Growth volatility and technical progress: a simple rent-seeking model∗

Charles Ka Yui Leung, Sam Hak Kan Tang and Nicolaas Groenewold October 26, 2005

Abstract Recent empirical evidence demonstrates that a higher level of technical progress is associated with a lower level of growth volatility and higher expected economic growth. This paper builds a simple growth model which combines the insights of Angeletos and Kollintzas (2000) and Tse (2000, 2001, 2002) with endogenous productivity growth and rent-seeking behavior to account for these stylized facts. Our model also complements the literature that focuses on the heterogeneity of different agents. Future research directions are also discussed. JEL classification: E30, O11, O40 keywords: volatility of economic growth, technical progress, rent-seeking, stabilization policy, institution.

∗ Acknowledgement: The authors thank Kai-Sun Kwong, and especially the anonymous referees, the Editor, for helpful comments, Fulbright Foundation (Leung), the Chinese University of Hong Kong Direct Grant for financial support (Tang). The usual disclaimer applies. † Correspondence. Leung: Department of Economics, The Chinese University of Hong Kong, Shatin, NT, Hong Kong, (Phone) (852)-2609-8194; (Fax) (852)-2603-5805; Email:; Tang and Groenewold: Department of Economics, The University of Western Australia, 35 Stirling Highway, Crawley WA 6009, Australia, Email:,,




This paper attempts to develop a simple growth model which relates the volatility of economic growth and technical progress (or total factor productivity growth, or simply TFPG ). Recently, there has been a growing interest in the relationship between the volatility of the aggregate output and other macroeconomic variables.1 For instance, Ramey and Ramey (1995) find that the (average) economic growth rate and the volatility of GDP are negatively correlated. Knack and Keefer (1995) and Acemoglu, Johnson, and Robinson (2001, 2002), among others, find that higher institutional quality has a positive effect on economic growth. Acemoglu et. al. (2003) find further that higher institutional quality helps to reduce growth volatility and mitigate economic crises. De Hak (1999) and Blackburn and Galindev (2003) provide growth models which relate macroeconomic volatility and the rate of economic growth. They do not, however, directly relate the technical progress, economic growth and growth volatility. Aghion and Banerjee (2005) argue that it is the interaction between the entrepreneurs and the imperfect capital market which leads to both the business cycle and economic growth. However, most, if not all, of these papers have not related volatility to the productivity of the economy. Tang (2002) is perhaps the first cross-country study which establishes that the higher the rate of technical progress, the lower will be the volatility of economic growth rate. Tang, however, does not provide a theory for this empirical regularity. This paper attempts to explain the relationship between technical progress, macroeconomic volatility and growth by using a simple endogenous growth model with rent-seeking behavior. The analysis is clearly connected to a growing literature on how the endogenous career choice between productive entrepreneurship and rent-seeking behavior can influence aggregate output. For instance, Baumol (1990) provides many historical examples which show that the returns to rent-seeking can affect the allocation of entrepreneurship. Murphy, Shleifer and Vishny (1991) study how the heterogeneity of agents can interact with the rent-seeking and lead to a lower economic growth rate.2 Acemoglu and Verdier (1998) present a rich model on the endogenous choice of rent-seeking. However, perhaps due to the finite-horizon nature of their model, they do not explore how rent-seeking activities interact with the determination of economic growth.3 Tse (2000) carefully calibrates a two-sector general equilibrium model and finds that GDP can increase by more than 2 times if monopoly is removed from the labor market. In a companion paper, Tse (2002) shows that monopoly in the capital equipment market can also affect employment as well as wages, and hence has important welfare consequences. Tse (2001) shows that the distribution of demand can significantly change the incentives of firms to invest
instance, see Aghion and Saint-Paul (1998) for a survey. addition, Kwong et. al. (2001) provide evidence that the industrial policies in Singapore and South Korea have led to discrimination against small and medium size enterprises, a lack of initiative and government-dependency among the citizens. In the case of South Korea, it has been argued that it even leads to corruption. 3 See Rebelo (1998) for a detailed discussion on the determinants of economic growth.
2 In 1 For


