Redistribution with education and endogenous crime

Charles Ka Yui Leung Department of Economics and Finance City University of Hong Kong Hong Kong Kuzey Yilmaz Department of Economics Koc University Istanbul, Turkey Nelson Lee Department of Economics and Finance City University of Hong Kong Hong Kong

The authors are grateful to Jun Zhang, the seminar participants of HKEA meeting for comments and discussion, and City University of Hong Kong for nancial support. The usual disclaimer applies. Department of Economics and Finance, The City University of Hong Kong, Kowloon Tong, Hong Kong SAR, China. Phone: (852) 2788-9604, Fax: (852) 2788-8806, Email: kycleung@cityu.edu.hk. Department of Economics, Koc University, Sariyer, Istanbul, Turkey. Phone: (212)-338-1229; Fax: (212)-338-1651; E-mail: kuyilmaz@ku.edu.tr; Homepage: home.ku.edu.tr/~kuyilmaz.

Abstract Recent literature has acknowledged the importance of criminal activities in the determination of aggregate output. This paper introduces a criminal option for the unskilled workers as an alternative career choice into the overlapping generation model developed by Fender and Wang (2003, IER). On the positive side, this model generates empirically plausible enrollment and skill premium. On the policy side, this model compare the implications of dierent redistribution policies on labor force composition, skill premium and income inequality. JEL Classication Number: D30, H20, I20, J24, O15 Keywords: endogenous crime, TO BE ADDED.

Introduction

Since Becker (1962) and Mincer (1958), the academic literature on human capital quickly expands, and it is then related to a number of important issues, such as nutrition, technological innovation, economic growth and formal schooling system. For instance, Benhabib and Spiegel (2005) present evidence that a critical level of aggregate human capital is associated with the technological change. Klenow and Rodriguez-Clare (2005) nd that a model with positive externality in human capital is much more plausible than the one without in explaining cross-country growth experience. At the individual level, Psacharopoulos (2004) provides empirical evidence on the returns to investment in education and nds that the world average private rate of return for those who received secondary education is around 19% while those who received higher education is 17%. While other researchers may obtain alternative estaimtes, most agree that individual return on the investment in human capital are in general signicantly positive.1 The theoretical literature are consistent with these empirical ndings. For instance, Iwaisako (2002) introduces a two-period overlapping generation model to show that an economy may have endogenous cycle, a poverty trap or permanent growth. Moreover, the economy can "escape" from the poverty trap and turns to permanent growth. Boldrin (2005) presents an over-lapping generation model where school attendance and nancing of schooling are endogenized. When dierent initial conditions are given, he shows that the economy can suer from poverty trap or keep on growing. Unfortunately, not everyone study and work hard. In fact, an individual could choose a "criminal career path". From the aggregate point of view, criminal activities can adversely aect the economy in several aspects including the inow of FDI, tourism, concentration of experts, and social breakdown.2 While criminal activities are not avoidable, it is negatively related to economic growth. Becker (1968) provides a theoretical analysis on the issue of crime and law enforcement. He concludes that although criminal activities are harmful, it is also costly to prevent all the criminal activities. So, there is an optimal policy determining the equilibrium level of crime rate. Recently, a signicant amount eorts has been devoted on the macroThere is a very large literature on estimating the return of education, see Hanushek and Welch (2006) for areview of the literature. 2 Please see Burnham et al. (2004), Levantis and Gani (2000), Axarloglou (2005), Conley and Wang (2006) for survey.
1

economic implications of crime. For instance, Wong (1995) nds empirically that "growing economic prosperity and rising educated standard contributed to the overall decline in crime rate". Axarloglou (2005) make use of the states level data of USA to show that relative crime rate among dierent states is one of the most important determinants to aect the inward FDI. Rodrik (1999) nds empirically that external shock, social unrest and the eectiveness of the government are interacting with each others, which can interrupt the economic growth process. Under an external shock, a country with weak government and high level of social unrest suered most. The research direction is pointed to the relationship between inequality and crime rate On the theoretical side, Imrohoroglu et. al. (2000), among others, tries to explain the positive correlation between inequality and crime rate observed in US data through a general equilibrium model. Mehlum et. al. (2005) provides a static-theoretical framework explaining how social conditions and economic performance are related. They show that "the economy may end up in a poverty trap with high crime and low production or obtain full modernization with low crime rate". They did not consider the education choice. Burdett et. al. (2004) presents a dynamic search theoretic model in which poverty trap may endogenously arise in the face of a high crime rate. Lochner (2004) provides a partial equilibrium model of work, crime and human capital investment. The model predicts that dierent kinds of crime decrease with the level of human capital (as the higher the human capital level, the higher the income and thus, the higher the cost of incarceration.) Huang et. al. (2004), in addition, explains the above relationship based on a continuous time model of search and matching without capital market. It shows that one possible equilibrium will exhibit high crime, low levels of education attainment, unemployment and poverty. While a review of the literature is beyond the scope of this paper, the following table nevertheless provides perhaps an over-simplifying summary of some previous papers. P lease Insert table 1 here To complement the previous literature, this paper takes a preliminary step to investigate how occupational decisions, criminal activities and physical capital accumulation are related in a dynamic general equilibrium model. More specically, this paper introduces a criminal option for the unskilled workers as an alternative career choice into the overlapping generation model 4

