نیرومندی، محدودیت های پردازش اطلاعات، و حساب جاری در اقتصاد باز کوچک

In this paper we examine the effects of two types of “induced uncertainty”, model uncertainty due to robustness (RB) and state uncertainty due to finite information–processing capacity (called rational inattention or RI), on consumption and the current account. We show that the combination of RB and RI improves the model's predictions for (i) the contemporaneous correlation between the current account and income and (ii) the volatility and persistence of the current account in small open emerging and developed economies. In addition, we show that the two informational frictions improve the model's ability to match the impulse response of consumption to income and the relative volatility of consumption to income growth.

Current account models following the intertemporal approach feature a prominent role for the behavior of aggregate consumption (see Sachs 1981). For given total income, consumption is the main determinant of national saving, and the balance of national saving in excess of investment is the major component of the current account. This important role for consumption has naturally led researchers to study current account dynamics using consumption models.1 For example, the standard intertemporal current account (ICA) model is based on the standard linear–quadratic permanent income hypothesis (LQ–PIH) model proposed by Hall (1978) under the assumption of rational expectations (RE). Within the PIH framework, agents can borrow in the international capital market and optimal consumption is determined by permanent income rather than current income; consequently, permanent income also matters for the current account. For example, consumption only partly adjusts to temporary adverse income shocks, which makes the current account tend to be in deficit. In contrast, consumption fully adjusts to permanent income shocks, with little impact on the current account. However, many empirical studies show that the standard RE–ICA models are often rejected in the post-war data.2 In addition, the standard models also cannot explain the different behavior of the current account and consumption in emerging and developed countries.3 It is not surprising that the standard RE–ICA models are rejected because the underlying standard PIH models have encountered their own well-known empirical difficulties, particularly the well-known ‘excess sensitivity’ and ‘excess smoothness’ puzzles. Specifically, the main problems with the standard RE–ICA models are as follows. First, the models cannot generate low contemporaneous correlations between the current account and net income (net income is defined as output minus investment and government spending).4 If net income is a persistent trend-stationary AR(1) process,5 the model predicts that the current account and net income are perfectly correlated, whereas in the data they are only weakly correlated.6 Note that in the data the current account is countercyclical with real GDP and more countercyclical in the emerging economy. (For example, see Neumeyer and Perri, 2005, Aguiar and Gopinath, 2007 and Uríbe, 2009). Second, they cannot generate low persistence of the current account.7 The standard RE models predict that the current account and net income have the same degree of persistence, whereas in the data the persistence of the current account is much lower than that of net income in emerging countries and insignificantly lower than that of net income in developed countries (see Table 1).8 Third, the models cannot generate observed volatility of the current account (Bergin and Sheffrin, 2000 and Gruber, 2004). Fourth, they cannot generate more volatile consumption growth in emerging countries (Aguiar and Gopinath, 2007). Finally, the assumption of certainty equivalence in these models ignores some important channels through which income shocks affect the current account. As shown in Ghosh and Ostry (1997) in post-war quarterly data for the US, Japan, and the UK, the current account is positively correlated with the amount of precautionary savings generated by uncertainty about future net income. Fogli and Perri (2008) also show that in OECD economies changes in country-specific macroeconomic volatility are strongly correlated with changes in net external asset position. Table 1. Emerging vs. developed countries (averages). A: Emerging vs. developed countries (HP filter) σ(y)/μ(y) 3.19(0.20) 1.83(0.07) σ(Δy)/μ(y) 3.82(0.19) 2.07(0.06) ρ(yt, yt− 1) 0.50(0.03) 0.44(0.03) σ(Δc)/σ(Δy) 1.35(0.08) 0.98(0.04) σ(ca)/σ(y) 1.53(0.09) 1.60(0.08) ρ(c, y) 0.33(0.04) 0.46(0.04) ρ(cat, cat − 1) 0.30(0.05) 0.41(0.03) ρ(ca, y) 0.05(0.05) 0.06(0.05) View the MathML sourceρcay,y 0.04(0.04) 0.15(0.04) B: Emerging vs. developed countries (linear filter) σ(y)/μ(y) 9.03(0.43) 4.37(0.18) σ(Δy)/μ(y) 3.82(0.19) 2.07(0.06) ρ(yt, yt− 1) 0.80(0.02) 0.79(0.02) σ(Δc)/σ(Δy) 1.35(0.08) 0.98(0.04) σ(ca)/σ(y) 0.80(0.06) 1.35(0.06) ρ(c, y) 0.68(0.04) 0.63(0.04) ρ(cat, cat − 1) 0.53(0.04) 0.71(0.02) ρ(ca, y) 0.13(0.05) 0.17(0.05) View the MathML sourceρcay,y 0.03(0.05) 0.16(0.05) Table options It is, therefore, natural to turn to new alternatives to the standard RE–ICA model and ask what implications they have for the joint dynamics of consumption, the current account, and income. In this paper, we show that two types of informational frictions, robustness (RB) and information–processing constraints (rational inattention or RI), can significantly improve the model's ability to fit the data discussed above. Specifically, these two types of information imperfections interact with the fundamental shock (the income shock in our model) and give rise to closely related “induced uncertainty”: (i) model uncertainty and (ii) state uncertainty. These two types of induced uncertainty can affect the model's dynamics even within the linear–quadratic (LQ) framework. 9 We adopt Hall's LQ–PIH setting in this paper because the main purpose of this paper is to inspect the mechanisms through which the induced uncertainty affects the joint dynamics of consumption, the current account, and income, and it is much more difficult to study these informational frictions in non-LQ frameworks. 10 After solving the models explicitly, we then examine how the induced uncertainty due to RB and RI can improve the model's predictions on these important dimensions of the joint dynamics of the current account, consumption, and net income in emerging and developed countries we discussed above. In particular, we are interested in two key features of emerging market: consumption volatility exceeds income volatility and less procyclical current accounts with net income found in the data. 11 Hansen and Sargent, 1995 and Hansen and Sargent, 2007a first introduced robustness (a concern for model misspecification) into economic models. In robust control problems, agents are concerned about the possibility that their model is misspecified in a manner that is difficult to detect statistically; consequently, they choose their decisions as if the subjective distribution over shocks was chosen by a malevolent nature in order to minimize their expected utility (that is, the solution to a robust decision-maker's problem is the equilibrium of a max–min game between the decision-maker and nature). Robustness models produce precautionary savings but remain within the class of LQ–Gaussian models, which leads to analytical simplicity.12 A second class of models that produces precautionary savings but remains within the class of LQ–Gaussian models is the risk-sensitive model of Hansen et al. (henceforth HST, 1999).13 We show that even if the parameter value of robustness is the same for all small open countries, the RB model has the potential to lead to the observed different joint behavior of consumption and current accounts across the developed and emerging economies. The reason is that the amount of model uncertainty that affects the model's dynamics is determined by the interaction of the preference for robustness and income uncertainty; consequently, the model with the same parameter value of robustness can still lead to different behavior of consumption and the current account because income uncertainty is different across countries.14 Furthermore, we find that incorporating robustness can improve the model by along the following three dimensions in all small open countries: generating lower contemporaneous correlation between the current account and net income, lower persistence of the current account, and higher relative volatility of consumption growth to income growth. In addition, after calibrating the RB parameter using the detection error probability, we find that RB can help generate the different stochastic properties of the emerging and developed economies. Specifically, the current account in the emerging economy is (1) less correlated with net income, (2) less persistent, and (3) less volatile than that in the developed economy. However, quantitatively, we find that RB by itself cannot fully explain the joint behavior of consumption and the current account in the two small-open economies. We therefore consider the model with imperfect state observation (state uncertainty) due to RI. Sims (2003) first introduced RI into economics and argued that it is a plausible method for introducing sluggishness, randomness, and delay into economic models. In his formulation agents have finite Shannon channel capacity, limiting their ability to process signals about the true state of the world. One key change relative to the RE case is that consumption has a hump-shaped impulse response to changes in income. 15 Using the results in Luo (2008), it is straightforward to show that RI by itself still leads to counterfactual strongly-procyclical current accounts and cannot generate precautionary savings in the LQG setting. 16 However, the combination of RB and RI produces a model that captures many of the facts that are seen as anomalous through the lens of an RE model, while producing consumption dynamics that are consistent with the data. The intuition is that RI introduces (i) slow adjustment to the income shock and (ii) an endogenous noise into the model, which amplifies the importance of model uncertainty in determining the model's dynamics and further improves the model's predictions on the joint behavior of consumption and the current account. We briefly list the results of the RB–RI model. First, we can produce a low correlation between the current account and net income, and in fact can even produce negative correlations for some parameter settings; the key requirement to get low correlations is that the agent has a strong fear of model misspecification. Second, we can produce low persistence in the current account, a consequence of the slow movements in consumption that RI produces. Third, if information–processing is sufficiently restricted, current account volatility can match that observed in the data for emerging markets, although not for developed economies. Fourth, the model produces a hump-shaped consumption response to income, a consequence of RI, and can produce highly volatile consumption growth in emerging economies. Fifth, the precautionary savings effect generated by RB is consistent with the positive correlation between income volatility and average current accounts. We detail in the main body of the paper the intuition for all of these results. The remainder of the paper is organized as follows. Section 2 presents key facts of small open economy business cycles. Section 3 reviews the standard RE–ICA model and discuss the puzzling implications of the model. Section 4 presents the RB–ICA model and discusses some results regarding the joint dynamics of consumption, the current account, and income. Section 5 solves the RB–RI ICA model and presents the implications for the same variables. Section 6 concludes.

