Borrowing Constraints and the Marginal Propensity to Consume

Available evidence suggests that the average marginal propensity to consume (MPC) from the 2001 tax rebate in the US was not nearly as large as that from previous tax cuts. We examine if this phenomenon can be explained by the fact that the widespread use of credit cards has made borrowing accessible for most US households by constructing a model that simulates the dynamic effect of relaxed borrowing constraints. Our model uses Kreps-Porteus preferences which account for independent measures of relative risk aversion and the elasticity of intertemporal substitution, both of which can theoretically affect the willingness to save or spend. Our model shows that the average MPC drops substantially immediately after borrowing constraints are relaxed because few consumers have binding borrowing constraints at that time. The model also shows that consumers gradually reduce their wealth after borrowing constraints are relaxed, causing more of them to have binding constraints over time, which in turn causes the average MPC to rise gradually to a new steady state value that is slightly lower than the original value. This dynamic pattern of the MPC suggests that a greater ability to borrow with credit cards could explain the lower effectiveness of the 2001 tax rebate. In addition, the model predicts that consumers choose to hold lower amounts of liquid assets for precautionary reasons when they have a greater ability to borrow unsecured debt. 실증연구들은 2001년 미국의 세금 환급에 서 나타난 평균적인 한계소비성향이 이전의 조세 감축에서 보여진 한계소비성향보다 감소 되었음을 암시하고 있다. 우리는 이와 같은 현상이 신용카드의 광범위한 사용으로 대부분 의 미국 가계에서 차입이 쉬워짐으로써 발생 하였다는 것을 차입제약 완화의 동태적 효과 를 분석하는 시뮬레이션을 통하여 고찰한다. 우리 모형은 위험 기피도와 시점 간 대체탄력 성을 독립적으로 결정하는 Kreps-Porteus 선호를 사용하여 차입제약이 완화된 직후 한 계소비성향이 크게 하락하였음을 보여주고 있다. 우리 모형은 또한 차입제약이 완화된 후, 소비자들이 자산을 감소시켜 완화된 차입 제약의 구속력이 서서히 커져서 한계소비성 향이 이전보다 약간 낮은 새로운 정상상태로 서서히 상승됨을 보여주고 있다. 이러한 한계 소비성향의 동태적 패턴은 신용카드 사용으 로 인한 차입제약 완화가 2001년 세금환급에 서 이전보다 낮은 한계소비성향을 야기하였 음을 의미한다. 그리고 우리의 모형은 차입제 약이 완화되었을 때 소비자들이 예비적 유동 자산을 낮은 수준에서 유지한다는 것을 보여 주고 있다. Borrowing Constraints and the Marginal Propensity to Consume 3


Ⅰ. Introduction
To counteract the prolonged effects of the financial crises in several countries, many governments have tried to cut taxes and to raise their spending. G20 countries in 2009 agreed to stimulus packages worth an average 2% of GDP. The effectiveness of such fiscal stimulus policies depend on the marginal propensity to consume (MPC) from changes in income for the average consumer, but economists continue to debate the empirical value of the average MPC and the effectiveness of fiscal stimulus policies more generally. In particular, some empirical studies have argued that the MPC from income shocks has declined during the last one or two decades by estimating the MPC from the 2001 tax rebate and comparing it to that from previous tax-cuts. For example, Shapiro and Slemrod (1995) found that 43% of surveyed consumers were willing to spend the temporary increase in their take-home income in response to the changes in income tax withholding in 1992, even though the temporary increase in income was likely to be offset by a decrease in a tax refund or an increase in tax payments in 1993. Using a similar survey, Shapiro and Slemrod (2003) reported that only 22% of respondents were willing to spend the initial 2001 tax rebate in 2001. Although the authors use the same survey methodology and similar questionnaires, the differences in the responses are perplexing. 1 The differences in the survey responses indicate that the average MPC has changed. Furthermore, when we use assumptions about the distribution of the MPC across consumers in Shapiro and Slemrod (2002), we are able to calculate that the average MPC has fallen from approximately 0.47 in 1992 to approximately 0.33 in 2001. 2 Other evidence is consistent with the hypothesis that the average MPC has declined over time. A University of Michigan survey, cited in the Christian Science Monitor, reported that only $8.36 billion out of the $38 billion 2001 tax rebate checks was spent. Also, a New York Times/CBS News poll in May 1982 found that approximately 50% of consumers in a survey said that they would spend the increase in take-home income due to the tax cuts proposed by the Reagan administration (Souleles (2002)), while Gallup Poll in July 2001 reported that only 17% of respondents said that they would spend the 2001 tax rebate (Shapiro and Slemrod (2002)).
Other empirical studies also indicate that the recent average MPC is no larger 4 韓國開發硏究 / 2011. Ⅳ than the average MPC from two decades ago. Souleles (2002) estimates the average MPC in response to the 1982 tax cut to lie between 0.662 and 0.998 at a 5% significance level one year after the tax cut was implemented. Johnson, Parker, and Souleles (2006) estimate the average MPC in response to the 2001 tax cut to lie between 0.2 and 0.4 at a 10% significance level for the first three month period when the rebate was received. The authors then show that the estimated overall MPC rose to about 0.66 at a 10% significance level six months after the 2001 rebate was received, and that the MPC thereafter was small and insignificant. Souleles (2002) and other authors have speculated that the apparent differences in the average MPC over time can be explained by a mental accounting hypothesis, where consumers save a large portion of a large lump-sum payment, but spend a large portion of incremental amounts from paychecks. (See Thaler (1990) for a general explanation of this hypothesis.) The foundation for this speculation is that the 1982 tax cuts and the 1992 withholding change were delivered to households through a reduction in taxes withheld from paychecks, while the 2001 tax cuts were delivered by mailing tax rebate checks.
However, we investigate a different explanation for the estimated fall in the average MPC out of temporary income shocks by using the fact that widespread use of credit cards has made borrowing accessible for most US households. The 2001 Survey of Consumer Finances (SCF) reports that 76.2% of the US households have at least one credit card and two thirds of households hold positive amounts of credit card debt (see Aizcorbe, Kennickell and Moore (2003) and Laibson, Repetto and Tobacman (2003)). In addition, credit card debt has grown over 10% per year since the 1970s, implying approximately a 250% growth rate per decade (see Yoo (1998)). Using results from Castronova and Hagstrom (2004) and the 2001 SCF, we show in Table 1 that the ratio between the median total credit limit from credit cards (as a measure of unsecured borrowing potential) per household and the median income per household has risen from approximately 0.3 in 1992 to nearly 0.5 in 1998 and 2001. 3 To analyze the theoretical relationship between the average MPC and borrowing constraints, Carroll and Kimball (1996) compare a model with uncertainty and complete borrowing constraints to one where consumers have perfect foresight and can borrow as much as they like, as is typically assumed in the permanent/life-cycle income hypothesis. The authors show that the introduction of uncertainty and borrowing constraints causes the predicted average MPC to rise relative to the perfect foresight and unconstrained case and that the predicted MPC rises more for consumers with low amounts of liquid wealth than for those with high amounts of liquid wealth. Ludvigson (1999) shows that consumption responds to expected changes in consumer credit and suggests that increases in access to credit may induce less excess sensitivity of consumption to predictable changes in income. Using the constant relative risk aversion (CRRA) utility function, however, Carroll (2001) argues that a high growth rate in income and/or a high rate of discounting enjoyment in the future (or specifically the -impatience‖ of consumers) are the main factors that determine the MPC, rather than borrowing constraints per se. Kimball and Weil (2009) separate the effects of risk aversion and the willingness move resources across time and examine how these two effects determine precautionary saving in a two-period Kreps-Porteus model.
We extend Kimball and Weil's (2009) analysis by using a specific form of Kreps-Porteus preferences called Esptein-Zin preferences, which can be used in a multiperiod simulation of how consumers may want to save or to borrow over time. More specifically, this paper also studies the theoretical effect of relaxing borrowing constraints on the average MPC, but it extends previous work in two directions. It uses the Kreps-Porteus preferences 4 in a multi-period model 5 instead of the commonly used CRRA preferences and analyzes the dynamics of the average MPC as well as its steady state predicted values.
Kreps-Porteus preferences are useful because they allow independent representations of risk aversion and intertemporal substitution, both of which may independently affect how much people want to spend and save. Relative risk aversion represents how much people dislike changes in the amount of resources they have over time due to external risks that they have no control over (such as a job loss caused by company wide layoffs). The elasticity of intertemporal substitution represents how willing people are to save and borrow over time (to substitute resources intertemporally) given a change in the relevant interest rate. Furthermore, as borrowing constraints are relaxed, our model shows that the willingness to save for precautionary reasons will decrease more when risk aversion is low, thereby raising the MPC from additional income available today. But as borrowing constraints are relaxed, willingness to borrow future resources may increase more when the elasticity of intertemporal substitution is high, thereby lowering the MPC from additional income available today. Thus, these two effects may offset each other under CRRA preferences due to the inverse relation between relative risk aversion and the elasticity of intertemporal substitution. With Kreps-Porteus preferences, however, we are able to control the size of these two effects independently by identifying separate parameters for the elasticity of intertemporal substitution and relative risk aversion. (See Kreps and Porteus (1978)).
We also consider how the average MPC from temporary income changes and the amount of assets may change over time until these values become stable in a selfdefined steady state, where the mean, median and standard deviation of the distribution of cash-on-hand do not significantly change after many iterations (up to 100 periods). When borrowing constraints are relaxed in the model to simulate greater borrowing capacity through credit cards, the model predicts two effects. First, fewer consumers should have binding borrowing constraints at that time, so that more consumers would be able to smooth intertemporal consumption by saving or borrowing given preferences about intertemporal substitution. Also, after borrowing constraints are relaxed, consumers with a precautionary motive to save 6 韓國開發硏究 / 2011. Ⅳ can afford to reduce the level of precautionary assets given preferences about risk aversion. Thus, we study the dynamics of the average MPC immediately after a change in borrowing conditions and thereafter as consumers adjust their precautionary wealth. We have found that the decline of the average MPC immediately after relaxing borrowing constraints is comparable to the estimated drop in the MPC, while the decline of the MPC at the new steady state can explain 10%-20% of the estimated drop. To the degree that borrowing constraints were relaxed in the US in early 2000s, our model can then partially explain empirical findings from previous studies.
The structure of this paper is as follows. Section 2 presents an optimization problem for a representative consumer with uncertain labor income and with a specified amount of credit card borrowing potential; it then explains how the analysis in this paper is conducted. Section 3 examines simulation results from the initial steady state before borrowing constraints are relaxed to the new steady state after they are relaxed, and then examines the path of the average MPC over time. Section 4 offers concluding remarks.

