M&E investment across all three sectors. (The t-value for a one-sided test of the hypothesis that µ0 = µ1 is 1.76 at the 5 percent level of significance.) The test statistic for structures investment in manufacturing investment is positive and significant at the 10 percent level but not 5 percent. However, when we consider the broadly defined business sector and total industries, we find that we can reject the null hypothesis of µ0 = µ1 in favor of µ1 > µ0. Indeed Figure 3 shows that structures investment for broader industry groupings remained high in Quebec owing to the large investment boom in manufacturing structures between 1988 and 1990, due to a strong anticipatory reaction to the 1988 Canada–US Free-Trade Agreement (FTA). Following Alesina, Spolaore and Wacziarg (2000, 2004), this suggests that as economic integration increases, the cost of political separation for Quebec may decrease, contrary to our original assertions. However, overall we find that the episode of political risk in the 1990's is associated with a significant decline in the investment-capital stock ratio for investment in M&E in Quebec relative to Ontario and similarly for investment in structures for manufacturing industries.
|
|
3 A theoretical framework |
|---|
|
TopAbstract1 Introduction2 A case study of...3 A theoretical framework4 Numerical results4.1 The long-run impact5 ConclusionsAppendixA Data AppendixB The regime shift...C Parameterizing the modelReferences |
|---|
The capital stock may be highly firm-specific or industry-specific such as in aeronautics or aluminum processing, and industry level uncertainty may affect all firms similarly. Hence, if firms wish to sell their excess capital in response to an adverse shock, they may not be able to find buyers. Even for less firm- or industry-specific capital goods, there may exist a "lemons" problem of adverse selection in the market for used capital that may similarly prevent firms from disinvesting. In what follows we abstract from resale markets altogether (i.e., we assume complete irreversibility) which allows for a simpler framework within which to analyze the impact of political risk. In this setting uncertainty has a particularly important impact on investment. (See Demers (1991) and Altug, Demers and Demers (1999).)
3.1 The basic model
We consider a monopolistically competitive risk neutral firm which makes variable input and investment decisions each period. At time t it produces output, Yt, using its beginning-of-period capital stock, Kt, and a variable labor input, Lt. The firm's production function is given by Yt = F(Kt,Lt,At), where At is a stochastic technology shock and F is twice continuously differentiable, increasing, concave, and satisfies the Inada conditions. Let pt denote the stochastic output price. We assume a constant elasticity demand function: 
where 
< –1 is the price elasticity of demand and 
t is a stochastic parameter representing the state of demand. Denoting the wage by wt define the short-run profit function at t as: 
. 
is continuous; increasing in Kt, At, and t, decreasing in wt; strictly concave in Kt, t, and At and bounded for finite Kt,t, At, and wt. Let r denote the real rate of interest. Define the tax-adjusted price of investment as 
, where 
denotes the purchase price of investment goods, zt is the present value of tax deductions on new investment at date t, and 
t is the investment tax credit at time t as a percentage of the price of the investment good14. The firm's after-tax cash flow at time t, Rt, is given by 
where 
t is the corporate tax rate, and It is the firm's rate of gross investment measured in physical units at time t. Let ht (At,t). Assume that ht takes on values in the finite set H and define f(ht + 1|ht) as the conditional density of ht + 1. Let primed variables denote next period's values. The firm's problem can be expressed recursively as
|
| (3.1) |
)K + I, the irreversibility constraint I 
0 and K given, where V denotes the value function, 0 < < 1 is the depreciation rate, and β = (1 + r)–1 is the discount factor. Reverting to time subscripts, the first-order necessary and sufficient conditions for the optimization problem at t are
|
| (3.2) |
be such that: 
When investment is irreversible the firm can be viewed as holding a call option to invest. Define 
as the call option of investing an additional unit at t + 1.15 Also,
|
| (3.3) |
, 3.2 becomes:
|
| (3.4) |
K is the partial derivative of with respect to Kt+1. On the right-hand side of (3.4) the marginal benefit of investing an additional unit appears as the sum of next period's discounted expected marginal profits and of the expected value of the undepreciated portion of the capital stock assuming the firm does not invest in the next period. The sum of these benefits can alternatively be expressed as Qt + 
t where Qt represents the expected marginal value of capital if the firm never invests or disinvests from t + 1 onward and simply allows its capital stock to depreciate and t captures the increase in value due to only future call options. On the left-hand side is the total cost of investing which is the sum of the purchase price of investment and of the loss of option value of waiting when the firm invests an additional unit. The option to invest will be exercised only when ht is sufficiently high or when business conditions are sufficiently favorable. If, on the other hand, ht 
the firm finds itself in the inaction zone so that Kt + 1 = (1 – )Kt. Hence, irreversibility, uncertainty and expandability lead to a zone of inaction. When business conditions become more unfavorable, the value of waiting increases, the size of the inaction zone rises and the amount invested falls as the marginal loss of option value becomes more important.