6 Among others. Economists have long been aware of the problem of rent-seeking and the literature which has accumulated is voluminous. Tullock (2003) provides an account of the term “rent-seeking”. they need to be endogenized in the model. to complement the existing contribution and to highlight the rent-seeking activities among firms. Eicher and Garcia-Penalosa eds (2005). we suppress this concern. It follows that firms indeed have incentives to influence the distribution of demand. Lau (1999).7 First. externality effects. 8 This follows a referee’s suggestion which we gratefully acknowledge. Finally. Tollison and Tullock (1980). Nevertheless. Thus.9 A merit of this framework is that it can endogenously derive the relationship between rent-seeking and the economic growth rate within a representative household framework. the role of constitutions and institutions. a robust instance. 5 It 4 For 3 . there are some minor amendments as well.4 The existing literature. we endogenizes the economic growth by introducing a simple R&D technology. which would create some market power for the firms.and innovate. namely. may not only affect the rate of the economic growth. such as the average growth rate and the volatility of GDP. Quah (1996. 9 This follows another referee’s suggestion which we also gratefully acknowledge. 1997). however. footnote 8) for more discussion on the role of heterogeneity in understanding the relationship between rent-seeking and growth. which is important for the convergence result in Angeletos and Kollintzas (2000). the model necessarily belongs to the family of stochastic. 7 To be internally consistent. and the importance of the initial distribution of wealth. is more concerned with the time series implications of a typical stochastic endogenous growth model and how these time series implications can be verified. Instead. see Leung and Quah (1996). but also its volatility. Since both the economic growth rate and its volatility are the subjects of our investigation. see Drazen (2000) and Persson and Tabellini (2000) for a textbook treatments. such as the strategic interactions among potentially rent-seeking entrepreneurs. income distribution and redistribution issues are important in understanding the role of rent-seeking in practice. endogenous growth models. this paper abstracts from the capital accumulation. while this paper focuses on how different institutions could affect the (stochastic) time series properties of a growth model.8 Since this paper focuses on the mean and variance of the economic growth rate.6 The present paper builds on the work of Angeletos and Kollintzas (2000) but modifies their model in two important ways. resource allocation consideration which leads to the emergence of rent-seeking. thus adversely affecting capital accumulation and long-run economic growth. Certainly. this simple model abstracts from these factors and focuses merely on the simple. Among others. 1 0 See Angeletos and Kollintzas (2000. all of which would affect the aggregate economic outcomes. the heterogeneity of agents.5 Recent efforts have explored several important issues. This paper takes a preliminary step to show how rent-seeking activities. is beyond the scope of this paper to survey that literature. Tse (2004) shows how informational friction can create market power for firms. Tullock (1993). see Buchanan.10 It should be noticed that the significance of this research may go beyond establishing and explaining another stylized fact in growth.

negative relationship between TFPG and growth volatility. Groenewold and Leung (2003) (henceforth TGL). and at the same time negatively correlated with the volatility of economic growth rate. The paper will provide conditions under which (endogenous) technical progress is positively correlated to the rate of economic growth. 1 3 There has been some discussion in the literature of the diff erence between the private and social return of R&D. It is beyond the scope of this paper to survey that literature and interested readers may consult Stock and Watson (1999). they find that better institutional quality accelerates technical change.15 1 1 Throughout this paper. one of the measure of the TFPG can only be calculated for 56 countries. if we regard growth fluctuation as a “bad” thing for the economy. al. the point estimates of TFPG contribution range from −0.13 The organization of this paper is simple. The next section presents some stylized facts.12 Consequently. Building on the work of Tang (2002).14 a robust negative relationship between growth volatility and TFPG is found. which in turn reduces growth volatility and mitigates economic crises. Depending on the measure of TFPG being used. Their results are robust to a large number of alternative specifications and controls for simultaneity. since these subsidies promote TPPG and therefore help to “stabilize” growth fluctuations.75. has been stablized.S. crises and growth. Tang (2002) presents perhaps the first systematic analysis of the relationship between growth volatility and technical progress. and whether the government policies have been effective in this respect. see Jones and Williams (1998. Using different samples and control variables. by improving TFPG of the overall economy. 1 5 See TGL for details. (2003). 2 and table 1 provide a summary of their empirical results. Figures 1. and data from 98 different countries. initial human capital stock and initial productivity gap are controlled for. 1 2 There is a debate on whether the post-war aggregate fl uctuations of the U. 1 4 Due to the data availability.11 Specifically. Among others. Using four different measures of TFPG.A. 2 Some Stylized Facts There is only limited empirical work on the relationship between growth fluctuations and technical progress. 2000). the term "productivity" and "technological progress" will be used interchangeably. Then a simple stochastic. dynamic general equilibrium model is built. all statistically significant. Detailed proofs are provided in the appendix. The relationship holds even when the initial per worker GDP.33 to −0. and this section draws heavily on Tang (2002) and Tang. it may provide an additional justification for various forms of government subsidies to R&D. 4 . TGL evaluate the role of technical change in affecting the macroeconomic volatility. then it may be a reasonable government objective to eliminate at least part of this “bad” thing. We summarize and conclude in the last section. and following the methodology of Acemoglu et.