The Model
Rather than relying on behavioral assumptions, we examine whether a model with forward-looking consumers who respond optimally to changes in credit availability can explain the apparent decline in the average MPC out of temporary income shocks. Formally, we model a representative consumer who is assumed to want to maximize the benefit from consumption resources over time according to the following specification of Kreps-Porteus preferences: (1) subject to: Borrowing Constraints and the Marginal Propensity to Consume 7 where t E denotes the conditional expectation given information at time t, ) 1 , 0 (   is related to the future discount factor in the Epstein-Zin specification of Kreps-Porteus preferences, 6 R denotes the gross interest rate on a single, risk-free asset, t C denotes a composite measure of consumption expenditure at time t, t Y denotes labor income at time t, and t X denotes resources, or "cash-on-hand", available for consumption. t P is the expected long run average or -permanent‖ component of income from labor services, and t  and t N are temporary and permanent changes in labor income, respectively. t  can be interpreted as temporary bonuses, lay-offs or illnesses without sick leave, while t N can be interpreted as promotions or demotions in one's career. G, the gross growth rate of t P , is assumed to be constant and is meant to reflect the long run average growth rate of the macroeconomy and real income. The logarithm of labor income shocks, t  ln and t N ln , are assumed to be independently, identically, and normally distributed with mean zero and variances 2   and 2 N  , respectively. This assumption implies that zero income shocks will not occur with positive probability. Equation (1) shows the time-inseparable Epstein-Zin (1989) specification of Kreps-Porteus preferences, which allows separate parameters for the elasticity of intertemporal substitution and relative risk aversion, both of which may affect the willingness to spend or save. As borrowing constraints are relaxed, the willingness to spend additional income may fall on average because fewer consumers are completely constrained from borrowing. This effect would lower the average MPC from additional income available today as borrowing constraints are relaxed because of a greater ability to borrow from future resources instead. Additional income available today could be saved or used to repay debt, another form of saving, instead of used to increase consumption expenditure. In addition, the relaxation of borrowing constraints weakens the precautionary saving motive, as Carroll and Kimball (2001) show with a CRRA model. Consumers should feel less of a need to maintain assets to protect against unforeseen income shocks when they are able to borrow in the event of unexpectedly low income. Predicted levels of assets therefore fall as borrowing constraints are relaxed as long as consumers are impatient, so that when the optimal consumption function is concave, the slope of the function-the MPC-will rise as wealth falls. The first effect, which we call the intertemporal substitution effect may be influenced by the elasticity of intertemporal substitution if its value affects the willingness to borrow, in particular at low levels of wealth when 8 韓國開發硏究 / 2011. Ⅳ the ability to borrow is more likely constrained. The second effect, which we call the precautionary dissaving effect, increases when consumers have lower risk aversion, since precautionary saving-an example of prudence-is directly related to relative risk aversion.
In a CRRA model, these two effects may interfere with each other, since the elasticity of intertemporal substitution is constrained to be the inverse of relative risk aversion. However, even if these two effects offset each other, it is possible to obtain a clearer prediction of how these two effects change when the elasticity of intertemporal substitution and relative risk aversion independently change in a Kreps-Porteus model.
The parameter  is negatively related to relative risk aversion, which is equal to 1-. The parameter  is directly related to the elasticity of intertemporal substitution, which is equal to 1/(1-). The commonly used CRRA utility function is a special case of Kreps-Porteus preferences when    .
While equations (1) through (4) (and (6)) are conventional, equation (5) is different from previous models in that it allows the representative consumer to borrow up to a constant fraction of permanent labor income ) ( t kP , where 0  k and k is known to consumers. The model simplifies reality by assuming that k (as well as preferences, interest rates and the growth of real income from labor services) is exogenous and constant across time and across consumers, although to make the model more realistic we may want to allow k to depend on the endogenous level of permanent income, which does vary across time and across consumers. A change in the borrowing limits of credit cards or in consumer loan scores would change the borrowing capacity for consumers and is modeled as changes in k. When 0  k , the representative consumer is not allowed to borrow at all, and as k increases, the borrowing constraint (5) is relaxed. Thus, borrowing constraints are modeled as quantity constraints rather than price constraints, i.e., rather than a gap between borrowing and lending rates. Consistent with this model, Jappelli (1990) presents evidence that consumers who are unable to borrow or -discouraged‖ from borrowing from financial institutions frequently are young (without an established credit history) and have low income, two characteristics that can proxy for permanent income.
Equation (6) says that the representative consumer must pay back all debts before he dies. In other words, he cannot declare bankruptcy during his lifetime after borrowing exogenous resources.
Maximization problems like the one above have no analytic solution due to uncertainty in future labor income, and thus require numerical analysis in order to obtain a solution. In the analysis, we take advantage of the recursive nature of the problem and then by normalizing all variables by t P to reduce the number of state variables.7 After normalization, the constrained maximization problem (using a value function that represents the maximized utility function) is written as subject to: Lower-case variables represent upper-case variables normalized by the value of permanent income: We solve the above problem under two sets of "impatience" parameter values, β = 1/1.05, R = 1.02, G = 1.02 and β = 0.9598, R = 1.0344, G = 1.03. The parameter values in the first set are from Ludvigson and Michaelides (2000) and those in the second set are either estimates in Gourinchas and Parker (2002)  In the analysis, we use approximations of the distributions of t  and t N by selecting a finite set of discrete points from the distributions. 9 In order for the finite horizon results of the numerical analysis to converge to the infinite horizon solution, the -impatience condition‖ must hold. Epstein and Zin (1989) outline two impatience conditions for Kreps-Porteus preferences, depending on the elasticity of intertemporal substitution. When the elasticity of intertemporal substitution is greater than 1 (when  is greater than zero), the impatience condition When the elasticity of intertemporal substitution is less than one (when  is less than zero), the impatience condition equals 10 韓國開發硏究 / 2011. Ⅳ where eis equals the elasticity of intertemporal substitution. We focus on (11b) in the Kreps-Porteus simulations because (11a) fails to hold for our specification of the interest rate and the growth rate of income. We do, however, use the case where the elasticity of intertemporal substitution equals one, a knife-edge case between (11a) and (11b). Bishop (2008) derives a non-separable impatience condition that is independent of the value of the elasticity of intertemporal substitution when the growth rate of income is bounded: where δ is a function of the elasticity of intertemporal substitution, relative risk aversion, exogenous variables that define the state of nature for consumers (the level of assets, the expected or average growth rate of income and changes in income), the benefit (utility) that we feel from consuming resources, the interest rate that we can earn from saving resources or the interest rate that we must pay when borrowing resources, and β, which represents how we feel about the trade-off between consuming resources today versus in the future. 10 An interesting characteristic of δ is that it depends on the state of nature and it approaches zero when consumers are borrowing constrained, making (12) easy to satisfy for a finite interest rate. In simulation results provided below, we satisfy (11b) and (12) to be sure that the optimal consumption functions converge after several iterations.
The impatience conditions are labeled such because they are satisfied when consumers are impatient (i.e.,  is low); but they can also be satisfied when the growth rate of income is high, making the future income look high relative to present income. Although the effect of the interest rate is ambiguous in (11b), when the return on saving is low in (11a), present consumption looks attractive relative to future consumption, and consumers become more impatient. Regardless of the underlying cause, when these conditions are satisfied and borrowing constraints are relaxed, the representative consumer is willing to borrow from future resources rather than save for the future.

