Business Cycle Dynamics under Rational Inattention

Transcription

1 Business Cycle Dynamics under Rational Inattention Bartosz Maćkowiak European Central Bank and CEPR Mirko Wiederholt Northwestern University First draft: June 27 This draft: September 29 Preliminary Version Abstract This paper develops a dynamic stochastic general equilibrium model with rational inattention Households and decision-makers in firms have limited attention and decide how to allocate their attention We study the implications of rational inattention for business cycle dynamics We find that the impulse responses of prices under rational inattention have several properties of empirical impulse responses: (i) prices respond slowly to monetary policy shocks, (ii) prices respond faster to aggregate TFP shocks, and (iii) prices respond very fast to disaggregate shocks As a result, profit losses due to deviations of the actual price from the profit-maximizing price are an order of magnitude smaller than in the Calvo model that generates the same real effects We also find that consumption responds slowly to monetary policy shocks For standard parameter values, deviations from the consumption Euler equation are cheap in utility terms, implying that households devote little attention to the consumption-saving decision We thank for helpful comments: Paco Buera, Larry Christiano, James Costain, Martin Eichenbaum, Christian Hellwig, Giorgio Primiceri, Bruno Strulovici as well as seminar and conference participants at Bank of Canada, Chicago Fed, Columbia, Cowles Foundation Summer Conference 29, DePaul, Duke, European Central Bank, ESSIM 28, EUI, Harvard, MIT, Minneapolis Fed, Minnesota Workshop in Macroeconomic Theory 29, NBER Summer Institute 28, NAWMES 28, Princeton, Richmond Fed, Riksbank, SED 28, Stony Brook, Toronto, UCSD, University of Chicago and University of Hong Kong The views expressed in this paper are solely those of the authors and do not necessarily reflect the views of the European Central Bank bartoszmackowiak@ecbint and m-wiederholt@northwesternedu

2 Introduction This paper develops a dynamic stochastic general equilibrium model with rational inattention We model the idea that decision-makers have limited attention and decide how to allocate their attention Following Sims (23), we model attention as an information flow and we model limited attention as a constraint on information flow We let agents choose the allocation of information flow We study the implications of rational (in)attention for business cycle dynamics The economy consists of households, firms and a government Firms produce differentiated goods using a variety of types of labor Households supply the different types of labor, consume a variety of goods, and hold nominal government bonds Decision-makers in firms take price setting and factor mix decisions Households take consumption and wage setting decisions The central bank sets the nominal interest rate according to a Taylor rule Prices and wages are physically fully flexible and there is no habit formation in consumption The only source of inertia in the model is the limited attention by decision-makers We compute the impulse responses of prices and quantities to monetary policy shocks, aggregate technology shocks, and micro-level shocks under rational inattention by all decision-makers We find that the model can match several features of empirical impulse responses We find that, in our model and for our parameter values, rational inattention by decisionmakers in firms has the following implications The price level responds slowly to monetary policy shocks More precisely, the impulse response of the price level to monetary policy shocks under rational inattention by decision-makers in firms resembles the impulse response in a Calvo model with an average price duration of 75 months At the same time, the price level responds fairly quickly to aggregate technology shocks, and individual prices respond very quickly to micro-level shocks The reason is the optimal allocation of attention Decision-makers in firms decide to devote little attention to monetary policy disturbances, about twice as much attention to the state of aggregate technology, and a lot of attention to market-specific conditions Therefore, prices respond slowly to monetary policy shocks, prices respond fairly quickly to aggregate technology shocks, and prices respond very quickly to market-specific shocks Furthermore, profit losses due to deviations of the actual price from the profit-maximizing 2