developed by Fender and Wang (2003). A merit of that framework of Fender and Wang (2003) is that, as in the case of many developing countries, young agents who would enroll in college may need to borrow. Thus, tuition loan competes with capital investment for funds in the market, and the latter is important in determining the marginal product of skilled workers, which in turn aects the career choice of young agents to enroll in education or not. Clearly, multiple equilibria are possible, as previous theoretical works have pointed out. Since Fender and Wang focus on the analytical results, it seems natural for this paper to complement the literature by providing some simple calibration results. Interestingly, in all of our numerical simulations, unique equilibrium is obtained. We then further proceed to policy analysis. In particular, we re-examine the "ordering" of the redistribution policies studied by Hanushek et al. (2003). Clearly, this paper is related to the literature on human capital and economic growth.3 Recent literature seems to be very positive on the return of human capital investment. For instance, Rauch (1993) justies that investment in education induced positive externality and Acemoglu and Angrist (2000) uses the Compulsory Schooling Laws as instrumental variables and nds that the positive external rate of return range from 1-3%. Thus, this may give a justication for the government intervention in the education market4 . Since this is issue is still under debate, this paper simply studies a case without externality and examines how dierent public policy would aect the individual incentives in human capital investment, as well as the aggregate outcomes (such as skill premium, income inequality, etc).5 The structure of this paper is simple. Section 2 presents the theoretical model where the economy is a small open economy. We try to further illustrate the steady-state equilibrium through numerical examples. In Section 3, we extend the benchmark model by changing the small open economy into a closed economy, in this case, the real interest rate would be an endogenous variable instead. At the same time, government policies would be evaluated in terms of the redistribution eects. The last Section concludes.
See Schultz (1997), Lee et. al. (1997), Benabou (1994, 1996, 2002), Galor & Moav (2004), Bond et. al. (2003), Li and Zhu (2004), Hanushek, Jamison and Jamison (2006), among others. 4 See Hanushek et al. (2003, 2005) for a review of the literature. 5 For a review of the literature, see Hanushek and Welch (2006).
3

A Baseline Model: a Small Open Economy

Our model is an extension of Fender and Wang (2003), which is a twoperiod overlapping generation (OLG) model. To facilitate the comparison, our assumptions and notations would be kept as close to Fender and Wang as possible. To simplify the analysis, we focus on the case of a small open economy, and thus the interest rate is xed. In this economy, each agent is endowed with one unit of labor. Agents are identical and the only dierence lies in the disutility towards education and robbery.6 The model focuses on occupation decision and investment in physical capital. Agents live for two periods7 and in the rst period, they make the job decision. Each generation is divided into three groups, educated agents, uneducated agents and robbers. To simplify the analysis, the population of the economy is normalized to unity i.e. Xe,t + Xu,t + Xr,t = 1 (1)

where t is the time index and Xe , Xu and Xr are the proportion of educated agents, uneducated agents and robbers in the whole population respectively. There are two kinds of disutility and representing the disutility for those who receive education and those who choose to be robbers and both of them are uniformly distributed on the interval [-, ]. A product of these two uniform distributions formulate a joint-distribution determining the equilibrium portion of education agents and robbers. Each people denoted one period to go to work inelastically (either in the rst period or second period depending on the job decision) and then receive income8 . The utility function is assumed to be linear and only the disutility incurred together with the amount of consumption, C, aect the utility level.
As it is an extension of Fender & Wang (2003) model, we try to use similar notations to represent the same variables in order to avoid unnecessary complication. 7 We implicitly assume that population of the economy is constant over time i.e. the birth rate equals to the death rate. 8 Although robbers do not participate in any formal employment, in our model, we assume that the attempt to rob the others during the matching process is the working hour for robbers.
6

where = e, u, r, with represents the disutility incurred to the individual if he/she enrolls in college, and represents the disutility incurred to the individual if he/she chooses to be a robber. If an agent chooses to be skilled labor, he pays for the school fees, , in the rst period. In this model, the young generation were born without bequest and hence they need to borrow in order to nance the school fees. After that, he receives education in the rst period, incurs disutility and then go to work in the second period, receive their wage and then repay the education loan. For those who choose not to receive education, they become unskilled workers or robbers. Unskilled labors go to work in the rst period and then invest their wage either in physical capital or become lenders. In the second period, they receive principal and interest return. For robbers, they wait until the second period. In the second period, after skilled and unskilled workers receive their income, they go to market to look for goods for consumption.9 The interaction between all agents are governed by a random matching process. If either a skilled or unskilled worker meet a robber, robbery exists and certain proportion of the victims income become the income of robbers. If a robber meet another robber, no robbery exist and their income (consumption) would be zero. Once the total income for each of the agents is nalized, they spend all of their income for consumption.10 Figure 1 provides a visualization of the setting. P lease Insert F igure 1 here. With the probability of meeting a robber, individuals make the occupational decisions base on the expected income and disutility. The expected disposable income, which equals to expected consumption level, of skilled (unskilled) labors and robbers are denoted as EWe,t (EWu,t ) and EWr,t and
In practice, there are some high-tech robbery through stealing and using passwords of others and then re-allocate money to their own accounts. Due to the lack of data, and the possible complexity by introducing high-tech robbers in the model, we can only abstract them away from the model. 10 For the sequence of actions of dierent kinds of agent, please refer to gure 1.
9

C (e) if = e C (u) if = u . U (; , ) = C (r) if = r

(2)

 () are the proportion of income will be robbed if skilled (unskilled) labor meet a robber in the second period. We have11 : EWe,t = We,t (1 Xr,t1 ) (1 + rt1 ) EWu,t = (1 + rt1 )(1 Xr,t1 ) and EWr,t = We,t Xe,t1 + (1 + rt1 )Xu,t1 Clearly, investment in human capital is attractive only if the utility level, evaluated in the second period, is higher than that of uneducated people and robbers. As we have a continuous distribution of disutility, in equilibrium, the marginal agent i should be indierent between choosing to be skilled or unskilled labor. Hence, EUe ( , ) = EUu (3)

where represents some critical values for disutility of education12 . Similarly, among the uneducated population, they can choose to be unskilled worker or robbers. As in the skilled-unskilled decision, the marginal agent is indierent between choosing to be skilled or unskilled labor. Hence, we have: EUr (, ) = EUu , (4)

where represents some critical values for disutility of robbery. The equilibrium occupation decision for each generation can be summarized in a two stage diagram.13 P lease Insert F igure 2 here. Each individual makes the job decision, by comparing the expected utility between being robbers, being an unskilled labor, or being a skilled labor.
For details of the derivation, please refer to appendix A. To solve for the shole system, we have to determine the critical value for disutilities. Appendix C provides us the details about the relationship between the critical value for disutilities and equilibrium level of skilled labor and robber. For details, please refer to appendix C. 13 For details, please refer to gure 2.
12 11