Clearly. a typical agent z . Lx t = N X z =1 x lt (z )/N. We will discuss the government budget constraint later. Time is discrete and the horizon is infinite. The output of agent z only depends y on the productive effort of agent z .16 The representative agent maximizes (1) subject to the budget constraint and the time constraint ct (z ) ≤ St (z )yt (z ). if θ2 = 0.(Figure 1. lt (z ) > 0. 2 and Table 1 about here) 3 A Baseline Model where 0 < β < 1. lt (z ). the utility function takes the log form: u (cs (z )) = ln (cs (z )) . Each agent operates their own production technology. (4) where Lx t is the average rent-seeking effort of the economy at time t. ct (z ) is the consumption of the agent z at period t. rent-seeking activities. and yt (z ) is the output of agent z . (6) yt (z ) = At (lt 1 6 In the current setting. discounted sum of utility ∞ X Et β s−t u (cs (z )) . then individual rent-seeking efforts are not effective at all. and R&D respectively. and will be determined at the equilibrium. lt (z ) are the fraction of time employed in productive. lt (z ). The economy is populated by a large number N of identical infinitelived agents. The production side is very simple here. θ y (z )) 1 . θ2 x St (z ) = (lt (z )/Lx t) . For simplicity. (2) (3) This section presents a simple growth model. N . lt (z ). 5 . then agent z is receiving y x d subsidy (being taxed) in net terms. lt (z ). we follow Angeletos and Kollintzas (2000) in assuming that the net amount of subsidy depends on the rent seeking effort of agent z relative to the average rent-seeking effort in the economy. if St (z ) is larger (smaller) than unity.. (5) Lx t is taken as given in agent z ’s maximization problem. maximizes the expected. lt (z ). To highlight the importance of rent-seeking activities. 2.. . relaxing this assumption will only complicate the algebra without any additional insights.. Clearly. We assume that 0 < θ2 < 1. At . (1) s=t where St (z ) is the subsidy factor for agent z . which are y x d all restricted to be positive. θ2 is a parameter to measure the importance of rent-seeking efforts in determining the government subsidy (or tax). z = 1. In each period. and the aggregate productivity at time t. y x d lt (z ) + lt (z ) + lt (z ) ≤ 1.

whose distribution takes the following form. lt = lt (1) . which is to pin down the R&D effort. ∀z . The first step is essentially static and it is easy to show that first order conditions imply y x lt (z ) = (θ2 /θ1 ) lt (z ). 0 < p < 1. p is an exogenously given probability. γ t ≡ At+1 /At . where γ t is the growth rate of productivity between period t and t + 1. Notice that while the probabilities of different "states of Nature" are exogenous. see Kahn (1990). Furthermore.. where φ = θθ 1 θ 2 (θ 1 + θ 2 ) Now we move to the second step. the aggregate productivity growth rate is a random process. Leung (2001).. we take the R&D effort lt (z ) as given. → − d Γg > 1/β > 1. which is to determine the optimal time allocated to R&D activities. To solve the individual maximization problem. θ1 + θ2 (8) −(θ1 +θ2 ) 1 θ2 > 0. (9) d with probability 1 − p Γb lt Thus. we proceed in two steps. the net output of agent z is St (z )yt (z ). and the natural restriction that the growth rate of a good state is higher than the bad state implies that Γg > Γb > 0. We take this optimal allocation into the second step. Since R&D affects future rather than current productivity.. the levels of the realization are endogenous.. we have ∂ Γb lt /∂lt (z ) > 0.. Formally. the maximization problem is necessarily dynamic in nature. and that aggregate productivity growth depends on the R&D efforts of all agents. and lt is the vec→ − ¡ ¢ d d d d tor of all the individuals’ efforts in R&D. lt (N ) . ¡ ¢θ1 +θ2 x −θ2 d St (z )yt (z ) = At φ 1 − lt (z ) (Lt ) . we assume that Γb = Γg 1 7 For z =1 a comparison of the moral hazard type model. We further impose the restriction that θ1 + θ2 < 1. to determine the optimal R&D effort. and only determine the allocation between production and rent-seeking activity. it is necessary to know the marginal return of investing in R&D. d First.17 To capture the idea that the level of ³− ´ → d d "bad state" is increasing in any individual’s effort. we assume that the research effort for each individual is observable to all other agents. 1 − lt θ1 + θ2 ¶ µ ¡ ¢ θ2 x d lt (z ) = 1 − lt (z ) . For simplicity. N Y ¡ d ¢1/N lt (z ) . ( Γg ³ ´ with probability p → − γt = ... and Γg is a parameter describing the growth rate in a "good state". lt (z ) . (7) Substituting this into (3).with 0 < θ1 < 1. we have µ ¶ ¢ ¡ θ1 y d lt (z ) = (z ) . (10) 6 .

+1 t+1 (11) → − d (−z ) is the vector of research efforts for all agents except agent z where lt → − ¡d ¢ d d d d (−z ) ≡ lt (1) . ∀z . 7 . lt ( − z ) . we will restrict our attention to the time-invariant labor allocation equilibrium. lt (−z ) . by (8)... are functions of parameter and other agents’ R&D efforts. d x lt (z ) = ld (z ) . Notice that at period t. lt = V A t t ¡ ¢ = a0 (z ) + a1 ln At + a2 ln 1 − ld (z ) + a3 ln ld (z ) + a4 ln Lx . .. Definition 1 In a time-invariant labor allocation equilibrium. Clearly. the desired conditions are all satisfied. The proposition is basically proved by substituting the conjecture into equation (11). all proofs can be found in the appendix): 1 8 We. it is then straightforward to compute the individual optimal R&D effort (again. lt (z ) = ly (z ). and other constraints) can be written as µ ¶ ³ → − ¡ ¢θ1 +θ2 x −θ2 ´ d d V At .. Lx ln A φ 1 − l ( z ) (Lt ) = max t t t d (z ) lt ∙ µ ¶ − − → d x +β p · V Γg At . and hence the time-invariant condition may not be as restrictive as it y x (z ) = lx (z ).. (11) is very difficult to solve in general. ∀z . This timeseems. lt with probability p. Given all these assumptions. lt invariant condition (12) leads to several results. L V At . µ ¶ µ ¶ → − → − d d x (−z ) . In this paper. we can characterize the value function. Lx t = L . L +1 t+1 µ µ ¶ ¶¸ − − → → − d d x + (1 − p) V Γb lt ( − z ) . ∀z . (13) where a1 ... ∀t.. however.e. a4 < 0.18 Consequently. period t + 1 productivity will increase to At+1 = Γg At . as described by (9). (3). lt (N ) . Lx . subject to (2).i. In addition. Given this explicit functional form.. maximizing (1). this formulation will prove to be very tractable. it is easy to see that the individual’s dynamic optimization problem (i. First. (12) Notice that there is no capital in this model and the productivity shock is i. lt (z − 1) . L At .d Since 1 > lt (z ) > 0.. cannot prove that it is the only equilibrium in this model. and is detailed in the appendix. for expositional purposes. lt . lt (z + 1) . l ( − z ) . a3 > 0. Proposition 2 The value function takes a simple log-linear form. ∀t.d.