3 price are an order of magnitude smaller than in the Calvo model that generates the same real effects of monetary policy shocks More precisely, profit losses due to sub-optimal price responses to aggregate conditions are 23 times smaller than in the Calvo model; and profit losses due to sub-optimal price responses to idiosyncratic conditions are 57 times smaller than in the Calvo model that generates the same real effects of monetary policy shocks The main reason for this result is the optimal allocation of attention, implying that prices respond slowly to monetary policy shocks, but prices respond fairly quickly to aggregate technology shocks, and prices respond very quickly to idiosyncratic shocks By contrast, in the Calvo model prices respond slowly to all those shocks The other reason for this result is that under rational inattention by decision-makers in firms deviations of the actual price from the profit-maximizing price are less likely to be extreme than in the Calvo model When we add rational inattention by households, we find that households devote little attention to the consumption-saving decision because for standard parameter values deviations from the consumption Euler equation are cheap in utility terms Since households devote little attention to the consumption-saving decision, consumption responds slowly to shocks It turns out that the impulse responses of consumption to shocks look similar to the impulse responses of consumption in a model with habit formation This paper is related to two strands of literature: (i) the literature on rational inattention (eg Sims (23, 26), Luo (28), Maćkowiak and Wiederholt (29), Van Nieuwerburgh and Veldkamp (28), and Woodford (29)); and (ii) the literature on business cycle models with imperfect information (eg Lucas (972), Woodford (22), Mankiw and Reis (22), Lorenzoni (28) and Angeletos and La O (29)) The main innovation with respect to the existing literature on rational inattention is that we solve a dynamic stochastic general equilibrium model The main innovation with respect to the existing literature on business cycle models with imperfect information is that the information structure is the outcome of an optimization problem The paper is organized as follows Section 2 describes all features of the economy apart from the information structure Section 3 characterizes the steady state of the non-stochastic version of the economy In Section 4 we derive the objective that decision-makers in firms maximize when they decide how to allocate their attention In Section 5 we derive the 3

4 objective that households maximize when they decide what to focus on Section 6 describes issues related to aggregation Section 7 characterizes the solution of the model under perfect information Section 8 presents numerical solutions of the model under rational inattention by decision-makers in firms Here we maintain the assumption that households have perfect information Section 9 presents numerical solutions of the model under rational inattention by decision-makers in firms and households Section concludes 2 Model In this section, we describe all features of the economy apart from the information structure Thereafter, we solve the model for alternative assumptions about the information structure: (i) perfect information, and (ii) rational inattention 2 Households There are J households in the economy Households supply differentiated types of labor, consume a variety of goods, and hold nominal government bonds Time is discrete and households have an infinite horizon Each household seeks to maximize the expected discounted sum of period utility The discount factor is β (, ) The period utility function is where U (C jt,l jt )= C γ jt γ i= ϕ L+ψ jt +ψ, () Ã IX! θ C jt = C θ θ θ ijt (2) Here C ijt is consumption of good i by household j in period t, C jt is composite consumption by household j in period t and L jt is labor supply by household j in period t The parameter γ> is the inverse of the intertemporal elasticity of substitution and the parameters ϕ> and ψ affect the disutility of supplying labor There are I different consumption goods and the parameter θ> is the elasticity of substitution between those consumption goods The assumption of a constant elasticity of substitution between consumption goods is only for ease of exposition One could use a general constant returns-to-scale aggregator 4

5 The flow budget constraint of household j in period t reads IX i= P it C ijt + B jt = R t B jt +(+τ w ) W jt L jt + D t J T t J (3) Here P it is the price of good i in period t, B jt are holdings of nominal government bonds by household j between period t and period t +, R t is the nominal gross interest rate on those bond holdings, W jt is the nominal wage rate for labor supplied by household j in period t, τ w is a wage subsidy paid by the government, (D t /J) is a pro-rata share of nominal aggregate profits, and (T t /J) is a pro-rata share of nominal lump-sum taxes We assume that all households have the same initial bond holdings B j, > We also assume that bond holdings have to be positive in every period, B jt > Wehavetomakesome assumption to rule out Ponzi schemes We choose this particular assumption because it allows us to rewrite the model in terms of logs of all variables One can think of households having an account The account holds only nominal government bonds, and the balance on the account has to be positive In every period, each household chooses a consumption vector, (C jt,,c Ijt ),anda wage rate, W jt Each household commits to supply any quantity of labor at that wage rate Each household takes as given: all prices of consumption goods, all wage rates set by other households, the nominal interest rate and all aggregate quantities 22 Firms There are I firms in the economy Firms supply the differentiated consumption goods Firm i supplies good i The production function of firm i is Y it = e a t e a it L α it, (4) where JX L it = j= η η L ijt η η (5) Here Y it is output, L ijt is input of type j labor, L it is composite labor input and (e a t e a it) is total factor productivity of firm i in period t Total factor productivity has an aggregate component, e a t,andafirm-specific component,e a it Type j labor is the labor supplied 5