The resulting expected utility can be incurred in the previous sub-game. Each individuals make their own occupational choice by backward induction. Moreover, we adopt an assumption from Acemoglu and Verdier (1998), namely, the disutility of incurred as being a robber is a private information. If the disutility incurred as being robber were a piece of public information, each agents in the economy would know who is going to be robbers and thus, the random matching process may dissipate. Now, we turn to explaination of the production side of this model. There is only one good in this economy, and it can be produced by unskilled labor or skilled labor together with physical capital. Thus, we have Yt = Yu,t + Ye,t (5)

where Yt is the total output in period t and Yu,t and Ye,t are the amount of production produced by unskilled labor and skilled labor with physical capital respectively. The marginal product of unskilled labor is assumed to be a constant, and so Yu,t = Xu,t . For skilled labor, the production process requires both skilled labor and physical capital and it is assumed to follow a constant returns Cobb-Douglas production function:
1 Ye,t = AXe,t1 Kt1 = AXe,t1 kt

(6)

where A is the total factor productivity, or simply TFP, and kt is the capitalKt skilled labor ratio which is equal to Xe,t1 . With competitive factor markets, the returns of dierent production factors equal to their marginal products and thus:
rt = (1 )Akt ,

(7) (8) (9)

We,t =

1 Akt ,

Wu,t = ,

where denotes the depreciation rate of physical capital. Denoted the expected utility (for the cohort in their second period of their lives at time t) of skilled labors, unskilled labors and robbers as EUet , EUut and EUrt respectively, we have14 : EUet = We,t [1 Xr,t1 ] (1 + rt1 )
14

(10)

To see how the equations are derived, please refer to appendix A.

EUut = (1 + rt1 )Wu,t1 [1 Xr,t1 ] EUrt = We,t Xe,t1 + Wu,t Xu,t1

(11) (12)

where EUe (EUr ) is the expected utility of skilled labor (robber). While Xe , Xu and Xr represent the proportion of educated agents, uneducated agents and robbers in the economy respectively, Xe (Xu ) represents the probability for a robber to meet a skilled (unskilled) labor and Xr represents the probability for skilled (unskilled) labor to meet a robber in the random matching process.15 The next question is, given the wages and interest rate, whether there is a non-trivial equilibrium (that is, all Xe , Xu and Xr are interior solution). Notice that the returns of dierent production factors equals to their marginal products. If the proportion of skilled labor in the economy is zero, then the value of capital-skilled labor ratio would tend to innity. If there is only one robber in the economy, that robber must meet either a skilled or unskilled labor and the expected income and the level of disutility incurred would be extremely large. Thus, the proportion of robber of the economy is not likely to reach zero. On the other hand, the propotion of robber in the economy cannot be one, otherwise, the expected income for robber would be zero. Thus, we may have interior solution for Xe , Xu and Xr . In our numerical implementation, if there is an equilibrium, it is unique. However, there are some parameter combinations such that there is no solution. To close the model, it is necessary to pin down the interest rate of the model economy. We rst present the case of a small open economy, where the interest rate is exogenous given.

2.1

Results from a small open economy model

In this section, we focus on the small economy case, where the real interest rate is determined by the rest of the world and hence is regarded as given by
To derive the probability for a skilled or unskilled labor to meet a robber, we rst (X r assume the matching function, m, equals to (Xee +Xu )Xu ) . And the matching function +Xr +X indicates the probability for a meanful matching (in this case, a robber meet a worker). m Hence, the probability for a robber meeting a skilled labor equals to Xr (XeXe u ) . After the +X substitution of m into the probability, it becomes Xe .By similar method, the probability m for robber meeting unskilled labor equals to Xr (XeXu u ) = Xu . On the other hand, the +X probability for a labor to meet a robber equals to Xe m u and by substituting function m +X into the probability, it can be pinned down into Xr .
15

10

the model economy. In addition, we assume that the world interest rate is constant over time, and hence, all variables take on their steady-state values. From now on, we drop the time subscripts. The equilibrium allocation is easy to describe. First, the capital-skilled labor ratio can be determined immediately16 , 17 : 1 1 k= (13) r+ and since the wage return for skilled labor, We , is a function of k, it become an "exogenous" variable too. The variables remains to be solved are the composition of the population. As we have normalized the population into one, once we nd out the steady-state value of any two groups we can nd out the remaining from equation (1) . Our main task remain is how to determine the value of Xe , Xu and the critical value for disutilities, and . 2.1.1 Agent Behaviour

As mentioned before, occupational decisions are made base on the corresponding expected utility generated. Each agent compares the expected utility incurred if they participate in dierent occupations and then choose the one with the highest expected utility. Thus, we have: EU (, ) = max {EUe (, ), EUu (, ), EUr (, )} (14)

Recall that both kinds of disutilities are uniformly distributed, thus, the whole population can be represented by a cube. Figure 3a and 3b show the birds eye view of the cube and are divided into three areas showing dierent groups of agents under dierent situations and the boundaries dividing them18 . Base on the results generated, a nonlinear relationship between the proportion of skilled labor (robber) in the economy and the critical disutilities for education ( ) and robbery ( ) can be found. By assuming that > 0, we have: Z Z 1 ( + )2 1 Xe = dd (15) 42 2 42
K Recall that the capital-skilled labor ratio equals to Xe . Thus, the steady state capital stock, K = kXe and the physical accumulation process reduces into a function Xe . 17 For the derivation of k, please refer to appendix B. 18 To see how to derive the boundaries, please refer to appendix C. 16