Thus. it is analogous to the Prisoner’s Dilemma. And since we have already solved for the optimal ld in (14). by (8) and (12). θ1 + θ2 + βa2 + β (a3 + a1 (1 − p) (1/N )) (14) Clearly. given the participation of other agents. we have the following result. everyone consumes his/her own production. to close the model. ∀t. since ld (z ) = ld . γ t ≡ At+1 /At . ∀z . by (8). where every agent is worse off under "the rent-seeking game". Corollary 4 ld (z ) = ld . and is hence identical across different agents. by (4) and (5).Lemma 3 The optimal individual level R&D effort is given by the following expression. ∀z. Thus. Thus. 0 < ld (z ) < 1. the government budget will be trivially balanced. as the term (1 − St (z )) = 0. Corollary 6 Lx = lx (z ). we do not need to impose. by (9). In other words. although no individual has the incentive to exit from it. Corollary 5 Γb = Γg ld < Γg as ld < 1. and St (z ) = 1. l d (z ) = β (a3 + a1 (1 − p) (1/N )) . The rent-seeking activities are also equalized across agents at the equilibrium. we know the distribution of the productivity growth. even though there is no net transfer of consumption goods across agents. 8 . (6) is simplified as yt (z ) = At (ly (z ))θ1 . we only need to dictate the value of the average rent-seeking activities Lx and the government subsidy St (z ). Also notice that the value of ld (z ) does not depend on a0 (z ). there is no net subsidy across agents. ∀t. ct (z ) = St (z )yt (z ) = yt (z ). but rather derive symmetry under the time-invariant labor allocation equilibrium. Now. there is a non-trivial amount of labor efforts spent in rent-seeking activities. ∀z . (15) with the value of ly (z ) being pinned down by (8). lx (z ) = lx . ∀t. ∀t. ∀z . Notice however that. However. lx (z ) = lx > 0. Thus. substituting this fact into (10) delivers the following result. Now. ∀z . N X z =1 [(1 − St (z )) yt (z )] = 0. And in that case. Thus.

(20) ∂θ1 ∂θ2 ∂p ∂N 1 9 See Quah (1996. t ¡ ¢ ¡ £ ¢¤2 V ar γ Y . < 0. This result begs the question of why R&D effort differs across countries. this model reproduces the stylized facts discussed in the stylized facts section: Proposition 9 The mean productivity growth and mean economic growth rate are negatively correlated with the variance of the economic growth rate. Thus. ∀t. 1997) for a formal defi nition.Now. PN The aggregate output of this economy is Yt ≡ z =1 yt (z ) = N yt . With these definitions. where y θ1 Y yt = At (l ) . (ii) the mean economic growth rate is increasing in R&D effort. we have already shown that labor allocation is identical across agents. it is then easy to prove the following results: Proposition 8 In this economy. and that there is no net transfer of consumption goods across agents. it is easy to see that both mean economic growth and mean technical progress are increasing in R&D efforts. In particular. The appendix shows that (14) can be simplified as ld = β (1 − β )−1 (1 − p) (1/N ) −1 (θ1 + θ2 ) + β (1 − β ) (1 − p) (1/N ) . ¡ ¢ £ ¤ E γY = Γg p + (1 − p) ld . 9 . < 0. Corollary 7 yt (z ) = yt . γ t ≡ Yt+1 /Yt . ∀t. ∀z . as ld (z ) = ld . we are ready to study the stationary equilibrium growth path. (19) which implies the following result: Lemma 10 The equilibrium R&D effort decreases with the following parameters: ∂ld ∂ld ∂ld ∂ld < 0. γY (16) t = γ t . < 0. (18) Equipped with these results. And we denote the economic growth rate by γ Y t .19 By restricting our attention to the time-invariant labor allocation equilibrium. by (15). and that the variance of the economic growth rate is decreasing in R&D efforts. we deduce that output is identical across agents. = p (1 − p) (Γg ) 1 − ld t (17) (iii) the variance of the economic growth rate is decreasing in R&D effort. ly (z ) = ly by (8). Thus. (i) the economic growth rate is equal to the productivity growth rate.