6 by household j Since there are J households, there are J types of labor 2 The parameter η> is the elasticity of substitution between those types of labor The parameter α (, ] is the elasticity of output with respect to composite labor Nominal profits of firm i in period t equal ( + τ p ) P it Y it JX W jt L ijt, (6) j= where τ p is a production subsidy paid by the government In every period, each firm sets a price, P it,andchoosesafactormix, ³ˆLit,,ˆL i(j )t, where ˆL ijt =(L ijt /L it ) denotes firm i s relative input of type j labor in period t Eachfirm commits to supply any quantity of the good at that price Each firm then produces the quantity demanded with the chosen factor mix Each firm takes as given: all prices set by other firms, all wage rates set by households, the nominal interest rate and all aggregate quantities 23 Government There is a monetary authority and a fiscal authority nominal interest rate according to the rule R t R = µ Rt R ρr " µπt Π The monetary authority sets the φπ µ # ρr φy Yt e εr t, (7) Y where Π t =(P t /P t ) is inflation, Y t is aggregate output defined as à IX! Y t = P it Y it /P t, (8) i= and ε R t is a monetary policy shock The price index P t will be defined later Here R, Π and Y denote the values of the nominal interest rate, inflation and aggregate output in the non-stochastic steady state The policy parameters satisfy ρ R [, ), φ π > and φ y The government budget constraint in period t reads à IX! JX T t +(B t B t )=(R t ) B t + τ p P it Y it + τ w W jt L jt (9) i= j= 2 The assumption that all types of labor appear in the labor aggregator is for ease of exposition One could assume that a firm-specific subset of types of labor appear in the labor aggregator 6

7 The government has to finance interest on nominal government bonds, the production subsidy and the wage subsidy government bonds The government can collect lump-sum taxes or issue new We assume that the government sets the production subsidy, τ p, and the wage subsidy, τ w,soastocorrectthedistortionsarisingfromfirms market power in the goods market and households market power in the labor market In particular, we assume that τ p = where ϑ denotes the price elasticity of demand, and τ w = ϑ, () ϑ ζ, () ζ where ζ denotes the wage elasticity of labor demand 3 We make this assumption to abstract from the level distortions arising from monopolistic competition 24 Shocks There are three types of shocks in the economy: aggregate technology shocks, firm-specific productivity shocks and monetary policy shocks We assume that, for all i =,,I, the stochastic processes {a t }, {a it } and ε R ª t are independent Furthermore, we assume that the firm-specific productivity processes, {a it }, are independent across firms In addition, we assume that the number of firms is sufficiently large so that IX a it = (2) I i= Finally, we assume that a t follows a stationary Gaussian first-order autoregressive process with mean zero, each a it follows a stationary Gaussian first-order autoregressive process with mean zero, and ε R t the period t innovation to a t and a it by ε A t follows a Gaussian white noise process In the following, we denote and ε I it, respectively 3 When households have perfect information then ϑ = θ and thus τ p = θ θ By contrast, when households have imperfect information then ϑ (the price elasticity of demand) may differ from the parameter θ Therefore, the value of the production subsidy () may vary across information structures For the same reason, the value of the wage subsidy () may vary across information structures 7