11

and Xr =

1 42

( + )2 dd + 2

1 42

(16)

and the case for 0 can be handled similarly. (The details are in the appendix). Base on the results generated, we turn our attention to comparative statics. Recall equation (3) and (4), these equations represent the situation where expected utility for being 1) skilled and unskilled labor and 2) robber and unskilled labor for the marginal agent are the same respectively. By solving these two equations together with (15) and (16) above, the steady-state value for Xe and Xu can be determined and thus, all other endogenous variables can be solved correspondingly. To see how may change in parameters aects the endogenous variables, we have to investigate how the change in parameters aects the four equations. Taking the change in education cost, , as an example. A change in education cost aects only equation (3) , together with the relationship between the proportion of skilled labor (robber) in the economy and the critical value of disutilities generated above, we have: Xe (1 + r) = < 0 Xe Thus, when there is a rise in education cost, the equilibrium proportion of skilled labor decrease. However, the changes in unskilled labor and robber are unknown. Marginal agents may be forced to abandon their human capital investment decision because of a rise in education cost. Spared labor should then choose either to be unskilled labor or robber. Following the above analysis, occupational decisions are made base on the expected utility. Among the spared labor, those endowed with high level of disutility towards robbery become robbers while those who really hate to be robbers become unskilled labor. For those who are not signicatly aected by disutilities for education and robbery, they occupational decisions may be largely aected by changes in expected income for dierent groups. Note that the returns of human capital investment come in monetary as well as subjective terms. Thus, it is possible to have the equilibrium expected income for skilled labor is smaller than the unskilled labor counterpart. If 12

the expected income for skilled labor is higher than that of unskilled labor, a decrease in skilled labor decrease the expected income for robbers no matter the spared labor choose to be unskilled labor or robber. Thus, as a result of rise in education cost, the equilibrium proportion for robber may rise or fall while that of unskilled labor increase. Similarly, If the expected income for skilled labor is smaller than that of unskilled labor, a decrease in skilled labor rise the expected income for robbers no matter the spared labor choose to be unskilled labor or robber. Thus, as a result of rise in education cost, the equilibrium proportion for robber increase while that of unskilled labor is uncertain. P lease Insert F igure 3 here

Numerical Calibrations

Since closed form solution is unlikely available for this model, we will turn to the numerical simulations. In this section, we will explain how the numerical values of parameters are determined, and how the equilibrium allocation will change as a results of a change in education cost (), proportion of income can be robbed from skilled labor (), and unskilled labor () . As in any numerical studies, the parameter values are crucially important. In this paper, to impose more disciplines in our numerical works, we intend to adopt parameter values which are consistent with the empirical literature, which is summarized in T able 2. The rst variable that we want to discuss is the share of physical capital in the Cobb-Douglus production function, (1 ). According to Cooley (1995), that gure should be around one third for the U.S. data. Miller and Upadhyay (2002) examine the CobbDouglus production function base on the panel data of 83 countries for 30 years representing dierent regions of the world. According to their estimation, the capital share range from 0.31-0.61 depending on the income level of the targeted country we are considering. Lee (2000), on the other hand, tries to estimate the coecients base on dierent model specications with South Koreas data from 1966 to 1997. The resulting estimation are similar to Miller and Upadhyay (2002), ranging from 0.251-0.497. Takeuchi (2005), in addition to existing literature, estimate the capital share bases on data in Japan. The estimates range from 0.3 to 0.4. 13

P lease Insert T able 2 Here With serious data limitations and the lack of existing literature, we are not certain about the share of income a robber can rob from victims at a time. The estimates for and are thus not very clear.19 The best estimate we are aware of it is from Imrohoroglu et. al. (2000). They make use of the average cost of property crime to victim as a proxy to the criminal earnings from property crime. Base on the data of USA, they nd that the average cost for property crime reaches 16% in 1979 and 15% in 1989 of the average after-tax annual earning of worker. They also mentioned that estimation in Wilson and Abrahamse (1992) yields similar value equals to 0.2. 3.0.2 Base Results

Numerical examples in this sub-section mainly investigate the eects of changes Xe in , and to the ratio of skilled and unskilled labor, Xu and the ratio of unskilled labor and robber, Xu . is assumed to be 0.4 and the real interest Xr rate is assumed to be 4%. is assumed to be equal to 4 and equals to 4.5, so that the wage ratio for skilled and unskilled labor, which is know as the college-premium, would be within the range of previous estimates (for details about college-premium, please refer to T able 3).

Dear KUZEY: I am sorry that I forget the nal parameter values that you used. would you please update the table? and i will try to ll up the discussion part... Thanks a lot....
For instance, while widely cited works such as Barro and Lee (2000) provide rich information about education, it does not have much information about crime.
19

14

Parameter value A 5 4 0.5 1 0.1 0.1 4 3.5

P lease Insert T able 4 Here 3.0.3 Distribution of Equilibria

(to be modied) Figure 4 to gure 9 summerized the distribution of equilibria under dierent parameter values. This is an attempt to investigate under what parameter values we can not generate any exible equilibirum in the model and see if it provides any theoretical implication. Taking gure 8 as an example, this gure shows that exible equilibira exist below the curve which is concave to origin. Where is the education cost and is the porportion of income being robbed if skilled labor meet a robber, we nd that equilibrium exists only if a rise in is associated with a fall in education cost such that the return to skilled labor remain more or less the same. However, when any one of these value increase continuously, the equilibrium may dissipate and the model may collapse. For details of other variables, please refer to the gures.