If the marginal return to either goods production (governed by θ1 ) or rent-seeking activities (governed by θ2 ) falls. will R&D effort. 2 1 See. which means a decline in θ2 . can be interpreted as an "institution parameter". and so. this paper builds a stochastic. production activities and R&D efforts all being endogenized. which reduces growth volatility and increases the (average) economic growth rate. therefore. many simplifications have been made to keep the model tractable. If the probability of realizing a good state of nature (p) increases. If the number of research agents (N ) increases. see Hanushek and Welch (2005). Jones and Williams (1998. the subsidy can be in the form of a tax credit. And this mechanism exactly match the empirical evidence provided by Tang (2002) and Tang. and a reduction of growth volatility is typically interpreted as a “good” thing for society. among others. the R&D effort ld will increase (by (20)). For instance. then it is rational for the individual to "free-ride" on the others and hence the R&D effort will also decrease. And if the institution improves. that better institutional quality accelerates technical progress. and the variance of the economic growth rate will decrease. or a subsidy to higher education.” This task is achieved by pointing to the stylized facts. With rent-seeking behavior. the representative agent assumption suppresses the possibility of a liquidity constraint.21 This paper provides a supplementary justification: a higher level of technical progress indeed decreases growth volatility. then the importance of R&D efforts decreases. endogenous growth model which successfully mimic these stylized facts. If the accumulation of physical capital is allowed. In other words. And by (16). Among others. such as abstracting from the accumulation of physical capital and the heterogeneity of agents.The intuition is very clear. Relaxing these assumptions can lead to a much more realistic model and much richer aggregate dynamics through the evolution of the income distribution. then it is natural to re-allocate the labor input to those activities and hence decrease the R&D effort. (17) and (18). this paper claims that technical progress contributes to “output stabilization. Also.20 And a common justification for such subsidies is that the social return to research exceeds its private counterpart and it is the government’s duty to correct this market failure by providing subsidization. 2000) for a review of the literature and evidence. Future research can extend the analysis in several directions. Notice that the parameter θ2 which is a proxy for the importance of rentseeking activities in shaping the net transfer of output from the government. we know that the expected rate of technical progress and expected rate of economic growth will increase. and instance. 2 0 For 10 . 4 Concluding Remarks Research is typically subsidized. Groenewold and Leung (2003).

as demonstrated by Acemoglu (1996. liquidity constrained agents may be forced to allocate their talents to the latter. however.22 The two factors. there is a large debate on the relative roles of technology and human capital accumulation in driving economic growth which is too large to be reviewed here. 2 2 However. One can consider a framework in which these activities become choice variables and hence introduce a much richer structure of the economy. and suppress the importance of human capital. See. 1997. Barro (2001) and Prescott (2002). 1998). Thus. 11 . thereby affecting the income distribution and the aggregate dynamics. which is also essential to rent-seeking activities. The transitional dynamics in that situation can be very rich.yet the fixed cost of starting a business or research is much high than that of rent-seeking activities. can be complementary rather than substitute for each other. and also lobbies the government. the paper stresses the role of technical progress in promoting economic growth. In addition. it is assumed that each agent produces. In the model. one can follow Tse (2004) in explicitly modelling the informational frictions and to provide a much deep micro-structure of the market power and rent-seeking activities. the exchange between Bils and Klenow (2000) and Hanushek and Kimko (2000). Alternatively. carries out R&D. among others.23 Notice that the accumulation of human capital will demand a time input. 2 3 See also Leung (1995). All these possibilities are remained to be explored. One can also follow Eicher and Garcia-Penalosa (2005) to distinguish between “predatory lawyers” and “institution-building lawyers” and to endogenously determine the degree of property right protection and hence the rate of technical progress. which is part of our ongoing research agenda. introducing human capital accumulation into the current framework will lead to much richer dynamics.

279-303. 1055-1089. 64. (1998). D.” Labour.” CEPR Discussion Paper 2464. 111.References [1] Acemoglu. [13] Baumol. 117(4). Robinson (2002). (1997). [11] Backus. and destructive. R. [4] Acemoglu. 98. [8] Aghion. Tullock (eds. Thaicharoen (2003). Corruption and the Allocation of Talent: A General Equilibrium Approach. and P. D. D. D. D. “Why do new technologies complement skills? directed technical change and wage inequality. volatility and learning.. P.” Journal of Monetary Economics.” Review of Economic Studies.-M. [3] Acemoglu.. Johnson and J. Volatility and Growth. 893-921. Kehoe (1992). K. macroeconomic symptoms: volatility. Oxford: Oxford University Press.” American Economic Review. 91. S.” Economics Letters.. “Institutional causes. 108(450). 417-421. 12 . 49-123. TX: Texas A&M University Press. (1996). 377-409. “Growth. 1369-1401.). W. Johnson. [14] Bils.” American Economic Review. [12] Barro. “Entrepreneurship: productive. College Station. P. Robinson and Y. D. Robinson (2001). “Property Rights. J. (1980) Towards a Theory of the Rent-Seeking Society. 1381-1403. Saint-Paul (1998). Kehoe and T. 779804.. “Does Schooling Cause Growth?” American Economic Review. M.” Quarterly Journal of Economics. [9] Aghion. 79. (1990). 12. [6] Acemoglu. [5] Acemoglu. J. G.” Journal of Economic Theory. 91. and R.” Economic Journal. R.. 50.M. “Reversal of Fortune: geography and institutions in the making of the modern world income distribution.” Quarterly Journal of Economics. 1231-94. “Rent seeking/corruption and growth: a simple model. Klenow (2000). 90(5). 445-464.” Quarterly Journal of Economics. D. “Training and innovation in an imperfect labor market. and T. “Human Capital and Growth. “A microfoundation for social increasing returns in human capital accumulation. 12-17. and T.D. S. 58. S. Johnson and J. unproductive. [2] Acemoglu. 113.” Journal of Political Economy. [15] Blackburn. 1160-83. [7] Acemoglu. crises and growth. Verdier (1998). Kollintzas (2000). Galindev (2003). [10] Angeletos. Tollison and G. D. “In Search of Scale Effects in Trade and Growth. and G. (2001). “Uncovering some casual relationships between productivity growth and the structure of economic fluctuations: a tentative survey. [16] Buchanan. “Colonial origins of comparative development: an empirical investigation. and A. P. Banerjee (2005).