8 25 Notation In this subsection, we introduce notation that will be convenient Throughout the paper, C t will denote aggregate composite consumption C t = JX C jt, (3) j= and L t will denote aggregate composite labor input L t = IX L it (4) i= Furthermore, ˆP it will denote the relative price of good i and Ŵjt will denote the relative wage rate for type j labor Finally, Wjt will denote the real wage rate for type j labor and W t will denote the real wage index In each section, we will specify the definition of P t and W t ˆP it = P it P t, (5) Ŵ jt = W jt W t (6) W jt = W jt P t, (7) W t = W t P t (8) 3 Non-stochastic steady state We begin by characterizing the non-stochastic steady state of the economy described in the previous section We define a non-stochastic steady state as a solution of the non-stochastic version of the economy with the property that real quantities, relative prices, the nominal interest rate and inflation are constant over time In the following, variables without the subscript t denote values in the non-stochastic steady state 8

9 In this section, P t denotes the following price index P t = Ã IX i= and W t denotes the following wage index JX W t = j= P θ it W η jt! θ, (9) η (2) In the non-stochastic steady state, the households first-order conditions read R Π = β, (2) and C ij C j = ˆP θ i, (22) ψ η W j = ϕ ³Ŵ j L C γ j (23) The firms first-order conditions read ˆP i = W ³ ˆP θ α i C, (24) α and ˆL ij = Ŵ η j (25) The firms price setting equation (24) implies that all firmssetthesamepriceinthe non-stochastic steady state Households therefore consume the different consumption goods in equal amounts, implying that all firms produce the same amount Since in addition all firms have the same technology in the non-stochastic steady state, all firms have the same composite labor input It follows from the definition of the price index (9), the consumption aggregator (2) and the definition of aggregate composite labor input (4) that ˆP θ i = µ Cij C j θ θ = L i L = I (26) Furthermore, in the non-stochastic version of the economy, all households face the same decision problem, have the same information and their decision problem has a unique constant solution, implying that all households choose the same consumption level and set the same 9

10 wage rate in the non-stochastic steady state Firms therefore hire the different types of labor in equal amounts It follows from the definition of aggregate composite consumption (3), the definition of the wage index (2) and the labor aggregator (5) that We will use equations (2)-(27) below C j C = Ŵ η j = ˆL η η ij = J (27) One can show that equations (2)-(27), Y i = L α i, Y θ i = C i and C i = ˆP i C imply that all variables appearing in equations (2)-(27) are uniquely determined apart from the nominal interest rate, R, and inflation, Π For ease of exposition, we select Π = Equation (2) then implies R =(/β) It is worth pointing out in this context that the steady-state inflation rate has no effect on real variables in both the non-stochastic and the stochastic version of the economy In addition, in the non-stochastic steady state, the initial price level, P, is not determined We will assume that P equals some value P Finally, for given initial real bond holdings B j, / P, fiscal variables in the non-stochastic steady state are uniquely determined by the requirement that real quantities are constant over time The reason is that real bond holdings are a real quantity and real bond holdings are constant over time if and only if the government runs a balanced budget in real terms (ie real lump-sum taxes equal the sum of real interest payments and real subsidy payments) 4 Derivation of the firms objective In this section, we derive a log-quadratic approximation to the expected discounted sum of profits We will use this expression below when we assume that decision-makers in firms choose the allocation of attention so as to maximize the expected discounted sum of profits To derive this expression, we proceed in four steps: (i) we make a guess concerning the demand function for a consumption good, (ii) we substitute the demand function and the production function into the expression for profits to obtain the profit function, (iii) we make an assumption about how decision-makers in firms value profits in different states of the world, and (iv) we compute a log-quadratic approximation to the expected discounted sum of profits