The Basic Model under Closed Economy

Thus far, we have assumed that the real interest rate is given to the economy and determined by the rest of the world. Also, there is no tax and the private loan market can by itself nance the human capital investment. In practice, these assumptions may not hold. We relax these assumptions in this section. First, we will endogenize the interest rate and (13) will no longer holds. 15

Second, we will need to introduce income tax into the model. In particular, skilled and unskilled workers will be subject to income tax rates e and u respectively, and the expressions (10) and (11) will be adjusted to EUet = (1 e ) We,t [1 Xr,t1 ] (1 + rt1 ) , EUut = (1 u ) (1 + rt1 )Wu,t1 [1 Xr,t1 ] . (17) (18)

On the other hand, the factor market conditions, (7), (8) and (9) continue to hold. To complete the description of the model, we will need both the market clearing conditions and the government budget. Due to the Walras law, it suces to have the capital accumulation equation and the government budget. The following expression describes the capital accumulation process. Kt+1 (1 ) Kt = (1 u ) Wu,t XU,t (1 u ) Wu,t1 (1 + rt1 )XU,t1 +(1 + rt1 )Xe,t1 Xe,t = rt1 Xe,t1 (1 u ) rt1 XU,t1 = rt1 [Xe,t1 (1 u ) XU,t1 ] . (19) It says that the net increase in physical capital (left hand side) is the sum of the two terms. First, you have the amount of (after-tax) saving of the unskilled workers made in period t, net of the withdrawal of deposit of the unskilled workers who made in period (t 1), i.e. (1 u ) Wu,t XU,t (1 u ) Wu,t1 (1 + rt1 )XU,t1 . The second term is similar. It is the repayment of tuition loan (for skilled workers) made in period (t 1), net of the new tuition loan made in period t, i.e. (1 + rt1 )Xe,t1 Xe,t . The government budget, on the other hand, can take dierent forms, depending on the policy behind. We here highlight a few policies to be studied.

TO BE CHANGED 5 Government Policy Evaluation

In this section, we will conduct some policy experiment within the current framework. There are several justications for it. First, the laissez faire 16

equilibrium with endogenous robbery may not be ecient. Even if it is, government in practice may still express concern on the equality issue on top on eciency.20 In practice, governments are often very involved in providing subsidized education.21 Thus, it may be instructive to examine how dierent redistribution policies may aect dierent agents welfare. In this model, the government is very simple. She has to maintain a balanced budget and is constrained to rise revenue through a proportional income tax. She will then redistribute income to the agents who have legal jobs. In this section, we will consider two redistribution plans: 1) Education Subsidy and 2) Wage Subsidy. Similar policies have also be studied by Hanushek et. al. (2003). In a static general equilibrium model calibrated to match the U.S. data, they nd that unless the externality of education is very large, the wage subsidy scheme always dominates the education subsidy scheme, in the sense that with any given level of eciency, wage subsidy regime can always gives a higher level of income equality. In contrast, this paper builds a dynamic general equilibrium model with endogenous capital accumulation and crime formation, intended to replicate the situation of some developing countries, with reasonable success. We however, has abstracted away the possibility of unsuccessful education. It is interesting to see with all these dierence, whether such dominance will sustain. 5.0.4 Education Subsidy

In this case, the government implements only education subsidies as the redistribution program. With education subsidies, the government taxes the income of skilled labor and unskilled labor at the rate equals to e and u respectively. Since the government is restricted to maintain the balanced budget, the government simply equates the total expenditure to tax revenue. Moreover, the education subsidy is assumed to be distributed equally among skilled labor. Thus, the budget that government facing satises the following equation: e We,t Xe,t1 + u Wu,t Xu,t + Xe,t g = Xe,t where g equals to the education cost that an agent actually pays. Besides, g is assumed to be strictly positive which means that the education subAmong others, see Benabou (1996a, b, 2000, 2002), Benabou and Ok (2001), Benabou and Tirole (2005), Hanushek and Javier (2003), Hanushek et. al. (2003) for more discussion. 21 Among others, see Hanushek (2002), Hanushek and Welch (2006) for a survey.
20

17

sidy provided by the government does not cover the whole education cost. Obviously, the tax rate on skilled and unskilled workers can be equalized, e = u.

Dear Kuzey here g is NOT directly observable... so we can play with it a little bit... almost like a free parameter...
5.0.5 Wage Subsidy

A wage subsidy is a direct subsidization of wage income for unskilled labor. In our model, the proportional tax, denoted as e , levied only to skilled labor and the tax revenue are distributed equally among unskilled labor. The government budget constraint is thus: e We,t Xe,t1 = sXu,t , where s is the wage subsidy given to each unskilled labor. In this case, the payo equation for unskilled workers, (18), needs to be rewritten as EUut = (1 + s) (1 + rt1 )Wu,t1 [1 Xr,t1 ] . (20)

5.0.6

Base Case Outcome

TO BE CHANGED
PLEASE UPDATE!!!!!!!!!!!!!!!!!!!!1 After the discussion on the numerical examples, we try to compare the results estimated through our model with the estimates of existing empirical Xe literature. Table 6 summarize the value of Xu across countries in dierent years. In our model, skilled labor are designated as post-secondary degree holder or higher while unskilled labor are designated as school leaver. The labor ratio in dierent countries shows a rising trend from 1990 to 2000. For instance, the labor ratio of South Korea rises from about 0.16 to about 0.36, more than 100%, in 10 years. With the estimates in dierent countries, table 18

6 also shows that the resulting labor ratio of our numerical examples lies with the boundary of some countries such as South Korea and Taiwan and similar to that of Bulgaria, a transitional economy. Table 1, on the other hand, We summarize the value of college-premium and it is dened as Wu (Please refer to table 1 in the last section). According to the existing literature, collegepremium is generally higher in developing countries. For example, Blom et. al. (2001), based on the data in Brazil in 1998, nd that the collegepremium reached 2.92. Other estimates base on more developed countries such as Kim (2005) shows that the college-premium in Korea ranges from 1.35 1.9. When the economy of South Korea continues to develop, the college-premium decreases from 1.9 to 1.35. The numerical results generated in this section are within this range, from 1.008 to 1.2, depending on the value of parameters. In sum, despite its simplicity, the numerical performance of this model is reasonable well. This provides us some condence to the policy analysis, which will be conducted in the next section. P lease Insert T able 6 Here Similar to the previous sections, a comparison among alternative government policies require a substitution of numbers into the key parameters. The value of key parameters that we are going to use are largely based on empirical estimation from existing literature. P lease Insert T able 7 Here Under education subsidy, when the tax rate rise from 0% to 1.5%, the value of Xe rise from 0.135686 to 0.146 and that of Xu decrease from 0.77887 to 0.772. This pattern is associated with a slightly decrease in robbers (Please refer to Table 7). Note that the introduction of education subsidy attract people to invest in human capital who are originally robbers and unskilled labors. As a result, the marginal product (wage income) for decrease and the interest payment for education loan increase. However, the higher the tax rate under education subsidy resulted in a higher disposable income ratio in the economy as the higher the tax rate, the lower the education cost for each skilled labor has to pay and thus, the higher (lower) the disposable income for skilled (unskilled) labor. Besides, the higher the tax rate under education subsidy, the higher the total production of the economy as the productivity 19