3.2 Political risk, the value of waiting and the endogenous risk premium
In this section we modify the model of irreversible investment presented in Section 3.1 in order to introduce political risk. Political risk affects investment through its impact on the distributions of the state of demand, productivity and the discount factor. Let 
t = r + 
t where t denotes the risk premium for Quebec due to political risk.16 The firm's stochastic discount factor is βt = (1 + t)–1. The state of demand, the state of productivity and the discount factor have different distributions depending on the regime. Suppose that t, At and t take on values in the sets 
and 
, respectively, and that they follow first-order Markov processes. That is, letting ht (t, At, t), fst(ht + 1|ht) denotes the conditional density of ht, given the regime st at time t. As we explain below, investors face a less favorable distribution of h in the "bad" regime.
In the model with regime shifts, firms face two regimes, the current regime, regime 0, and a less favorable regime, regime 1, a transition to which may occur with positive probability. We define regime 1 as the separation of Quebec from Canada, a regime which is attainable only if the PQ is in power. We also assume that once separation occurs, it cannot be reversed: Quebec could no longer return to the Canadian federation and regime 0 can no longer be achieved; that is, regime 1 is an absorbing state. Given the PQ in power, the regime shift is governed by a two-state Markov chain with time-varying transition probabilities. Whether a regime shift occurs in this setting depends on the political process, that is, it depends on the realization of two distinct events: which political party is elected and if a separatist party is elected, its chances of winning a referendum. To obtain the total probability of a regime shift, we must take both the electoral process and the regime shifts into account, as follows. Denote by 
the total probability of remaining in the current regime at t + 1, and 
the total probability of a regime shift in period t + 1 given that Party a is in power and regime 0 is in effect at t, where 
is the time-to-election for the government in power. Appendix B describes in detail how the regime shift probabilities are calculated.
The state variables consist of Kt, ht,at, 
and the current regime st. The expectation of the future value function depends on which regime is expected to prevail in the future. Furthermore, the probability of elections is incorporated into the transition probabilities using the representation in (B.3) together with the law of motion of the time-to-election 
given in (B.2). Letting primes denote next period's values, the value function for the firm's problem is given by:
|
|
for all values of a. Even if the Liberals have just won the elections today, the probability of a shift to the "bad" regime in the next period is still positive because the total probability of separation takes into account that the PQ may get elected in the future with positive probability. By (B.3), 
. Hence, we find that political risk affects the firm's investment decisions regardless of the party in power.
To give content to our assumption that political risk leads to less favorable distributions for t, At, and t, we will assume that t, At are stochastically smaller and that t is stochastically larger (in the sense of first-order-stochastic dominance, or FSD) in the "bad" regime.17 As firms now assign a positive probability of facing lower states of demand and productivity and higher interest rates than in the current political regime downside risk increases. Let 
be the value of ht such that it is optimal not to invest at t.18 Under irreversibility a shift in the distribution in the range 
and 
induced by political risk will affect the decision as to whether to invest or not as well as the amount invested. This is Bernanke's "bad news principle". As the option value of waiting rises the incidence of a binding irreversibility constraint increases: that is, 
t rises. Furthermore, when it invests, the firm invests a lower amount. Looking at (3.4) we see the impact of political risk through three channels: (1) the loss of option value, CK, (the impact of constraining the firm in the future by adding an additional unit of capital–by investing the firm reduces its option value of waiting) is enhanced, as making a mistake (i.e. underestimating the probability of separation and the extent to which t, At, and t are less favorable) may be very costly. In the event that separation should take place the firm would be stuck with excess capital which it could not sell (if it is firm or industry specific). Even in the case of less industry specific capital, the possibility of resale even at high discount prices would be very limited, as the economy would be severely depressed. (2) Qt would fall as the expected marginal value of existing capital would fall; (3) t, growth options, would also be adversely affected under a scenario of shifting to a "bad" regime.19 Therefore, the marginal costs of investing rise while the marginal benefits fall so that investment diminishes.