Simulation Methodology
In the simulations, we focus on an increase in k from 0.3 to 0.5 to reflect the median total credit limit relative to the median income per household from 1992 to 2001, as shown in Table 1. 11 We compute the optimal consumption functions for both values of k using backward induction, starting from a terminal period T = 100, sufficiently long to 10 Specifically, Bishop (2008) solves for δ as β[Et[vt+1(RSt/GNt+1 + εt+1) α ]] (ρ-α)/α • Et[vt+1(RSt/GNt+1 + εt+1) α-1 • vt 1-ρ ] where St is the value of saved assets. 11 The dynamic pattern of the MPC is robust for other values of k. The complete results when k = 0, 0. 1, 0.2, 0.3, 0.5, and 0.6 for the highest level of normalized cash-on-hand, where t represents an arbitrary period. Figure 1 represents changes in the optimal consumption function as borrowing constraints change under the second set of impatience parameter values. 12 As k increases, the consumption function shifts up and to the left, implying less saving or greater borrowing to achieve a fixed level of consumption, or more intuitively, greater consumption at a fixed level of resources. This finding suggests that even consumers who are not currently constrained also increase spending as borrowing capacity increases, a claim consistent with Gross and Souleles (2002). Those authors find that consumers who are not currently borrowing constrained increase their consumption slightly in response to an increase in borrowing capacity from credit 12 韓國開發硏究 / 2011. Ⅳ cards. They argue that consumers with non-binding constraints raise their consumption due to a weakened precautionary saving motive after borrowing constraints are relaxed.
We use these converged consumption functions, which approximate the optimal consumption function for an infinite horizon problem, to define the MPC out of temporary income shocks and evaluate its properties as the borrowing parameter k and the parameters governing the elasticity of intertemporal substitution and relative risk aversion change. We calculate the MPC for each consumer from the following equation: Thus, equation (13) measures the MPC from -temporary‖ income changes, such as those from the tax code change in 1991 or the tax rebate in 2001. We calculate the average of (13) across consumers with different levels of cash-on-hand.
Using the optimal consumption function, we simulate the behavior of 4,000 consumers to examine the dynamic pattern of the MPC out of temporary income changes when the entire population experiences the same change in k but each individual experiences different simulated exogenous income changes. Income shocks are randomly drawn from log normal distributions based on empirical evidence from the Panel Study of Income Dynamics (See Carroll (1992)). To simplify the analysis, we assume that consumers do not start their working lives with any saved assets. 13 Given this zero initial endowment and beliefs about expected future income, each consumer is programmed to optimally decide how much to consume and save each period over his lifetime. Given this behavior, the population of consumers generates a simulated distribution of normalized assets and normalized cash-on-hand that achieves a stable mean, median and standard deviation over time when the impatience conditions are satisfied.
After calculating the average MPC from the stable distribution of normalized cash-on-hand for k = 0.3, we increase k to 0.5, compute the corresponding optimal consumption function and then recalculate the average MPC from the distribution of normalized cash-on-hand during each period after the change in borrowing capacity until a new steady state is reached.

Ⅲ. Simulation Results
In this section, we present the dynamic pattern of the average MPC from the simulation to understand whether greater borrowing capacity could reduce the effectiveness of tax rebates in stimulating consumption expenditure.  (2002), the average MPC is estimated to have fallen from 0.47 in 1992 to 0.33 in 2001. When the risk aversion coefficient (rra) is one (two or three) and the elasticity of intertemporal substitution (eis) is one (1/2 or 1/3), the results in Figures 2 and 3 can be interpreted as those under ordinary CRRA preferences with the CRRA coefficient being one (two or three). Because we can put independent values for rra and eis, we can better predict the effects of changes in the average MPC with the Kreps-Porteus preferences. Figures 2 and 3 show that when the borrowing constrains are relaxed from k = 0.3 to k = 0.5, the average MPC drops immediately and substantially because consumers have a weakened precautionary saving motive and therefore are predicted to reduce their assets. During this time, few consumers have binding borrowing constraints unless they suffer from an enormous negative income shock. As a result, additional temporary income from a tax rebate or tax cut is predicted to be mostly saved for future consumption so that the average MPC in the current period is predicted to be low.
However, consumers are predicted to reduce their assets in response to the loosening borrowing constraints when the impatience conditions are satisfied. As assets are spent, more consumers will face binding borrowing constraints even with greater borrowing capacity, so that the average MPC is predicted to rise gradually to a new stable value. Figure 4 shows the dynamic pattern of the fraction of the population with binding constraints, which corresponds to that of the average MPC: after falling substantially when borrowing constraints are relaxed, it subsequently rises gradually. The average cash-on-hand declines gradually as the weakened precautionary saving motive causes consumers to reduce their assets. Both a growing fraction of consumers who are borrowing constrained and lower cash-onhand cause the average MPC to rise after its initial drop. 14 Figures 2 and 3 also show that the magnitude of the initial decrease in the predicted average MPC from the simulation approximately equals the estimated decrease in the average MPC for some of the specified parameter values. For example, the magnitude of the decrease in the average MPC is slightly larger than the estimated decrease in the average MPC when rra = 1 and eis = 0.33 with the first set of impatience parameters.
After the initial drop, the average MPC gradually increases to a stable value, which we call a steady state value. Table 2

[Figure 4] Dynamic Pattern of the Binding Fraction and the Cash-on-Hand during the Transition
Notes: ‗eis' denotes the elasticity of intertemporal substitution and ‗rra' denotes the risk aversion coefficient in Kreps-Porteus preferences. than that with k = 0.3, the magnitude of the decline is much smaller than the estimated decline from the survey evidence. The magnitude of the decline in the average MPC at the steady state is between 0.013 and 0.030 depending on parameter values, which represents approximately 10% to 20% of the estimated decline. We conclude that as borrowing capacity increases the intertemporal substitution effect immediately lowers the average MPC from additional income available today, but this effect is substantially offset over time by the precautionary dissaving effect which raises the average MPC from additional income available today. While the initial decrease in the MPC is comparable with the estimated decrease in the MPC (from 0.47 to 0.33), the decrease in the MPC between two steady states is much smaller. That is, when borrowing constraints are initially relaxed, the ability to borrow at the level of currently available resources reduces the willingness to spend additional income. But over time, liquid wealth is predicted to be reduced due to impatience and the precautionary dissaving effect, so that the willingness to spend additional income increases relative to the initial decrease.
In addition, Figure 4 shows that consumers will hold lower amounts of cash-onhand when they have a greater ability to borrow unsecured debt due to the precautionary dissaving effect. This result is consistent with that of Bishop (2008).
Thus, to the degree that credit became easier to obtain in the early 2000s in the US through credit cards and other forms of unsecured credit, our model predicts that the MPC will initially decrease by a magnitude that is similar to the estimated decrease, although as consumers are predicted adjust their savings over time in the simulated model, the magnitude of the simulated decrease decreases to about 10%-20% of the estimated value from previous studies.

Ⅵ. Conclusion
This study examines whether a model with relaxed borrowing constraints can replicate the apparent decline in the average MPC in the US. We theoretically analyze the dynamics of the average MPC by using a model with Kreps-Porteus preferences, which accounts for independent measures of preferences about risk aversion and intertemporal substitution. If the widespread use of credit cards indicates that consumers are less borrowing constrained than they were a generation ago, they should be better able to maintain a steady level of consumption expenditure over time (which we call the intertemporal substitution effect) and should have less incentive to save (which we call the precautionary dissaving effect). The model shows that the first effect is dominant immediately after borrowing constraints are relaxed and that the first effect is offset by the second effect gradually as consumers reduce their assets. The model likewise predicts that consumers are substantially less responsive to changes in income immediately after they acquire a greater borrowing capacity, although this effect diminishes over time as they reduce their assets. The results of this paper therefore suggest that temporary tax cuts or tax rebates will be less effective in boosting consumption expenditure when consumers have a greater borrowing capacity.
In addition to credit card borrowing, other types of household borrowing capacity may have also grown during the past few decades. Securitization and decreasing interest rates may have given consumers a greater ability and incentive to use home-secured debt for greater consumption expenditure (See Aizcorbe, Kennickell, and Moore (2003)). Future work could enhance the model in this paper by including durable goods or illiquid assets that allow consumers to borrow more secured debt.