for skilled labors are higher than unskilled labor and the labor participation rate increase accordingly (Please refer to gure 11). On the other hand, the introduction of wage subsidy decrease the value of Xe from 0.135686 to 0.133 and rise that of unskilled labor from 0.77887 to 0.78155 when the tax rate rise from 0% to 1.5% (Please refer to Table 7 and Figure 10). The reason is straightforward. As a result of the introduction of wage subsidy, the expected return for skilled labor decrease while that of unskilled labor increase. The fall in relative income among these two Xe groups resulted in a fall in Xu . Besides, the total production in the economy decrease slightly when the tax rate increases due to the fall in skilled labor and slightly increases in robbery (Please refer to gure 11 and total production is normalized into 1). Comparing the numerical results presented in this section, we know that the introduction of education subsidy and wage subsidy generate opposite eects to wage ratio, equilibrium proportion for dierent groups and the total production for the economy. The introduction of education subsidy widen the income gap between skilled and unskilled labor but rise the total production of the economy, on the other hand, the introduction of wage subsidy narrow down the income gap but lowered the total production of the economy. Such phenomenon may complicate the policy making process if the government consider both the equality level and economy growth in the economy as the performance for dierent policy are counteracting. Thus, the implementation of policy may be very dierent depending on the welfare function that the social planner maximizes. P lease Insert F igure 10 and 11 Here

Conclusions

TO BE CHANGED
This paper attempts to contribute to the literature in two ways. In terms of modelling, this paper extends the framework of Fender and Wang (2003) to include endogenous robbery, as in Huang, Liang and Wang (2004). Interestingly, for a wide range of plausible parameter combinations, we always 20

obtain an unique equilibrium, which makes latter policy analysis much easier in some sense. And although the model is very stylized, our predictions in composition of labor force, college premium, among other statistics, are within the range of previous empirical research. Thus, the framework itself may contain independent interest and worth more attention in the future research. In terms of policy analysis, this paper attempts to revisit the comparison of education subsidy and wage subsidy regimes, which has been studied by Hanushek et. al. (2003). Based on the quantitative results generated, the performance for dierent policies are very dierent. In case of education subsidy, it widen the income gap between skilled and unskilled labor but rise the total production of the economy. If case of wage subsidy, it narrow down the income gap but lowered the total production of the economy. In terms of future research, this model can be extended in many important dimensions. First, the model can allow for endogenous labor-leisure choice. Second, the model can also extend to a multi-community setting and allow for local interactions, whether through the provision of local public goods, or through the externality of social interactions.22 Third, perhaps more importantly, the model can allow for bequest in terms of physical wealth and also skill (see Haunshek et. al, 2004). Some social studies have indicated that there are inter-generational transfer of skill (whether for legal or illegal work).23 Some ongoing projects are being pursued along these directions.

Among others, see Benabou (1993, 1994, 1996a, b, 2000), Benabou and Tirole (2003), and the reference therein. 23 Among others, see Benabou (2002), Benabou and Ok (2001), Benabou and Tirole (2005), and the reference therein.

22

21

References
[1] Acemoglu, Daron; Verdier, Thierry, 1998, Property Rights, Corruption and the Allocation of Talent: A General Equilibrium Approach, Economic Journal, 108, 1381-1403. [2] Acemoglu, Daron; Angrist, Joshua, 2000, How large are Human Capital Externalities? Evidence from Compulsory Schooling Laws, NBER macroeconomics annual 2000, Vol 15, Cambridge and London: MIT Press, 9-59. [3] Axarloglou, Kostas; 2005, What Attracts Foreign Direct Investment Inows in the United States, International Trade Journal, 19(3), 285-308. [4] Barro, Robert J., Jong Wha Lee, 2000, International Data on Educational Attainment: Updates and Implications, Appendix Data Tables, CID Working Paper No. 42. [5] Becker, Gary S., 1962, Investment in human capital: A theoretical analysis, Journal of Political Economy, 70(5), 9-49. [6] Becker, Gary S., 1968, Crime and Punishment: an Economic Approach, Journal of Political Economy, 76(2), 169-217. [7] Benabou, Roland; 1994, Human Capital, Inequality, and Growth: A Local Perspective, European Economic Review, 38(3-4), 817-26. [8] Benabou, Roland; 1996, Heterogeneity, Stratication, and Growth: Macroeconomic Implications of Community Structure and School Finance ; American Economic Review, 86(3), 584-609 [9] Benabou, Roland; 2002, Tax and Education Policy in a HeterogeneousAgent Economy: What Levels of Redistribution Maximize Growth and Eciency?, Econometrica, V. 70, 2, 481-517. [10] Benhabib, Jess; and Mark Spiegel, 2005, Human Capital and Technology Diusion, in Handbook of Economic Growth, volume 1A, P. Aghion and S. Durlauf, eds. (chapter 13). [11] Blom, Andreas; Holm-Nielsen, Lauritz; Verner, Dorte, 2001, Education, Earnings, and Inequality in Brazil 1982-98 Implications for Education Policy, The World Bank Policy Research Working Paper, No. 2686. [12] Boldrin, Michele; 2005, Public Education and Capital Accumulation, Research in Economics, vol. 59, 2, 85-109. 22