We have examined how introducing two types of information imperfections, robustness and rational inattention, into an otherwise standard intertemporal current account model changes the dynamic effects of income shocks on the joint dynamics of consumption and the current account. We have shown that a model with agents who have both a preference for robustness and limited information processing capacity has the potential to better account for the data along a number of dimensions. The model proposed in this paper can also be used to address the international diversification and consumption correlations puzzles (Backus et al., 1992). In Luo et al. (2011) we show that the model incorporating model uncertainty and state uncertainty reduces the correlation of consumption across countries, and can in fact produce consumption correlations lower than income correlations. RB will lower the international consumption correlations by generating heterogenous responses of consumption to income shocks across countries, provided countries differ in terms of their preference for robustness. In addition, in contrast to the intertemporal consumption approach we consider here, the ‘new rule’ approach to the current account assigns the preeminent role to portfolio choice (for conflicting views on the relevance of the new rule, see Kraay and Ventura, 2003). An interesting extension to our study would be to permit portfolio choice and study the dynamics of the current account in the RB–RI model. Finally, to explore the mechanisms through which the two informational frictions interact and work, in this paper we have set up the model in a parsimonious way so that we can obtain a closed-form solution. We think that the mechanisms and insights we have explored in this simple framework can be carried over to more general cases. In particular, extending the model to incorporate the global interest rate shock emphasized by Nason and Rogers (2006) will be critical for demonstrating conclusively the utility of the RB–RI framework.