Reducing the Number of State Variables in the Maximization Problem
The value function is multiplicatively separable with respect to expected long run average income, so it can be normalized by this value to reduce the number of state variables from two to one. In effect, randomness in nature for consumers is reduced from two dimensions to one dimension, and consumers are modeled to have a lifetime perspective and care only about available resources relative to their expected long run average income, rather than available resources and expected long run average income separately.
Consider the value function in the second to the last period: For the second to the last period, Using this solution, we can find 韓國開發硏究 / 2011. Ⅳ and so on. In general, Since the terminal period was chosen arbitrarily, we can find the solution to the original problem in each period if we solve the normalized one-state problem The problem for a representative consumer whose preferences are represented by the Epstein-Zin function therefore becomes where the lower case variables represent upper case variables normalized by t P .
Necessary conditions for maximization To derive the first order condition of this normalized problem, assume that we have already found the optimal value function, v(x), for all periods except for t and t + 1. Perturb the allocation of resources in t and t + 1 such that the derivative of unmaximized normalized utility (ut) with respect to the choice variable is equal to zero: by perturbation in only periods t and t + 1.
so that the first order condition is rewritten as have inverse powers and if α = ρ, the first order condition results in the expected utility first order condition. Note also that under certainty (when risk is irrelevant), the expected utility first order condition results and the value function is a transformation of the expected utility function.

Backward Induction in the Maximization Problem
In the last period of life, it is optimal to consume all resources if there is no bequest motive. (If there is a bequest motive, then one could specify that it is optimal to leave a fixed amount at the end of life.) Since the future value function is modeled to be zero after death, the value function in the last period is defined as . Given this extra constraint, we can iteratively solve for the optimal level of consumption expenditure in each period. Beginning with period T -1, we specify a level of the normalized state variable and find consumption as a function of 1  T x from the first order condition: If the representative consumer is not borrowing constrained, the first order condition holds exactly. If he is constrained, we set To find the optimal level of consumption, T T x c , we plug each solution to the above equation to find the one that maximizes However, because the marginal utility of consumption and the marginal utility of saving are convex over the entire domain, there is a unique solution to the first order condition t  . After finding this unique optimal level of consumption expenditure in T -1, we use the value to solve the first order condition in period T -2. Thus, we proceed iteratively until we reach the beginning of the representative consumer's working life. 24 韓國開發硏究 / 2011. Ⅳ

Discretizing the Distribution of Error Terms in the Maximization Problem
To approximate the integrals in the expected value operator, we construct a discrete approximation based on a one dimensional Gauss-Hermite quadature given that ln Nt+1 and ln εt+1 are normally distributed and independent. We use 10 discrete points from a one dimensional quadature for both ln Nt+1 and ln εt+1. Because multidimensional quadatures are extremely difficult to calculate, nonindependent errors would be difficult to approximate. Expectations are modeled to be a function of probabilities of changes in the normalized state variable, conditional on information available today. From the independent normal distributions of ln Nt+1 and ln εt+1, we select 10 Gauss-Hermite discrete points from each distribution and weight them accordingly. (See Secrest and Stroud (1966) for the points and weights for a one dimensional Guass-Hermite quadrature.) The approximation of expected value (utility) of random cash on hand using this procedure is based on the following: where CDF( ) is the cumulative distribution function of ( ), pdf( ) is the probability distribution function of ( ), and d represents the total derivative.
Thus, we can approximate the first order condition as: be the solution to this equation (and the approximate solution to the first order condition). The value function can be solved, given a level of liquid wealth t x for each solution ) ( t t x c and given that the problem has be solved for period t+1. 韓國開發硏究 ABSTRACT This paper investigates the extent of global and regional integration in East Asia using stock price index as a measure of economic performance. We employ a structural VAR model to separate the underlying shocks into "global", "regional" and "country-specific" shocks. The estimation results show that country-specific shocks still play a dominant role in East Asia although their role appears to have declined over time, especially after the 1997 financial crisis. Global and regional shocks are responsible for small but increasing shares of stock price fluctuations in all countries. The results indicate that the stock markets in East Asia remain dissimilar and are subject to asymmetric shocks in comparison to European countries.

Ⅰ. Introduction
The 1997-8 financial crises in East Asia have had far-reaching repercussions in the real economy, policy making, and academia. Within the region, countries have started showing a strong interest in the search of an exchange rate regime that would be more robust to financial crises. At the same time, East Asian nations have been working in earnest for regional economic integration in the past decades. To enhance the financing facilities in the aftermath of the financial crisis, the Chiang Mai Initiative was launched by 10 member countries of the Association of Southeast Asian Nations (ASEAN) plus China, Japan, and Korea (ASEAN+3) in May 2000. 1 In order to facilitate the channel for better utilization of Asian savings for Asian investments and enhance efficiency and liquidity in bond markets in Asia, a local currency-denominated bond market under the Asian Bond Markets Initiative has been developed. 2 In 2005, the East Asia Summit was established by ASEAN+3 plus Australia, New Zealand and India, for the total of 16 countries.
As an important element of financial integration, East Asian countries have been seeking the feasibility of an economic and monetary union. Key policymakers are increasingly vocal about the need to establish a monetary union in the region or create a single currency. Earlier attempt by Japan to create a monetary union died quickly due to strong oppositions from the IMF and the US Treasury. Inspired by the European Currency Unit, now replaced by the Euro, the Asian Development Bank has proposed the Asian Currency Unit (ACU) -a weighted index of currencies for ASEAN+3. Despite numerous technical and political obstacles, the ACU has been moving forward from an academic exercise to a real outcome, one that can be used in the market amid a growing consensus among academic and policy practitioners that intraregional exchange rate stability is desirable for East Asia and a monetary union is the ultimate form to ensure it. 3 One natural question is whether East Asian countries are well integrated financially in a global sense. Are they also regionally well integrated as they have tried to achieve in the aftermath of financial crises? These are important questions since regional integration may reduce the cost of forming a currency union or some form of commoncurrency pegging within the region. Individual member countries will lose the ability to independently use monetary and exchange rate policy when they form a currency 30 韓國開發硏究 / 2011. Ⅳ union, which may entail severe costs if they are subject to dissimilar macroeconomic shocks and go through different business cycles. If countries within the region are similar to each other, the cost from losing the independent monetary and exchange rate policy would be lower. Trade and financial integration of an individual economy with the region is likely to reduce the cost of such a common currency arrangement to the extent that it makes the economy more similar to that of the region.
It is well known that East Asian economies are well integrated in terms of intraregional trade. For instance, Bayoumi and Eichengreen (1994) suggest that, in terms of trade integration, East Asia can qualify optimum currency area (OCA) criteria as well as European countries. Evidence on financial market integration, however, is much less clear. The majority of studies claim that the degree of financial market linkage in East Asia still remains low compared to Europe. Using data on cross-border bilateral holdings of financial assets and liabilities, real interest rate differentials, and consumption risk sharing, Jeon et al (2005) show that East Asian economies became more financial integrated in the post-crisis period. The development is more in the direction of global integration than in regional integration. With similar and additional data such as equity portfolios, debt securities, and bank claims, Kim et al (2008) reach a similar conclusion that East Asian countries are financially less integrated in general than European countries. They also estimate the degree of consumption risk sharing in East Asia by regression analysis and tend to be relatively more linked to the global markets than integrated with one another regionally, particularly compared to Europe.
The purpose of this paper is to assess the extent of financial integration within East Asia and study whether countries in the region satisfy the conditions for an OCA. We employ the overall stock price index as an indicator of macroeconomic performance as well as the development of financial market in each country. The availability of high-frequency data is also a big advantage in our case where the sample period is short due to general data problems of developing countries and made even shorter as a result of the recent financial crisis and resulting structural breaks.
We use a structural vector autoregressive (VAR) method to investigate the extent of financial market integration in East Asia. Returns to investors in each country"s market are affected by three types of underlying shocks: country-specific shocks, regional shocks and global shocks. These structural shocks are identified by long-run restrictions developed by Blanchard and Quah (1979). To investigate the progress in financial integration, we also separate the sample into 8 non-overlapping 2-year subperiods before and after the crisis. We then compare the East Asian region with that of 15 European countries. The experiences of the Economic and Monetary Union (EMU) provide a natural benchmark as the member countries have followed the rigorous process of regional integration in trade and finance and successfully formed a monetary union.
The empirical results show that, in all East Asian stock markets, country-specific shocks are dominant although they became less important in the post-crisis period than in the pre-crisis period. 4 Regional shocks play a minimal role in most cases while the importance of global shocks varies across countries depending on the extent of financial openness and development. In European countries, in marked contrast, external shocks that combine both global and regional shocks appear to take over the dominant position. This suggests that, despite years of efforts toward financial liberalization and cooperation in the region, the East Asian economies are subject to asymmetric shocks and far less integrated financially compared to the European countries. The region seems sufficiently unique perhaps due to different resource endowments, growth experience or economic policies although the efforts for financial integration in the post-crisis period appear to have some effects on the economic and financial structure in the region. Theory of optimum currency area would predict that pegging to the same currency would be more costly in East Asia than it would be in European countries.
The rest of the paper is organized as follows. Section II reviews the current status of trade and financial integration in East Asia. Section III illustrates the data and methodology used in our empirical analyses. Section IV examines the degrees and patterns of regional shocks and country-specific shocks on domestic stock market by using forecast error variance decomposition. Section V investigates the robustness of the benchmark model. Section VI provides concluding remarks.