[13] Bond, Eric W.; Trask, Kathleen; Wang, Ping; 2003, Factor Accumulation and Trade: Dynamic Comparative Advantage with Endogeneous Physical and Human Capital, International Economic Review, Vol. 44, No. 3, 1041-1060. [14] Brainerd, Elizabeth; 1998, Winners and Losers in Russias Economic Transition, The American Economic Review, 88, 1094-1116. [15] Burdett, Kenneth; Lagos, Ricardo; Wright, Randall; 2004, An On-theJob Search Model of Crime, Inequality, and Unemployment, International Economic Review, v. 45, 3, 681-706. [16] Burnham, Ray; Feinberg, Robert M.; Husted, Thomas A.; ,2004, Central City Crime and Suburban Economic Growth, Applied Economics, V. 36, 9, 917-22. [17] Chang, Hsiao Chuan; 2003, International Trade, Productivity Growth, Education and the Wage Dierential: A Case Study of Taiwan, Journal of Applied Economics, Vol. VI, No. 1, 25-48. [18] Conley, John; Wang, Ping, 2006, Crime and Ethics, Journal of Urban Economics, 60, 107-123. [19] Cooley, T. ed., 1995, Frontiers of Business Cycle Research, Princeton: Princeton University Press. [20] Denny, Kevin; OSullivan, Vincent, 2004, Can Education Compensate for Low Ability? Evidence from British Data, WP04/19, The Institute For Fiscal Studies. [21] Fender, John, Wang, Ping, 2003, Education Policy in a Credit Constrained Economy with Skill Heterogeneity, International Economic Review, Vol. 44, No. 3, 939-964. [22] Ferreira, Sergio Guimaraes, 2004, The Provision of Education and its Impacts on College Premium in Brazil, Revista Brasileira de Economia, Vol. 58, 2, 211-233. [23] Galor, Oded; Moav, Omer; 2004, From Physical to Human Capital Accumulation: Inequality and the Process of Development, Review of Economic Studies, V. 71, 4, 1001-26. [24] Hanushek, Eric A.; 2002, Publicly provided education, Alan Auerbach and Martin Feldstein (ed.), Handbook of Public Economics Vol. 4, Amsterdam: North-Holland, 2045-2141. [25] Hanushek, Eric A.; Javier, Luque, 2003, Eciency and equity in schools around the World, Economics of Education Review, 22, 481-502. 23

[26] Hanushek, Eric A.; Leung, Charles KaYui; Yilmaz, Kuzey, 2003, Redistribution through education and other transfer mechanisms, Journal of Monetary Economics 50, 1719-1750. [27] Hanushek, Eric A.; Leung, Charles KaYui; Yilmaz, Kuzey, 2004, Borrowing constraints, college aid, and intergenerational mobility, NBER Working Paper 10711. [28] Hanushek, Eric A.; Jamison, Eliot; Jamison, Dean, 2006, The Eects of Education Quality on Mortality Decline and Income Growth, Economics of Education Review, forthcoming. [29] Hanushek, Eric A.; Welch, Finis ed.; 2006, Handbook of the Economics of Education, Volume 1, 2, Amsterdam: North-Holland. [30] Huang, Chien Chieh; Laing, Derek; Wang, Ping, 2004, Crime and Poverty: A Search Thoretic Approach, Internation Economic Review, Vol 45, No 3, 909-938. [31] Imrohoroglu, Ayse; Merlo, Antonio; Rupert, Peter; 2000, On the Political Economy of Income Distribution and Crime, International Economic Review, Vol. 41, No. 1, 1-25. [32] Imrohoroglu, Ayse; Merlo, Antonio; Rupert, Peter; 2004, What accounts for the Decline in Crime? International Economic Review, V. 45, 3, 70729. [33] Iwaisako, Tatsuro; 2002, Technology Choice and Patterns of Growth in an Overlapping Generations Model, Journal of Macroeconomics, 24, 211-231. [34] Iyigun, Murat F.; 1999, Public Education and Intergenerational Economic Mobility, International Economic Review, Vol. 40, no. 3, 697-710. [35] Kim, Dae Il; 2005, Growth in College Education and Wage Dierentials in Korea, Seoul Journal of Economics, Vol. 18, 87-123. [36] Klenow, Peter; Andrea Rodriguez-Clare, 2005, Externalities and Growth, in Handbook of Economic Growth, volume 1A, P. Aghion and S. Durlauf, eds., (chapter 11). [37] Lee, Lung Fei; Rosenzweig, Mark R., Pitt Mark M., 1997, The Eects of Improved Nutrition, Sanitation, and Water Quality on Child Health in High Mortality Population, Journal of Econometrics, 77, 209-235. [38] Lee, Nam Chul; 2000, Education and Economic Growth in South Korea 1966-1997, The Journal of Applied Business Research, Vol. 16, No. 4, 83-93. 24

[39] Levantis, Theodore; Gani, Azmat; 2000, Tourism Demand and the Nuisance of Crime, International Journal of Social Economics, v. 27, iss. 7-8-9-10, 959-67. [40] Li, Hongbin, Zhu, Yi; Income, Income Inequality, and Health: Evidence from China, Chinese University of Hong Kong, Working Paper. [41] Lochner, Lance; 2004, Education, Work, and Crime: A Human Capital Approach, International Economic Review, v. 45, 3, 811-43. [42] Miller, Stephen M., Upadhyay, Mukti P.; 2002, Total Factor Productivity, Human Capital and Outward Orientation: Dierences by Stage of Development and Geographic Regions, University of Connecticut Working Paper Series, No. 33. [43] Mincer, Jacob, 1958; Investment in human capital and personal income distribution, Journal of Political Economy 66, no. 4, 281-302. [44] Mehlum, Halvor; Moene, Karl; Torvik, Ragnar; 2005, Crime Induced Poverty Traps, Journal of Development Economics, v. 77, iss. 2, 325-40. [45] Psacharopoulos, George; Patrinos, Harry Anthony; 2004. Return to education: A Further Update, Education Economics, Vol. 12, no. 2, 111-134. [46] Rauch, James E; 1993, Productivity Gains from Geographic Concentration of Human Capital: Evidence from the Cities, Journal of Urban Economics, 34(3): 380-400. [47] Rodrik, Dani, 1999, Where did All the Growth go? External Shocks, Social Conict, and Growth Collapses, Journal of Economic Growth, 4, 385-412. [48] Schultz, T. Paul, 1997, Accessing the Productive Benets of Nutrition and Health: An Integrated Human Capital Approach, Journal of Econometrics, 77, 141-158. [49] Takeuchi, Fumihide, 2005, Cauese of Decline in Labors Share in Japan, JECR Researcher Report, No. 53. [50] Wilson, James; Abrahamse, Allan, 1992, Does crime pay?, Justice Quarterly, 9 (3), 359-377. [51] Wong, Yue Chim Richard; 1995, An Economic Analysis of the Crime Rate in England and Wales, 1857-92, Economica, Vol. 62, No. 246, 235246.