Impact of interest rates
One important impact of political risk even in the "good" regime is through interest rates that rise due to a premium. To evaluate theoretically the impact of the increase in interest rates, assume first that t is i.i.d. A shift in the distribution of t in the "good" regime will increase the incidence of a binding irreversibility constraint (t rises). Furthermore, if an interior solution exists at t, we obtain from the first-order condition
|
| (3.5) |
t is serially dependent there is an additional effect of a rise in the current interest rate, namely, the information effect that arises as an increase in t signals a change in future values. Assuming that , A, and are mutually independent, so that f s' (h'|h) = f , s' ('|) f A, s' (A'|) f ,s' ('|), the total effect is
|
| (3.6) |
F, s'/ F, s' (t + 1|t) t is the derivative of the cumulative distribution function of t + 1 with respect to the conditioning variable t. The first term in (3.6) is the negative the cost effect as in (3.5). The second term captures the information effect and will be positive when is positively serially correlated, since F/t 0 by FSD, and VK < 0. While the total effect is ambiguous, the information effect tends to be quantitatively small so that the cost effect dominates and investment is negatively affected. When is negatively serially correlated, the sign of VK is theoretically indeterminate, but f/t 0. In this case, if VK < 0 then the information effect is negative, so that both the cost and the information effect of an increase in the interest rate depress current investment.20
To gain some additional insight, we rewrite (3.2) as follows. Using the envelope theorem,
|
|
at t can be re-arranged as
|
| (3.7) |
and of a time-varying marginal irreversibility risk premium when the firm invests an additional unit, namely,
|
| (3.8) |
As can be seen, the risk premium in interest rates raises the current cost of capital thus lowering investment. In addition, in the i.i.d. case, a shift in the distribution of t in the "good" regime makes it more likely that investment will be constrained in the future so that 
t rises. In turn, a higher irreversibility risk premium lowers investment at time t. Under reasonable assumptions, a similar conclusion will hold when is serially correlated. Thus, the irreversibility risk premium rises with stochastically higher interest rates. In other words, the cost of capital effect and the irreversibility premium effect of higher interest rates both contribute to reducing current investment even in the "good" regime.
|
4 Numerical results |
|---|
|
TopAbstract1 Introduction2 A case study of...3 A theoretical framework4 Numerical results4.1 The long-run impact5 ConclusionsAppendixA Data AppendixB The regime shift...C Parameterizing the modelReferences |
|---|
Table 3 shows the parameter values used in the simulations. As in the Real Business Cycle literature, we use the steady-state properties of the model or the results of existing studies to determine the values of the parameters.21 We use data on the determinants of investment such as tax policy, labor costs, and demand and productivity for Quebec to calibrate the model and consider alternative specifications for the regime shift probabilities.
|
We consider the behavior of a benchmark investment model with one which incorporates some aspect of the experience of political risk in Quebec. Table 4 shows the average investment response denoted E(I), the average investment-capital stock ratio, I/K, and the coefficient of variation (cv) of investment, CV = STD(I)/E(I) based on 1000 simulated histories for the demand and productivity shocks that are of 150 periods length. We assume that the
t and At are lognormally distributed and approximate the continuous distributions with discrete Markov processes in each regime using the procedure in Tauchen (1986). We consider 11 points in the discrete grid for the exogenous shocks. The first part of Table 4 presents the investment response for our baseline specification, the case without regime shifts.22 The second part of Table 4 presents the average investment response when there exists a positive perceived probability of transiting to a less favorable regime by firms. In the simulations reported in Part II, the "bad" regime is assumed to be characterized by distributions for demand and productivity that have level means that are 10 percent lower than in the "good" regime, holding the level variances constant. We draw realizations of the random variable st denoting the current regime by making use of the properties of the electoral process as described in Section 3.2. For the period 1990–1998, we take 
equal to 50 percent. The probability of a regime shift, conditional on a = PQ,L being in power, is given by the relation in (B.3).
|
For the parameterization in part II of Table 4,
varies as 0.5, 0.4090, 0.3768, 0.3306, and 0.3013 as the number of periods until elections falls from 5 to 1. By contrast, 
increases as 0.0, 0.0910, 0.1232, 0.1694, and 0.1987 as lt falls. When the PQ is in power at t, the probability of achieving separation falls as the election date approaches and a change in governing party becomes more probable. The opposite is true when the federalist party is initially in power at t since in this case the probability of achieving separation increases as the election date approaches and a change in governing party becomes more probable. |
4.1 The long-run impact |
|---|
|
TopAbstract1 Introduction2 A case study of...3 A theoretical framework4 Numerical results4.1 The long-run impact5 ConclusionsAppendixA Data AppendixB The regime shift...C Parameterizing the modelReferences |
|---|
Table 4 illustrates the long-run behavior in an economy subject to stationary shocks to demand and productivity and to regime shifts arising from the nature of the party in power. In the steady-state I/K is stochastic in view of serially correlated shocks to demand and productivity.23 Table 4 shows that averaging over all values of the shocks, the average investment-capital stock ratio is approximately equal to the exogenous rate of depreciation for all specifications of the model. Our interest lies in examining the average investment response.