Trade Integration in East Asia
The extent of regional integration through trade in East Asia has been rising fast over the last twenty years. Wyplosz (2001) uses a gravity approach to determine a "normal level" of bilateral trade among Asian and European economies and finds that East Asia is more, while Europe is less, integrated than one would expect. According to the theory of OCA, a high degree of intraregional trade can increase the efficiency gain of using a common currency while lowering the cost of losing monetary policy autonomy. There is some evidence that joining a currency union can increase trade among member countries, which will further strengthen the case for the formation of the currency union. 5 Table 1 summarizes the changes in the share of intraregional trade for various regions in the world over the period of 1980-2006. 6 For comparison, the fourth panel of the table lists the trade pattern for the Euro area within the region and with the periods. There are substantial increases in the role of global shocks in Korea, Hong Kong, Singapore, Malaysia, Indonesia, and Australia while little changes in China. 5 Rose (2000) reports that bilateral trade between countries that use the same currency is over 200 percent larger than otherwise, controlling for other effects. Lee and Barro (2007) find that a currency union can generate welfare gains from the additional trade with countries belonging to the same currency union, which in turn stimulates an increase in consumption growth rates. 6 In the paper, the intra-regional trade ratio is defined as exports or imports within the region as a share of total exports or imports with the world. Export Import Export Import Export Import Export Import Export Import outside world. It shows that the intraregional trade in the Euro area is stable and maintained at around 65 percent.
The first panel reports trade patterns in the ASEAN. Intraregional trade within the ASEAN increased steadily since 1980 except a slight downturn in exports after 1995, perhaps reflecting the recessionary consequences of the financial crisis that hit the region. The roles of the United States and Japan are still dominant but have declined over the whole period. In addition, there is a significant increase in intraregional trade ratio in a broader region. By adding China, Hong Kong, Japan and Korea to the region in the second panel, we find that nearly half of international trade of the region is with regional partners in 2006. The United States is still the largest importer in East Asian trade, but it is no longer the largest exporter. Trade with the Euro area increased early, peaked in 1990 at 17.9 percent for exports and 15.1 percent for imports. Since then, the trade preference of East Asia with Euro area seems to have declined. Exports to the Euro area dropped to 15.4 percent and imports from Euro area dropped to 10.1 percent in 2006.
In the third panel, Australia and New Zealand are added to East Asia. The intraregional exports and imports have risen dramatically from the 1980s through the 2000s. For instance, in 1980, 37 percent of total import and export were with the regional trading partners. By 2006, the figures rose to 48.6 percent and 51.9 percent, respectively. The table demonstrates, however, that the intraregional trade ratios among East Asian economies are still lower than those of the Euro area by more than 10 percent in 2006.

Financial integration in East Asia -the Chiang Mai Initiative
Before the Asian financial crisis broke out in 1997, few would have seriously argued for the creation of a new regional financial cooperation system. Economic integration in the region had been mostly a market-led process. One of the most noteworthy outcomes of the financial crisis would be the initiation of regional financial cooperation by the East Asian economies. The financial crisis gave East Asia a strong impetus to search for a regional mechanism that could forestall future crisis. Japanese financial authorities proposed the creation of an Asian Monetary Fund (AMF) as a framework for promoting financial cooperation and policy coordination in the region at the G7-IMF meetings in Hong Kong during September 20- 25, 1997. 7 The United States, European Union and the IMF opposed the proposition on grounds of moral hazard and duplication. In November 1997 the East Asian economies, together with the United States, Canada, Australia and New Zealand, agreed to establish the Manila Framework Group in order to develop a concerted approach to restoring financial stability in the East Asia. The Manila Framework took an initiative to create a mechanism for regional surveillance complimentary to the global surveillance by the IMF. 8 In October 1998, Japan pledged $30 billion to support the economic recovery of the crisis-affected countries. The initiative provided major assistance for restructuring corporate debt, reforming financial sectors, strengthening social safety 7 The intrepid proposal for a regional alternative to the International Monetary Fund (IMF) seemed to arise without warning and at the worst possible moment. Both the Philippines and Indonesia had floated their currencies and the Asian Financial Crisis was increasingly showing signs of contagion at the time. The proposal raised temporary hopes among the crisis-ridden economies of Asia but elicited a stringent rebuke from the IMF and the US Treasury and ultimately fell to the wayside in favor of a more IMFcentered approach. See Phillip (2003). 8 Manila Framework terminated its function in November 2004 after 12 meetings. The failure of the Manila Framework is said to be attributable to the lack of mutual trust and lack of a professional secretariat. 34 韓國開發硏究 / 2011. Ⅳ nets, generating employment and addressing the credit crunch. The initiative was called "New Miyazawa Initiative" and was highly successful. 9 In November 1998, the United States and Japan jointly announced the Asia Growth and Recovery Initiative (AGRI), which was a multilateral effort to stimulate economic growth in Asia. With support from the World Bank and the Asian Development Bank (ADB), AGRI supported corporate restructuring and restored market to private capital. It also strengthened bond guarantee functions of the World Bank and the ADB.
The idea of an AMF was revived when the finance ministers of China, Japan and South Korea, along with the ten ASEAN members, agreed on May 6th, 2000 in Chiang Mai, Thailand to establish a system of swap arrangements within the group. The regional scheme for financial cooperation known as the Chiang Mai Initiative (CMI) has been gathering momentum and opening the doors to possibly significant policy-led integration in East Asia. The CMI has two components: expanded ASEAN Swap Arrangements (ASA) encompassing the ten ASEAN countries; and a network of Bilateral Swap Arrangements (BSA) repurchasing arrangements basically encompassing the thirteen ASEAN + 3 countries.
At present, the total amount of BSAs covering all 13 countries is estimated to be around $83 billion. 10 The maximum amount that any individual country can draw varies a great deal. For instance, the maximum liquidity through the CMI to Thailand is about $12 billion while the BSA to Malaysia is $6.5 billion. Doubts have been raised as to whether the BSA system could truly be a credible and effective system of defense against future speculative attacks. The success of the CMI will depend on whether the surveillance system in East Asia can work as effectively as expected. A mechanism that enforces exchange of information and applies peer review and pressure through policy coordination is the right approach to boost the confidence of the countries in the region. It is expected that East Asia will reach deeper monetary and economic integration with gradual development of the CMI to a more effective and efficient regional arrangement. 9 The Japanese Ministry of Finance and the Ministry of Finance of Malaysia have reached an agreement regarding the basic features of the short-term financing facility under the framework of the "New Miyazawa Initiative". The facility is aimed at supporting credit-extending schemes which intend to promote economic activities in Malaysia, such as a trade financing facility, small and medium size enterprise credit line, etc. This will serve as a standby facility for the Malaysian Government should the need arise. In this short-term facility, the Japanese Ministry of Finance is committed to providing up to US$ 2.5 billion liquidity to Bank Negara Malaysia, if and when necessary, through swap transactions between the US dollar and the Ringgit.
10 Japan concluded six agreements with China, South Korea, Thailand, the Philippines, Indonesia and Malaysia: two-way arrangement with China, Korea, Thailand, and the Philippines and one-way arrangement with Indonesia and Malaysia. Korea concluded four agreements in addition to Japan-Korean BSA. China concluded four agreements in addition to its agreements with Japan and Korea except with Singapore. See Table 1 for details. Figure is from Ministry of Finance, Japan.