Technology Adoption and Growth. Labor-Force Quality. [21] Hanushek. and the behavior of asset returns. [22] Hanushek.. 395-408. K. 24. [29] Lau. and J. K. E. L. F.-S. 90.” in Robert Solow and Luigi Pasinetti (ed. forthcoming. 173-177. and T. [30] Leung.” Journal of Monetary Economics.). 1119-35.[17] De Hek. endogenous growth. Amsterdam: North Holland. [20] Eicher. Durlauf ed. T. [24] Jones. and A. [19] Easterly. [25] Kahn. and productivity disturbances. “On endogenous growth under uncertainty. Rodriguez-Clare (2004). Chau. 5. Lui and L. “I(0) in. Recursive Macroeconomic Theory. New York: North Holland/Elsevier. “Too Much of a Good Thing? The Economics of Investment in R&D.” International Economic Review.” Economics and Politics. 93. Cambridge: MIT Press. C." forthcoming in P. P. S. [26] Klenow. A. K. 38. [32] Leung. [23] Jones. C. “Measuring the Social Return to R&D. (1990). (1999).-C. 7.” Economics Letters. 48. C. 46. Williams (2000). Cambridge: MIT Press. S. 1184-1208. 124. Rebelo. R. “Institutions and economic performance: cross-country tests using alternative measures. [18] Drazen. “Moral hazard. L. 727-744. Sargent (2000). (2000). Handbook of Economics of Education. Y. C.” Journal of Econometrics. and C. “Educated guesses and income distribution. 207-227. “Convergence. Taiwan and South Korea. (1995). Welch eds. Qiu (2001). forthcoming. 27-44. E. Economic Growth and the Structure of Long Term Development. 535547. [27] Knack. 40. 65-85. King. T.-H. Industrial Development in Singapore. "Externality and growth. Kimko (2000). Garcia-Penalosa eds (2005) Institutions for Development and Growth. P. Keefer (1995). “Policy. J. and D. Princeton: Princeton University Press. Quah (1996). K. integration and cointegration out: time series properties of endogenous growth models. [31] Leung. 113. Y.. Handbook of Economic Growth. Williams (1998).” Journal of Monetary Economics. P. increasing income inequality and social insurance: the case of China?” Journal of Economic Behavior and Organization.. imperfect risk-sharing. New York: International Economic Association. R. and D. Aghion and S. and J.” American Economic Review. (2005). and the Growth of Nations. W. 1993. [33] Ljungqvist. [28] Kwong.” Quarterly Journal of Economics. (1999). C. Y. Political Economy in Macroeconomics. Singapore: World Scientific.” Journal of Economic Growth. and F. “Schooling. Levine and S. and P. 13 . “Productivity growth. (2001).