Consider the investment response of firms with regime shifts, which can be computed conditional on remaining in the "good" regime, so that the separation state is not realized in the sample, or by allowing for the separation state to occur in a probabilistic manner during the sample period. We observe that the probability of a regime shift at some future date induces firms to lower their investment today. Table 4 shows even if separation has never been realized in the sample, investment falls by 14.6 percent compared to the case without regime shifts. As in Danthine and Donaldson (1999), our results indicate that the presence of an absorbing state, namely, a state from which the economy cannot transit once it has entered into it, has a significant negative effect on investment even if such a disaster state has never been observed in the sample. Comparing these results with the actual investment response in Quebec, we note that the de-trended levels of investment per worker, fell by 22, 18, and 16 percent for M&E investment in manufacturing, the business sector and total industries during the period of political risk in Quebec relative to the period of political stability in the 1980's. Part (II.ii) of Table 4 also shows that if the separation state were realized during the sample period (in a probabilistic sense), then investment declines by nearly 37.8 percent in the long-run.24 It is in this context that the various subjective indicators of sentiment in favor of separation gain importance: to the extent that investors incorporated such beliefs into their real investment decisions, even a relatively low likelihood of transiting to a less favorable regime implies a significant decline in investment. Our exercise may also be interpreted in the following manner: Given the subjective indicators in favor of separation, by how much do demand or productivity need to fall in the "bad" regime to generate the observed decline in investment in Quebec? Our analysis predicts that when the regime shift probabilities are specified by making use of the nature of the actual electoral process and the opinion poll results, a 10 percent decline in demand and productivity in the "bad" regime is sufficient to lead to a decline in investment that is in line with observed investment shortfall in Quebec.
Our results also show that when the firm can adjust its capital stock optimally over the long-run, a non-zero probability of transiting to the "bad" regime implies that the long-run steady state capital stock is lower. This occurs because the investment-capital stock ratio is unchanged but the long-run level of investment has fallen. Abel and Eberly (1999) argue that there are two effects of an increase in risk with irreversibility: the first refers to the short-run or trigger effect while the second refers to the long-run or hangover effect. In the latter case, the inability to disinvest implies that firms may find themselves with too much capital. Our results imply that the trigger effect dominates so that the long-run capital stock with regime shifts is lower relative to the situation in which the current regime, namely, regime 0, is expected to prevail permanently. This occurs even if the "bad" regime has never been observed in the sample.
4.2 The short-run impact
In our previous analysis, we did not distinguish between the party in power at the initial date, namely date 1, because this does not affect the long-run investment response. Furthermore, in our simulations, the party in power switches in a probabilistic sense. However, it may be of interest to examine the investment response conditional on one of the parties remaining in power for a certain period of time. That is, we allow elections but we assume that the same party is reelected for several consecutive elections. In Part II (iii) and (iv) of Table 4, we simulate the response of investment conditional on the PQ or L remaining in power. For simplicity, we use the decisions rules contingent on having 5 periods left in office and consider shocks from the distributions for demand and productivity in the "good" regime. We observe that allowing the PQ to remain in power for the foreseeable future leads to a significant decline in investment as the likelihood of achieving separation increases. Investment is lower by nearly 6.5 percent compared to the situation where the party in power is likely to change, and by 20.12 percent relative to the situation without regime shifts (due to the risk of separation). By contrast, if the federalists were to remain in power for the foreseeable future, the probability of separation diminishes and investment would be nearly 6.6 percent higher compared to the situation with the normal switching of parties during elections and only 8.5 percent lower relative to the situation without regime shifts. Such was the case during 1985–89, which is a period of political quiescence as far as the separation issue is concerned, when investment rates are found to be significantly higher than in the 1990s. Thus, it is of interest to note that our model delivers predictions that can match these observations.