Financial integration in East Asia -Asian Bond Market Initiative
Due to the underdevelopment of capital markets, countries in East Asia have depended on short-term foreign currency-denominated financing. This causes "maturity" and "currency" mismatches which make the region vulnerable to volatility in short-term capital movements. The East Asian financial crisis vividly illustrates the risks of the double mismatches. It has been agreed that developing bond markets in the region would be effective in regional financing as wellfunctioning bond markets set the benchmark interest rates for all debts with varying maturities and risks and thereby promote efficient uses of resources for economic growth. The Asian Bond Market Initiative (ABMI) aims to develop efficient and liquid bond markets in East Asia, enabling better utilization of regional savings for investment within the region. 11 Its activities focus on the following two areas: (1) facilitating access to the market through a wider variety of issuer and types of bonds, and (2) enhancing market infrastructure to foster bond markets in Asia. 12 Asian governments, central banks and the Asian Development Bank are keen to see the expansion of Asian bond markets in order to help provide finance for the large infrastructural development that the region needs over the next decade. Alongside the expansion of the bond markets, Asian governments and central banks are currently discussing the creation of an ACU. The ADB has suggested that bonds may also be issued in ACU over the next few years which would help lower the financing costs for Asian issuers who have substantial trade links with other countries in the region. 13

Stock Markets in East Asia
Stock exchanges in Asia developed much later than those in Europe or America. The first Asian market for securities trading was in Shanghai which began in the late 1860s. The first share list appeared in June 1866. The Bombay Stock Exchange, launched in 1875, was the oldest organized market in the region, followed by the Tokyo Stock Exchange (TSE) three years later. In 1891 during the boom in mining 11 At the 6th ASEAN+3 Finance Ministers" Meeting in August 2003 at Manila, the Philippines, finance ministers agreed to promote Asian bond markets. 12 A robust primary and secondary bond market in Asia requires a wide variety of issuers and products that could be addressed by encouraging: (1) Sovereign bond issuance by Asian governments to establish benchmarks; (2) Asian government financial institutions to issue bonds in Asia to meet their financing requirements; (3) The creation of asset-backed securities markets, including collateralized debt obligations (CDOs); (4) Bond issuance in the region by multilateral development banks and government agencies; (5) Bond issuance in the region for funding foreign direct investment in Asian countries; and (6) The expansion of local currency-denominations of bonds and the introduction of currency-basket bonds.
13 At the ASEAN+3 Finance Ministers" Meeting (AFMM+3) on August 7, 2003, six voluntary working group (WG) on the ABMI have been established to address key areas of bond market development. Since the establishment of the six WGs, comprehensive efforts have been made to develop regional bond markets. 韓國開發硏究 / 2011. Ⅳ shares, foreign businessmen founded the "Shanghai Sharebrokers' Association" headquartered in Shanghai as China's first stock exchange.

36
Off to a late start amid dramatic historic events, Asian stock markets were quick to adopt cutting-edge strategies and have experienced rapid growth. They espoused technology, demutualized and listed their own shares long before U.S. markets did. The TSE is the second stock exchange in the world by market value, second only to the New York Stock Exchange. It currently lists 2,271 domestic companies and 31 foreign companies, with a total market capitalization of over 5 trillion dollars. The TSE was established in 1943, the exchange was combined with ten other stock exchanges in major Japanese cities to form a single exchange.
The Shanghai Stock Exchange was reestablished on November 26, 1990. A market capitalization of nearly $2.38 trillion makes it the fifth largest in the world. There are two types of stocks being issued in the Shanghai Stock Exchange: "A" shares and "B" shares. A shares are priced in the local Renminbi yuan currency, while B shares are quoted in U.S. dollars. Initially, trading in A shares is restricted to domestic investors only while B shares are available to both domestic (since 2001) and foreign investors. However, after reforms were implemented in December 2002, foreign investors are now allowed to trade in A shares with some restrictions under the Qualified Foreign Institutional Investor system and there is a plan to eventually merge the two types of shares.
Development of the stock markets in East Asia has been accelerated in the aftermath of the 1997 crisis. Willingly and also upon the International Monetary Fund and other external pressures, East Asian countries have become far more open. Table 2 shows the magnitude of market capitalization in our sample as of the end of 2011. China became the second largest stock market, surpassing Japan. The total size of the 12 East Asian stock markets included in the sample is very close to that of the United States, far exceeding that of the European Union.
In this paper we employ and focus on the overall stock price index as indicator of the overall performance of the economy. It is well known that stock prices are a good leading indicator of economic activity. Traditional models suggest that the price of a firm"s stock equals the expected present value of the firm"s future payouts or dividends. As long as these expectations reflect the underlying fundamental factors, they must ultimately reflect real economic activity. 14 14 Fama (1990) showed that stock returns are actually significant in explaining future real activity for the whole period from 1953 to 1987 in the United States stock market. Quarterly and annual stock returns are highly correlated with future production growth rates. According to the reported regressions past stock returns are significant in explaining current production growth rates and vice versa. Merton (1984) found that movements in the United States stock prices were positively correlated with real GNP. Schwert (1990) showed that Fama"s results could be replicated by using data that goes back as far as to 1889. He finds the correlation between future production growth rates and current stock returns to be robust for the whole period from 1889 to 1988. However, Binswanger (2000) concluded that traditional links between stock market performance and two major macroeconomic indicators, production and GDP, broke down in the most recent United States bull market. Although the regressions of stock returns on measures of real activity in the United States over the period from 1953 to 1997 seem to confirm the findings of Fama (1990), stocks returns do not reflect real activity in the current stock market boom from 1984 to 1997. In recent

Ⅲ. Data and Methodology
According to the theory of optimum currency areas (OCA), joining a single currency area brings in costs and benefits. The benefits include reductions in uncertainty and transactions costs that can arise under floating exchange rates. The costs are due to the inability to use monetary and exchange rate policy for economic stabilization. The magnitude of the costs is expected to be lower if business cycles in the member countries are closely correlated and their economic structures are similar.
The OCA criteria have been operationalized and quantified in a number of studies. Bayoumi and Eichengreen (1993), in a well-known study, examine the correlation of aggregate supply shocks to investigate the similarity of economic structure across potential member countries. Their assumption is that aggregate demand shocks are regime-specific while aggregate supply shocks are likely to be invariant with respect to changes in the exchange rate regime. In this study, we separate shocks to the economy into "global", "regional", and "country specific" shocks. The latter will be interchangeably called "domestic" shocks. Global shocks research, Mao (2007) found the links between stock prices and industrial production or GDP remained strong during the high-growth phase since 1980s in the Australian stock market. 38 韓國開發硏究 / 2011. Ⅳ affect economies both inside and outside the regional boundary. Commodity price shocks can be an example of such shocks. Regional shocks are common to the economies within the region. German unification of 1989 and the resulting fiscal expansion may constitute a regional shock for European countries. In East Asia, large fluctuations in the yen-dollar exchange rate seem to have been a common, important regional source of disturbances (Kwan, 1994). Country-specific shocks are unique to a particular economy. They may be either from aggregate demand shocks that are associated with monetary or fiscal policies or supply shocks on productivity or the terms of trade. Regional shocks are expected to be important in a small open economy or in an economy with an economic structure similar to its trading partners or neighbors in the region. External shocks can extend regional boundary. Global shocks affect all countries in the same direction.
Following Chow and Kim (2002), we assume that global, regional and domestic price indices  g t y , r t y , and d t y  are affected by three different types of shocks that arise from the global, regional and the domestic markets and are denoted as g t u , r t u and d t u , respectively. In a matrix form, it can be summarized as follows: For the identification of structural shocks, we employ the following 3 restrictions of the Blanchard-Quah (1989) type based on the assumption that the individual economy is small in the region and, in turn, the region is a small part of the world. 1) Regional shocks have long-run effects on the global index; 2) Country-specific shocks have long-run effects on the global index; 3) Country-specific shocks have no longrun effects on the regional index. We impose these restrictions only in the long-run responses but not on short-run responses.
The identifying assumptions imply that the cumulative effects of a d t u shock on r t y is equal to zero and so are the cumulative effects of the or r t u shocks on g t y .
The assumptions can be restated in terms of impulse responses, . We assume that each structural shock has unit variance and is uncorrelated to other shocks. The importance of regional shocks -which affect countries in the region in a symmetric fashion -is taken as the indicator of similarity of economic structure within the potential member countries since, by construction, they affect each country in the group. On the other hand, the costs associated with a loss of monetary independence and flexible exchange rate adjustments could be heavy if dominant shocks are country-specific shocks and therefore uncorrelated across the region. For global shocks, a global rather than regional arrangement might be a better course of action in dealing with such shocks. In the context of East Asia, for instance, if global shocks (say, affecting U.S. output) are relatively more important than regional ones (say, affecting Japanese output), forming a dollar bloc may be a better policy choice than forming a yen bloc.
The overall stock price index is used as an indicator of macroeconomic performance to identify the three underlying shocks. Stock price data are ideal for our purpose since the availability of high frequency data as a proxy for macroeconomic performance can help us overcome the serious problem of having to work with a short-time span such as the post-crisis period, for which at best 6-7 years of data are available. 15 We employ weekly price data from July 1, 1989 to November 11, 2011 for 12 stock exchanges in East Asia: Japan, China, South Korea, Hong Kong, Singapore, Malaysia, Taiwan, Indonesia, Thailand, the Philippines, Australia and New Zealand. For comparison, the model is first estimated for 16 European countries that consist of 11 EMU countries − Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, Netherlands, Portugal, and Spain − and 5 non-EMU countries − Denmark, Norway, Sweden, Switzerland, and the United Kingdom. 16 The proxy for the global price indexes is obtained from MSCI AC World Price Index. Similarly, we also employ MSCI AC Europe and MSCI AC Asia Pacific as the regional price index for Europe and East Asia.