[40] Rebelo. Y. H. (1996) “Convergence Empirics across Economies with (Some) Capital Mobility.[34] Murphy. “On the Determinants of Economic Growth. 1. 99. 335-341. and G. C. D. A. Leung (2003). C. Y. (2002). 500-521. Rent Seeking. “Long run policy analysis and long-run growth. [35] Persson. “The link between growth volatility and technical progress: cross-country evidence. (2001). Cambridge: MIT Press. 14 .” Journal of Economics (Zeitschrift-furnationalokonomie). S. “Business Cycle Fluctuations in US Macroeconomic Time Series. New York: NorthHolland. Groenewold and C. D. (1985). [44] Tang. [42] Rogerson. [39] Ramey. and M. 92. 95-124. Tabellini (2000). [43] Stock. 1357-67.” American Economic Review. “The first-order approach to principal-agent problems. 503-530. S. “Richard T. “The allocation of talent: implications for growth.” in Contemporary Economic Issues: Proceedings of the Eleventh World Congress of the International Economic Association. S.” Labour Economics. 138-56. “Search frictions. (2002). 137-174. N.. E. K. G. “The distribution of demand. “Cross-country evidence on the links between volatility and growth. (2004). [50] Tullock. 77. “Monopoly. and investment in technology. Elgar. [36] Prescott. K. K.. 275-297. 9(5). (1993). [38] Quah. Polarization.. Political Economics: Explaining Economic Policy. 53.” Quarterly Journal of Economics. 3-64. [49] Tse. H. 27-59. (1998). and long-run growth. “Monopoly. Michael Woodford eds. (1991). Y. Shleifer and R.” Journal of Political Economy. [48] Tse. Chinese University of Hong Kong. Ramey (1995). (2000). [45] Tang. Crises and Growth: A Robust Causation” mimeo. W. C. (2002). [37] Quah. T. Handbook of Macroeconomics. [41] Rebelo. Vishny (1991). 85. 106(2). 323-346. market structure. 681-697. 1138-1151. K. 73 (3). Ely Lecture: Prosperity and Depression.” Journal of Economic Theory. 2. Brookfield: E. G. C. Y. Volume 1A.” Economics Letters.” Econometrica. “Institutions.” in John Taylor. 61 (1). human capital accumulation and development. (1997) “Empirics for Growth and Distribution: Stratification. market power. and V. Y. T.” American Economic Review. Watson (1999). London: Macmillan Press. T. 1-15. J. [47] Tse. S.” Journal of Development Economics.” Journal of Economic Growth.” Journal of Economic Growth. and Convergence Clubs. employment and wages. Technical Change and Macroeconomic Volatility. 116(2). [46] Tse.

G.[51] Tullock. “The origin rent-seeking concept. 2(1). 15 .” International Journal of Business and Economics. 1-8. (2003).

a0 (z ) + ¡ hand ¢ is basically d d a1 ln A + a ln 1 − l ( z ) + a ln l ( z ) + a4 ln Lx . we have o P d a0 (z ) = (1 − β )−1 {ln φ +βa1 ln Γg +β (1 − p) a1 (1/N ) N ln l ( j ) . ³ ´ ³ ´ γ t ≡ Yt+1 /Yt = (N yt+1 ) / (N yt ) = yt+1 /yt = y θ1 y θ1 At+1 (l ) / At (l ) = (At+1 ) / (At ) = γ t . it will be available upon request) A A. where yt = At (l ) . Again.2 Proof of (14) The proof is straightforward. Yt ≡ z =1 yt (z ) = N yt . We substitute the functional form for V into equation (11).j 6=z and a1 a2 a3 a4 = (1 − β )−1 . a1 . d d 1 − l (z ) N l (z ) Simply re-arranging terms will yield (14). A. (18) PN y θ1 y y By definition. and differentiate with respect to ld (z ). Thus. which is (16). a3 > 0. by simply comparing coefficients on different terms.1 Proofs Proof of (13) be re-written as +a3 ln ld (z ) +a4 ln Lx ) . = (1 − β ) = (1 − β ) = (1 − β ) −1 −2 −1 The proposition is basically proved by substituting the conjecture into equation (11). a4 < 0. A. Y ∀z. (21) and clearly. β (1 − p) (1/N ) .Appendix (If this appendix is not published. by definition. .. which ¡ can d ¢ ¡ ¢ ln At + ln φ + (θ1 + θ2 ) ln 1 − lt (z ) −θ2 ln Lx +β (a0 (z ) +a2 ln 1 − ld (z ) PN +a3 ln ld (z ) +a4 ln Lx ) +βa1 ln At +βa1 ln Γg +β (1 − p) a1 (1/N ) z=1 ln ld (z ) .3 Proof of (16). j =1. and the left side the conjectured functional form... 16 . l (z ) = l . (17). to get ∙ ¸ 1 a1 (1 − p) 1 [(θ1 + θ2 ) + βa2 ] = β a3 + . (−θ2 ) . This confirms the initial conjecture. and the right hand side is t 2 3 ³ ´ ¡ ¢ ¡ ¢ θ1 +θ2 −θ ln At φ 1 − ld (z ) (Lx ) 2 +βp (a0 (z ) +a1 ln (Γg At ) +a2 ln 1 − ld (z ) ! Ã N Y ¢ ¡ ¢ ¡ 1 /N At +a2 ln 1 − ld (z ) ld (z ) +a3 ln ld (z ) +a4 ln Lx ) +β (1 − p) (a0 (z ) +a1 ln Γg z =1 (θ1 + θ2 ) .

Notice that the economic growth rate is an i. A. which is (18). We substitute (21) into (14) to get (19). ¡ ¢ E γY = E (γ t ) = pΓg + (1 − p) Γb t = pΓg + (1 − p) Γg ld . 17 . as Γb = Γg ld £ ¤ = Γg p + (1 − p) ld . The variance of the economic growth rate is also easy to calculate. process and hence the conditional mean and unconditional mean coincide and is given by. A.i.d.5 Proof of (20) Simply differentiate the expression for ld in (19) with respect to different parameters to deliver (20). ¡ ¢ ¡ Y ¢¤2 £ V ar γ Y = E γY t t − E γt £ £ ¤¤2 = p · Γg − Γg p + (1 − p) ld £ £ ¤¤2 + (1 − p) · Γb − Γg p + (1 − p) ld £ ¡ ¢¤2 = p · (Γg ) (1 − p) 1 − ld £ ¡ ¢¤2 + (1 − p) · (Γg ) (p) 1 − ld £ ¡ ¢¤2 = p (1 − p) (Γg ) 1 − ld . which is (17).4 Proof of (19) The proof is simple.