In a final experiment, we model the transition from the period of political stability in the 1980's to the period of political risk in the 1990's. Specifically, taking into account poll data, we simulate the investment response implied by the model by letting 
for the first 60 periods, and letting 
for the remaining periods. In other words, as the poll data indicate, had the PQ succeeded in gaining power during the period of political stability, the probability of their effecting separation would have been significantly less that 50 percent. Thus, we take 
By contrast, during the 1990's, an exogenous political event, namely the breakdown of the Meech Lake Accord in 1990, aroused separatist sentiment as reflected in poll data so that in the event of the PQ winning elections (which it did in 1994) the probability of a regime shift would be 50 percent. Thus, we take 
. The last row of Table 4 shows that the overall decline in investment relative to the situation without regime shifts is 10.5 percent. Since this scenario captures most accurately the situation in Quebec during the period of political risk, we find that what we termed the political risk channel of the risk of separation can lead to a decline in investment that is in line with observed values.
4.3 The risk premium in interest rates
In our framework, political risk affects investment through a perceived probability of a change in regime and also through a higher discount rate for firms’ present value maximization problems. Evidence from a number of studies suggests that political risk had a significant effect on asset returns in Quebec.25 While our results from the previous section imply that the perceived threat of transiting to the separation state could be responsible for the observed decline in investment in Quebec, another possibility is that investment fell due to the increase in the firms's cost-of-capital and irreversibility premium in the "good" regime. To examine this claim, we proxy the firm's required rate of return by a baseline interest rate plus the bond spreads. The mean and SD for the Quebec–Ontario bond spreads are given by 0.2589 and 0.0890 for the period March 1990 to December 1998. However, these bond spreads increased up to 0.3419 in July 1990 after the collapse of the Meech Lake Accord, to 0.4873 in July 1994 preceding the provincial election, and to 0.4562 in February 1995 in the run-up to the 1995 referendum. Hence, we define a 2-state process for the real interest rate which takes on the values r1t = 5.26 percent and r2t = 5.78 percent. We consider only i.i.d. interest rates, and assume that the probabilities for the high and low rates are given by p = 0.5. Under this assumption, the mean of the implied interest rate process is 5.52 percent, which equals the sum of the baseline interest rate plus the mean of the Quebec–Ontario spread, and the population standard deviation is 0.26.26 The first part of Table 5 shows that changes in the discount factor of the magnitude observed in the sample cannot account for the decline in investment. Compared to the baseline model, we note that fluctuations in interest rates of a magnitude observed in the data yield an investment decline of only 4.47 percent. Clearly this is too low to match any of the features of the actual data.
|
4.4 A "sudden stop"
The results of the previous section suggest that the financial market channel alone seems incapable of rationalizing the observed investment response in Quebec. Another possibility is that investors may have expected large increases in market risk premia in the event that separation had occurred, provoked, for example, by the problem of the apportionment of the Canadian federal debt between Canada and Quebec, uncertainty about the future of the Canadian dollar, a potential banking crisis, capital flight and the institution of capital controls by a separate Quebec, and so on.
Here, we consider an alternative version of the political risk channel that operates through large expected increases in the (exogenous) risk premium facing firms in the "bad" regime. Specifically, we assume that rt = 5.26 percent in the "good" regime, but that it fluctuates with equal probabilities between r1t = 5.26 percent r2t = 15 percent in the "bad" regime. As before, we assume that the distributions for demand and productivity have 10 percent lower means in the " bad" regime. Table 5 shows that if investors expect higher interest rates or higher risk premia in the " bad" regime, investment declines by close to 40 percent relative to the situation without regime shifts and by 30 percent relative to the situation without increases in risk premia. Thus, investment falls both because expected future cash flows are lower, and also because investors expect to discount the cash flows at a higher rate. This occurs even if separation has not been observed in the sample.27 If separation does occur in a probabilistic sense, then the decline in investment is 57 percent if no regime shifts are expected. As in Table 4, we can also consider a transition from the period of relative political stability in the 1980's to a situation with political risk in the 1990's. Assuming that the transition is known ex ante, then investment declines around 35 percent if investors expect worse fundamentals for demand, productivity, and real interest rates in the "bad" regime. Combining these results with those reported in Table 4, we conclude that the expectation of higher interest rates or higher risk premia in the "bad" regime is an important channel through which political risk must have made its effects felt.