Ⅳ. Empirical Results
We estimate a structural vector autoregressive model for two groups of countries separately: East Asia and Europe. 17 The results of the forecast error variance 15 Kaminsky and Reinhart (1999) show that stock price indices are a significant predictor of currency crisis. As the crisis nears, changes in stock prices are about 40 percent below those observed in non-crisis periods. Weakening equity prices reflects both deteriorating cyclical position of the economy and reduced foreign demand as capital inflows are reversed and worsening balance sheets of firms. The beginning of a recession is also reflected in the stock market, which collapses a year before the crisis. 16 All stock price index data in this study are retrieved from Data stream (Thompson Financial). 17 For unit-root tests, not reported here for space reasons, we employ the augmented Dickey-Fuller (ADF) test, the Schmidt-Phillips (SP) test, and the Kwiatkowski-Phillips-Schmidt-Shin (KPSS) test. The null hypothesis for the ADF and the SP tests is that the stock price index is non-stationary. A time trend is included in all regressions; the number of lags used in the unit root tests is determined using the optimal lag length tests based on the Akaike Information Criterion (AIC), Hannan-Quinn Criterion (HQC), and Schwarz Criterion (SC). (Typically, all three criteria report the same results. When they are different, we take the result indicated by the AIC criterion.) The null hypothesis that the stock price index is nonstationary cannot be rejected for any East Asian countries with the ADF and SP tests. For the KPSS test, the null hypothesis is that the stock indices are stationary, which is rejected at the conventional significance level. When the same tests are applied to the first differences of the series, the unit-root null is strongly 40 韓國開發硏究 / 2011. Ⅳ  Tables 2 and 3, respectively. Global shocks, regional shocks and country-specific shocks are denoted as "U-G", "U-R" and "U-D" respectively. For brevity, we report only the variance decompositions of the domestic price index since the regional and global indices are mostly explained by regional and global shocks themselves. The pre-crisis and the post-crisis periods are defined as 1989:7:1 to 1997:6:30 and 1999:1:1 to 2010:11:11. The 18-month intervening period is dropped in estimation as a period of crisis and extreme volatility. Dividing the sample into the two sub-periods as mentioned above may not seem as compelling for European countries as for East Asian countries. Nonetheless, we maintain the same divisions for Europe and East Asia for consistency and also because a few previous years before the official introduction of the euro January 1999 were marked by extreme uncertainty and market volatility. For Europe, the two sub-periods are termed period I and period II. Table 3 shows that global, regional, and country-specific shocks on average explain 33, 26, and 41 percent of the variations in the domestic stock price throughout the whole sample period. The table also shows that the role of countryspecific shocks has declined over time in all countries without exception. Thus, in the rejected with the ADF and SP tests and the stationarity null is not rejected with the KPSS test. These results suggest that all the series contain a unit root and thus should be first differenced to achieve stationarity. The empirical results are available upon request.

<Table 3> Variance Decomposition of Domestic Index for European Countries
Financial Integration in East Asia: Evidence from Stock Prices 41 post-crisis period, 34 percent of variations in the average stock price can be explained by its own market shocks in all European countries while they did more than 61 percent before the crisis. There seem to be large variations in the extent of integration among the countries. France, the UK, Germany and the Netherlands are among the most integrated in the post-crisis period. On the other hand, the stock prices in such countries as Greece, Austria, Ireland, Norway, and Portugal exhibit strong influence of country-specific factors.
The corollary of the above change in the role of country-specific shocks is the increase in the sum of the roles played by global and regional shocks. Global shocks became more important virtually in all countries. They explain 14 percent of domestic stock market price variations in the pre-crisis period. More than 45 percent of domestic price variations are explained by global shocks in period II. Regional shocks became less important slightly on average from 25 percent to 21 percent. The tendency appears to be nearly universal. (Exceptions are Belgium, Greece, Portugal, and Sweden, where the role of regional shocks increased by a small insignificant magnitude.) In short, there appears to be clear evidence of greater financial market integration in Europe over the past two decades or so. Financial integration has progressed mainly globally. The role of regional shocks seems to have declined somewhat.
One of interesting questions for Europe is whether the participation in the Eurozone make difference in the progress of financial integration.
According to Frankel and Rose (1998), the economic criteria for OCA such as highly correlated business cycles are evolving over time. Due to the strong and positive effects of a common currency on international trade which in turn have positive effects on business cycle correlation across countries, countries may satisfy the condition after than before they join in a currency area. It is interesting to note that, in terms of financial integration, there is little difference between the Eurozone countries and the rest as a group. All but five countries listed at the bottom of the table became members of the EMU. (Denmark, Sweden, and the U.K. decide to opt out. Norway and Switzerland are not part of the European Union.) A comparison of the two groups does not reveal any significant differences. There is no evidence that countries specific shocks are less important -and thus financially more integratedin the Eurozone countries. Similarly, there is no indication that regional shocks are more important in those countries.
These results are reasonable given the fact that financial market openings pursued in European countries beginning in the 1980s have caused the stock market in each country to be more exposed to external/global shocks. It is also interesting to note that the introduction of the euro has accelerated the globalization of each stock market whether the country has become a member of the EMU or not. At same time, the fixed exchange rate arrangement under the European Monetary System (EMS) and the efforts of individual countries to participate in the single currency area seem to have gradually increased the extent of financial integration among the EMU and non-EMU members alike as indicated by the increasing role of regional shocks in virtually all European countries in the recent periods.  Asian economies, country-specific shocks are dominant in the determination of the domestic price index for the whole period estimation. They are responsible for 74 percent on average and for nearly 80 percent or more of changes in the local stock price index in all countries except Japan, Hong Kong, Singapore, and Australia. (This is not surprising given the open and advanced nature of the financial systems in these economies.) After the financial crisis, their role seems to have declined in all countries without any exception. In some cases, the decline is remarkable as in Korea, Hong Kong and Singapore. Nevertheless, they are still far more important than that can be observed in the European countries, explaining 70 percent or more in the majority of cases. China is also exceptional in that country-specific shocks continue to be dominant and there is little change in the post-crisis period explaining more than 90 percent variations in China"s stock prices are explained by its own domestic shocks. 18 Regarding the role of external shocks, we find that there is a substantial increase in the role of global shocks in the post-crisis period in Korea, Hong Kong, Singapore, Taiwan, and Australia. For instance, 35 (24) percent of variations in the domestic stock price are explained by global shocks in Singapore (Korea) in the post-crisis period, up from 16 (5) percent in the period before the crisis. On the other hand, little change is observed in China and Malaysia. The two countries are well known in their response to the crisis, in particular, reinforcement or new imposition of capital controls. Regional shocks became more important in the post-crisis period. On average, their contribution doubled from 7 to 14 percent. The increased role is noticeable in Japan, Korea, Hong Kong, and Singapore. Malaysia is the only exception to the trend.

<Table 4> Variance Decomposition of Domestic Index for East Asian Countries
In European countries, financial development and the opening of the country"s stock market seems mainly in the form of the increased exposure to global shocks. Regionalization has already been established in period I as a result to long sustained efforts to achieve economic and political integration within the region. Thus, in terms of the percentage contribution, regional shocks became less important in period II. In East Asia, in marked contrast, both global shocks and regional shocks became more important in a balanced fashion although their joint contribution explains on average a third of variation in the stock price.