25

APPENDIX

Proof of (10) , (11) and (12)

Under the search theoretic framework, the income for skilled, unskilled labor and robbers are not exact and each agent make their occupational decisions base on expacted disposable income. To derive the expected income for skilled labor, we have: We,t (1Xr,t1 )+(1)We,t Xr,t1 (1+rt1 ) = We,t [1 Xr,t1 ](1+rt1 ), where the left hand side of the equation shows that the expected disposable income for skilled labor equals to the sum of two fragment showing that 1) a skilled labor can hold the total amount of income if he does not meet a robber, and the probability of not meeting robber is exactly (1 Xr,t1 ) as we normalize the total population to unity, and 2) a certain portion of the income, , will be robbed if they meet a robber. As a result, they will be left with the proportion (1 ) and the corresponding probability is Xr,t1 . They alos need to repay the tuition loan. The amount of tuition was , and was paid at time t1, and hence the total debt to be repaid (the interest plus the principal) is (1 + rt1 ), where rt1 is the interest rate on debt carried made at period t1 and to be repaid at period t. Notice also that the robbery that would happen at time t, are conducted by robbers who made their decisions at time t 1, and hence the proportion of robbers is Xr,t1 . By similar method, we can nd out the expected disposable income for unskilled labor: (1 Xr,t1 ) + (1 )Xr,t1 = (1 + rt1 )Wu,t1 [1 Xr,t1 ] . Notice that the unskilled labor does not need to repay any tuition loan. Finally, for robber, the expected disposable income equals to the portion of income robbed from skilled labor if he meets a skilled labor plus the portion of income robbed from unskilled labor if he meets a unskilled labor and thus generating: Wr,t = We,t Xe,t1 + Wu,t Xu,t1 Besides, the expected utility for unskilled labors equals to the sum of expected disposable income and the expected utility for skilled labors and robbers equal to the expected disposable income plus the disutility terms. Base on the above results, we can generate equation (10) , (11) and (12) .

26

Capital Stock with xed real interest rate

rt = A(1 )kt

As mentioned in section 2, the real interest rate is xed and since return to physical capital depends on the marginal product, we have:

and thus,

k=

which is a constant as r is given in this case.

A(1 ) r+

Mathematical derivations of Figure 3

Agent behaviour: Unskilled Labor

To nd out the boundaries for dierent groups in gure 3, we rst consider the case for unskilled labor. Recall equation (10), (11) and (12) , we have: EUe = We,t [1 Xr,t1 ] (1 + rt1 ) and EUu = (1 + rt1 )Wu,t1 [1 Xr,t1 ] EUr = We,t Xe,t1 + Wu,t Xu,t1

For those who choose to be unskilled labor, we must have: U (; , ) = max {EUe (, ), EUu (, ), EUr (, )} = EUu (, ) EUu > EUe and EUu > EUr and since in equilibrium, the marginal agent follows equation (3) and (4), by substitution, we can generate the following two constraints: > and > 27

Those with disutilities satisifying these two conditions would like to be an unskilled labor rather than other occupations and is represented by area EJQL in gure 3a and 3b.

Agent behaviour: Skilled Labor

By similar method, those who want to be skilled labor must follow: U (; , ) = max {EUe (, ), EUu (, ), EUr (, )} = EUe (, ) EUe > EUu and EUe > EUr On the other hand, the equilibrium conditions showing the behaviour of the marginal agents generate the following two conditions: = We,t [1 Xr,t1 ] (1 + rt1 ) (1 + rt1 )Wu,t1 [1 Xr,t1 ] (21) and = We,t Xe,t1 + Wu,t Xu,t1 (1 + rt1 )Wu,t1 [1 Xr,t1 ] (22)

again, by substituting these equilibrium conditions, we can generate the following two constraints: > and > Those who satisies these two conditions would like to be skilled labor rather than robbers or unskilled labor and is represented by area BEJI in gure 3a and 3b.

Agent behaviour: Robber

For those who want to be robber, we have: U (; , ) = max {EUe (, ), EUu (, ), EUr (, )} = EUr (, ) EUr > EUu and EUr > EUe Thus, by substituting these equilibrium conditions, we can generate the following two constraints: < and < Those who satisies these two conditions would like to be robber rather than skilled or unskilled labor and is represented by area MKDEBMC in gure 3a and 3b. 28

In these cases, the situations are dierent from that of the unskilled labor as we do not know the critical values for the disutilities beforehead. Thus, to nd out the area representing skilled labor in the diagram, we rst assume that > 0 and then < 0 and the corresponding area is shown in gure 3a and 3b respectively. Base on the above conditions, we can nd out the equilibrium proportion of skilled labor in the economy by integration and we have: ( R R 1 2 1 dd ( +) 42 2 2 1 R R 4 Xe = 2 1 dd + ( +) 42 42 2 if > 0 , if < 0 if > 0 if < 0

(23)

and thus generating a nonlinear relationship between the proportion of skilled labor (robber) and the critical disutilities for education ( ) and robbery ( )

R R 1 2 1 42 dd + ( +) 42 2 1 R R 1 Xr = 2 42 dd ( +) 42 2

(24)

29