The sequence of events following the separation decision might appear similar to those in a "sudden stop".28 Evidence from a number of "sudden stop" situations suggests that real interest rates can register dramatic increases over brief periods. As a final exercise, we consider a known transition to the "bad" regime with higher interest rates. Specifically, we assume that the real interest rate is rt = 5.26 percent in the "good" regime, but in the "bad" regime, it rises to r2t = 50 percent during a brief transition of 20 periods, and then randomly fluctuates between the values of rt = 5.26 percent and rt = 15 percent afterwards.29 Figure 4 shows that investment essentially collapses during the transition beginning with period 60 and the long run value of the CV increases by 30 percent even compared to a situation in which the "bad" regime is realized in a probabilistic sense. More interestingly, we observe that a transition of this sort also reduces the investment-capital ratio.30 These results appear to be line with our view of events that would accompany such a transition. Our model predicts that once a financial crisis is precipitated, perhaps as a function of political events, we would tend to observe large declines and high volatility in real investment and real output. Note that the simulations indicate that there is an increase in investment as the economy exits from the acute crisis stage. Many economies do display some increase in investment following a crisis as the economy works back to so-called "normal" capacity.31
|
|
5 Conclusions |
|---|
|
TopAbstract1 Introduction2 A case study of...3 A theoretical framework4 Numerical results4.1 The long-run impact5 ConclusionsAppendixA Data AppendixB The regime shift...C Parameterizing the modelReferences |
|---|
In this article, we have studied the impact of political risk on real investment decisions. There is a large literature that studies the negative effect of political risk on investment behavior. Much of this literature lacks an explicit theoretical focus, and uses regression-based or reduced-form techniques to uncover the importance of political risk on investment behavior.32 Unlike much of this literature, we have used a fully articulated model of firms’ investment decisions that incorporates a perceived probability of transiting to a regime with worse fundamentals. Allowing for actual features of the electoral process that govern the regime switch, we have examined the role of expectations regarding demand, productivity, or interest rates in a future unobserved regime in determining current investment outcomes. Our analysis is in the spirit of a "reverse calibration" exercise. We find that when the regime shift probabilities are specified by making use of the nature of the actual electoral process and the results of opinion poll data, a 10 percent decline in demand and productivity in the "bad" regime is sufficient to lead to a decline in investment that is in line with observed investment shortfall in Quebec.33 Allowing for increases in risk premia in the "bad" regime causes larger potential declines in investment today. In the absence of expectational effects, we find that the investment shortfall due to increases in interest rates in the current regime, for example, are too small to rationalize the observed declines in investment in Quebec.
Much of the literature on the determinants of FDI or capital flight has been concerned with identifying the nature of the risk premia due to political risk. In our model of irreversible investment, we have shown that political risk increases downside risk and raises the option value of waiting. Political risk increases the endogenous irreversibility premium and, hence, the marginal cost of investing, while at the same time it leads to a decline in marginal benefits as it reduces the marginal value of capital and growth options. As marginal costs increase and marginal benefits fall, the incidence of the binding irreversibility constraint rises and total investment falls. Using numerical simulations that capture key features of the Quebec economy, we have shown how the irreversibility premium interacts with a perceived probability of transiting to a regime with lower demand or productivity or higher interest rates to yield significant declines in investment that could not be explained by changes in current fundamentals alone. We have also shown how the same mechanisms can lead to "sudden stop"—type phenomena—with sharp and prolonged contractions in investment and output—as an economy experiences a temporary period of high and volatile interest rates in the transition to a regime with permanently worse fundamentals.
While, for simplicity of exposition, we have abstracted from financial market imperfections, in the presence of such imperfections, the impact of a financial crisis on investment will be magnified. Hence our results may be viewed as providing a more conservative estimate of the impact of political risk on investment. To simplify the exposition, we have also assumed that total factor productivity is exogenous. In an earlier version we have explored the implications for TFP or TFP growth to depend on investment. This framework provides an additional channel for political risk to affect growth through lower current productivity or productivity growth.
In this article, we have studied a specific case of political risk. However, our framework could also be used to study the impact of uncertainty about the sustainability of monetary, fiscal and exchange rate policy—all of which are important risk factors facing individual firms. It could also be used to examine the investment response following the adoption of structural adjustment programs when there is uncertainty about the policy-maker's resolve or its ability to implement the reforms.
|
Appendix |
|---|
|
TopAbstract1 Introduction2 A case study of...3 A theoretical framework4 Numerical results4.1 The long-run impact5 ConclusionsAppendixA Data AppendixB The regime shift...C Parameterizing the modelReferences |
|---|
|
A Data Appendix |
|---|
|
TopAbstract1 Introduction2 A case study of...3 A theoretical framework4 Numerical results4.1 The long-run impact5 ConclusionsAppendixA Data AppendixB The regime shift...C Parameterizing the mod |
|---|