Ⅴ. Robustness Check
The results reported in the previous section appear reasonable. However, they may rest on some assumptions that may not be tenable. We thus investigate the robustness of the empirical results by considering various alternatives. For space reasons, we provide the results for post-crisis East Asia only. In Model B, we employ the Choleski decomposition as the method of identification, which imposes restrictions on the presence (or absence) of contemporaneous effects. The results are very close to the baseline model except the sharp decline in the role of regional shocks. Model C estimates the regression in levels with the Choleski decomposition. Given the fact that determining the presence of unit roots and whether the variables are cointegrated or not is difficult and subject to ambiguity, we estimate for Model C using levels in all variables. The results are also broadly similar to those of the baseline model except that the relative roles of global and regional shocks vary more than 10 percent in Japan, Hong Kong, Singapore, Thailand, and Australia.
In Model D, we employ the regional indexes that are obtained as weighted average of all individual country indexes using the market value of publicly traded shares as of December 31, 2010 (from CIA World Factbook). 19 For the global index, we use the simple average of the US S&P 500 and the MSCI AC Europe-Price Index. Otherwise, it uses the same setup as the baseline model. The change in the definition generally increases the role of regional shocks and reduces that of global shocks while the combination of the two explains roughly the same fraction of the local stock price index as in the baseline. China is the most interesting case of all. Reflecting the influence of its economic size and international trade, its own stock price movements seem to be heavily reflected in the regional index. Thus more than 60 percent of its own index is identified by regional shocks themselves. The increase in the role of regional shocks in Hong Kong, Singapore and Taiwan seems to have the same root as that of China. Japan is the opposite case which loses its influence in the regional index to a substantial degree. It also appears that the Philippines, Australia and New Zealand are underrepresented in the construction of the regional 韓國開發硏究 / 2011. Ⅳ index such as MSCI AC Asia Pacific. This result suggests that the nature of global and regional shocks is subject to a great deal of uncertainty. The absolute level of the role of their contributions on the local stock price index is less important than its variation over time.
In Table 6, we estimate the baseline model and its variation using monthly stock price index. Monthly averaging eliminates larger parts of idiosyncratic daily price movements than does weekly averaging used in the baseline model. The roles of the three shocks change substantially. First of all, country-specific shocks are much less important while global shocks become far more influential. On the other hand, the role of regional shocks is reduced to a minimum level. Once the influence of global shocks is taken into account, regional shocks play almost no role. This suggests that the universality of the stock price trend implied by the US and European stocks.

Ⅵ. Conclusion
This paper investigates the extent of global and regional financial integration in East Asia in the stock market. We employ a structural VAR model to separate the underlying shocks into "global", "regional" and "country-specific" shocks. The estimation results show that country-specific shocks still play a dominant role in East Asia although their role appears to have declined over time, especially after the 1997 financial crisis. The roles of global and regional shocks have increased in the postcrisis period.
Comparison with the stock markets of the European countries reveals some interesting differences between the two groups of countries. First of all, East Asian stock markets are much less integrated globally or regionally. For instance, global and regional shocks account for two thirds of stock price movements in Europe but only one third in East Asia. Secondly, high level of regional integration in Europe seems to have already been achieved in the 1990s even before the introduction of the euro and further integration of the financial market has progressed in the direction of globalization. On the other hand, stock market integration in East Asia seems to be more balanced in that both global and regional factors have become more important over time.
The empirical results remain largely unaffected if we use levels instead of differences of variables or different identification schemes such as the Choleski decomposition instead of the Blanchard-Quah type long-run restrictions. However, the relative weights of global and regional factors strongly depend on the definitions of the global and regional indexes. For instance, using the weights based on the most recent market value of capitalization dramatically increases (reduces) the role of China and Hong Kong (Japan) in the regional factor. We also find that the results are sensitive to the frequency of data employed in the study. For instance, using monthly data instead of weekly significantly increases the role of global factors, which explains more than 50 percent of fluctuations in the stock prices. On the hand, regional shocks become negligible. This suggests that over the long term stock markets in East Asia follow the global trend Financial Integration in East Asia: Evidence from Stock Prices 45  Our results also contradict the previous study by Eichengreen and Bayoumi (1999) who find that East Asian countries are almost as qualified as the EMU countries in terms of OCA criteria. We find that regional shocks tend to play increasingly more important role in East Asia as financial markets become more integrated with those of the United States and Japan. However, their roles are not as important as found in the EMU countries. We also find that the hypothesis of the endogenous OCA criteria may not apply to the financial market. In Europe, the extent of globalization or regionalization of the stock market seems to be hardly different whether a country participates in the EMU or not. 46 韓國開發硏究 / 2011. Ⅳ Some caveats remain. An investigation of robustness of the empirical results indicates that the global-regional-country specific decomposition depends on the definition of the regional and global indexes. Finding ideal indexes for the purpose remains a subject of future study. Likewise, the fact that decomposition depends on the frequency of the data -e.g., weekly or monthly -poses an important issue in empirical analysis. ABSTRACT Ever since the UN Summit agreed on the MDGs in 2000, OECD/DAC member countries have taken poverty reduction as the main goal of their aid. To achieve this goal, all donors and recipient countries agreed on the Paris Declaration on Aid Effectiveness in 2005. To monitor and evaluate the progress in the targets of the Declaration, all donors and recipients got together periodically, and the 2011 conference was held in Busan, Korea. As part of this effort, this paper aims to assess the extent to which DAC donors have allocated their aid to achieve the MDGs during the latest millennium era: [2005][2006][2007][2008][2009]. In addition, to compare the aid allocation performance between DAC members and non-DAC emerging donors, this paper also assesses the aid allocation performance of Korea (KOICA) for the same period. The analysis of this paper shows evidence contrary to the recent literature findings that donors tended to select, as their aid recipients, those countries that warranted more aid on account of their acute development needs, and good policies and institutions. The difference between the recent literature and this paper is attributed to the different sample periods and/or the weaknesses of the estimation models and methods adopted in the literature. This paper shows why a different estimation method is adopted and why its estimation results are more reliable and convincing. This paper also shows the difference between DAC and non-DAC donors in the aid allocation performance by analyzing aid allocations by the representative aid agency of Korea (KOICA), and recommends some policy measures to be taken by both DAC and non-DAC donors.

ABSTRACT
This paper uses the horizontal regulation system as the base analysis framework. The study clearly defines the regulatory goals of the followings: the horizontal cross-ownership regulations on program provider (PP) and platform provider, the vertical regulation on cross-ownership between PP and platform operator, the regulation on cross-ownership of program provider by terrestrial broadcasting company, and the regulation on cross-ownership between terrestrial broadcasting company and platform provider. Then, by analyzing the conformity between goals and criteria of regulations and the adequacy of the regulation level according to regulatory purposes, this paper examines the justifiability of each regulation and extracts improvement measures that suite regulatory purposes. This analysis finds following appropriate measures: replacing the horizontal cross-ownership regulation on PP with conduct regulations, such as designating major broadcasting programs or replacing the current criterion of cross-ownership regulation from sales to the audience market share; reshaping the horizontal cross-ownership regulation on platform provider so that system operator (SO), satellite broadcaster and Internet protocol television (IPTV) operator would be applied by the same regulation based on the number of subscribers of pay television services; and discontinuing other cross-ownership regulation. In this way, the study shows that with appropriate regulations on cross-ownership of PP, there would be no need for additional regulation on vertical integration between PP and platform operator. On the other hand, given that the regulation on terrestrial broadcasting cross-ownership of PP could be justified only by regulatory purpose of the protection of the diversity of public opinions, it would be desirable to replace the current criteria of the number of PPs with the criteria of the audience market share. Lastly, the study shows that when platform operator is targeted by the cross-ownership regulation based on the number of subscribers of pay television services, the regulation on cross-ownership between terrestrial broadcasting company and platform provider should be replaced with conduct regulations, such as designating must-offer channels and major broadcasting programs.     The storm and flood insurance, which was introduced recently to substitute the disaster relief system to moderate government's financial burden and enhance people's effort to reduce damage, seems to suffer various problems. This paper conducts a theoretical analysis on various aspects of the storm and flood insurance to provide insight on those issues and draw policy implications. First, the coexistence of disaster relief with the storm and flood insurance is likely to harm the penetration of the storm and flood insurance. Second, the current premium system is likely to induce people to make less efforts to reduce damage due to moral hazard problem. Third, current support for damage-reducing efforts may not fulfill its purpose and hence should be scrutinized carefully.  Upon selecting preferred bidder in Public-Private Partnership projects, multi-dimensional procurement auction, where price factor and non-price factor are evaluated, is used. This paper tries to analyze bidding data in BTO road projects. It is shown that a winner tends to get higher score in bidding evaluation, which is partly due to increase in base score as well as fiercer competition among bidders. It turns out that score margin in non-price factor was determinant in selecting winner. Also, there was no competition when the level of bonus point was set too high. For price factor, it costs 730 million KRW per score in construction subsidy by government, while it costs 2.43 billion KRW per score in toll revenue. For non-price factor, it was estimated to cost 2.30 billion KRW. Based on the results, it was suggested that we should have appropriate level of bonus point for first initiator, change in scoring rule in construction subsidy part, adjustment of base score in evaluation.   [ Figure 17] Decomposition of Winning