04 Total Factor Productivity Growth .08 -0.02 0.Figure 1: Economic Volatility and Technical Change Economic Volatility and Technical Change 25 Standard Deviation of Annual GDP Per Capita Growth 20 15 10 5 0 -0.02 0 0.06 -0.04 -0.

08 -0.04 -0.02 -2 -4 -6 Total Factor Productivity Growth 0 0.02 0.Figure 2: Economic Growth and Technical Change Economic Growth and Technical Change 8 6 4 2 0 -0.04 Average Annual GDP Per Capita Growth .06 -0.

32) -0.24 60 0.29 (0.89) Equation Equation Equation Equation Equation Equation (2) (3) (4) (5) (6) (7) Dependent variable is standard deviation of annual GDP per capita growth rate -0.88) (-2.60) (-1.76) -69.75) -77.49) 0.15 (-0.05 0.28) 0.30) Equation (8) -0.61) (0. Openness and Financial Development) Equation (1) Initial Constraint on Executive TFPG Africa Asia Latin America Latitude Coastal Area Mean Temperature Human Capital Openness Financial Development Observations R-Square -0.65** -83.31) Equation (9) -0.83** (-2.34 58 0.52) (-1.Table 1: Determinants of Macroeconomic Volatility (Robustness to Regional Dummy.66) 48 0.58 (-0.27 (-0.06 -0.28 -0.46 (-1.38 50 0.30** -72.22) (-1.73* (-2.36 -0.29 .37 59 0.26) -81.17) (0.06) -0.01 (1.11) (0.00** (-1.25 -0.37) (-1.49) (0.37) (-1.39 (-0.90 -0.73) (-3.29) 0.11* (-2.06) -73.28 -0.21) -0.02) -0.35 60 0.35 57 0.14) (0.65 0.40 60 0.08 0.72) (-2.28* (-1.19) 0.96) (-2. Geography.34 60 0.19* -71.25 (0.98 (0.02* -75.24 (-1. Human Capital.11) 0.27* (-1.31 (0.25 -0.90** -79.59) (-1.07) (-0. Weather.61) (-0.

score of 3 indicates slight to moderate limitations. Figures are averages for the period 1970-90 taken from International Financial Statistics (IFS) lines 32d/line 99b. 9. 0 otherwise.The World Fact Book 2002. which is a measure of distance from the equator. Data are obtained from Acemoglu. Data are available from Penn World Table Mark 5. German and Scandinavian) and the logarithm of real GDP per capita in 1970. Johnson and Robinson (2001). French. respectively. scaled to take values between 0 and 1. Africa: Dummy variable taking value 1 if a country belongs to Africa. Latitude: Absolute value of the latitude of the country. Financial development: Natural logarithm of value of credits by financial intermediaries (banks and non-banks) to the private sector divided by GDP. From McArthur and Sachs (2001). This variable is instrumented by the mortality rate of European settler in the ex-colonies. 1996. TFPG is instrumented by legal system (English. . TFPG is total factor productivity growth over the period 1970-90. 2. 8. with higher scores indicating more constraints. Scores of 2. Asia: Dummy variable taking value 1 if a country belongs to Asia. It has a scale from 1 to 7. We average the scores for 1950.6. Data are available from Polity III dataset compiled by Keith Jaggers and Ted Robert Gurr. 5. 4 and 6 indicate intermediate values. for a one-tailed test. Inter-University Consortium for Political and Social research.Notes: 1. 7. 0 otherwise. ** and * indicate statistical significance at one or five percent level. score of 7 indicates executive parity or subordination. Openness: The average ratio of nominal imports plus exports to GDP in Purchasing-Power-Parity US dollars (PPP GDP) for the period 1970-90. which is taken from Tang (2002). Latin America: Dummy variable taking value 1 if a country belongs to Latin America or the Caribbean. See La Porta et al. where 0 is the equator. 1960 and 1970 to obtain the average initial constraint. score of 5 indicates substantial limitations. Data are available from Barro and Lee (1993). 13. Coastal area: Proportion of land area within 100 km of the seacoast. Human capital: Natural logarithm of one plus the average years of schooling for the period 1970-90. 4. 10. 0 otherwise. Initial constraint on executive measures the institutional and other constraints that are placed on presidents and dictators. Macroeconomic volatility is measured by the standard deviation of annual real GDP per capita growth rate for the period 1970-98. 3. 11. Score of 1 indicates unlimited authority. 12. Data for legal system are taken from La Porta. This TFPG is estimated using a cross-country regression with the growth of output per worker as the dependent variable and a constant and the growth of physical capital stock per worker as the independent variables. 6. Mean temperature: 1987 mean annual temperature in degrees Celsius from McArthur and Sachs (2001). Lopez-de-Silanes and Shleifer (1998) and the website of CIA. Parentheses contain t-ratios. (1999).