11 First, we guess that the demand function for good i has the form µ ϑ Pit C it = ς C t, (28) P t where C t is aggregate composite consumption, P t is a price index satisfying the following equation for some function d that is homogenous of degree one, symmetric and continuously differentiable P t = d (P t,,p It ), (29) and ϑ> and ς>are undetermined coefficients satisfying ς ˆP ϑ i = ˆP θ i (3) When we solve the model for alternative assumptions about the information structure below, we will always verify that this guess concerning the demand function is correct 4 Second, we substitute the production function (4)-(5) and the demand function (28) into the expression for nominal profits (6) to obtain the profit function We begin by rewriting the expression for nominal profits (6) JX JX ( + τ p ) P it Y it W jt L ijt =(+τ p ) P it Y it L it W jt ˆLijt, (3) j= where ˆL ijt =(L ijt /L it ) is firm i s relative input of type j labor Theterminsquarebrackets on the right-hand side is the wage bill per unit of composite labor input equations (4)-(5) yields and L it = = µ Yit e at e a it JX j= j= Rearranging α, (32) η η ˆL ijt (33) Substituting the technology (32)-(33), Y it = C it and the demand function (28) into the expression for nominal profits (3) yields the profit function ³ ϑ µ α ϑ Pit ς Pit P Ct J t X ( + τ p ) P it ς C t e a t e a it W jtˆlijt + W Jt P t j= J X j= η η ˆL (34) 4 To give the simplest example, when households have perfect information then P t is given by equation (9), ϑ = θ and ς = ijt η η

12 Profits of firm i in period t depend on the following variables that the decision-maker in the firm chooses: P it, ˆL it,,ˆl i(j )t ; and on the following variables that the decision-maker in the firmtakesasgiven:p t,a t,a it,c t,w t,,w Jt Third, we make an assumption about how decision-makers in firms value profits in different states of the world Since the economy described in Section 2 is an incompletemarkets economy with multiple owners of a firm, it is unclear how firms should value profits in different states of the world Therefore, we assume a general stochastic discount factor In particular, we assume that, in period, decision-makers in firms value nominal profits in period t using the following stochastic discount factor Q,t = β t λ (C t,,c Jt ) P t, (35) where P t is the price index appearing in the demand function (28) and λ is some twice continuously differentiable function with the property 5 λ (C,,C J )=C γ j (36) Then, in period, the expected discounted sum of profits equals " X ³ E i, β t F ˆPit, ˆL it,,ˆl i(j )t,a t,a it,c t,,c Jt, W t,, Jt # W, (37) where E i, is the expectation operator conditioned on the information of the decision-maker of firm i in period and the function F is given by ³ F ˆPit, ˆL it,,ˆl i(j )t,a t,a it,c t,,c Jt, W t,, Jt W JX = λ (C t,,c Jt )(+τ p ) ς C jt ˆP ϑ it j= JX ς ˆP ϑ it C jt j= λ (C t,,c Jt ) e a t e a it α J X W jtˆlijt + W Jt j= J X j= η η ˆL ijt η η (38) 5 For example, the stochastic discount factor could be a weighted average of the marginal utilities of the different households (ie λ (C t,,c Jt )= J λ jc γ jt with λ j and J λ j =) Equation (36) j= j= would be satisfied because all households have the same marginal utility in the non-stochastic steady state 2

13 In the following, small variables denote log-deviations from the non-stochastic steady state For example, c jt = ln(c jt /C j ) Expressing the real profit function F in terms of logdeviations from the non-stochastic steady state and using equations (), (24)-(25), (27), (3), Y i = L α i, Y i = C i and C i = θ ˆP i C yields the following expression for the expected discounted sum of profits " X ³ E i, β t f ˆp it, ˆl it,,ˆl i(j )t,a t,a it,c t,,c Jt, w t,, w Jt #, (39) where f ³ˆp it, ˆl it,,ˆl i(j )t,a t,a it,c t,,c Jt, w t,, w Jt = λ (C e c t,,c J e c ϑ Jt ) ϑ α WL i J JX j= e ( ϑ)ˆp it+c jt λ (C e c t,,c J e c Jt ) WL i e ϑ α ˆp it α (a t+a it ) J J X J X e w jt+ˆl ijt + e w Jt J J j= j= e η η ˆl ijt η η JX j= e c jt α (4) Fourth, we compute a log-quadratic approximation to the expected discounted sum of profits around the non-stochastic steady state We obtain the following result Proposition (Expected discounted sum of profits) Let f denote the real profit function defined by equation (4) and let f denote the second-order Taylor approximation to f at the non-stochastic steady state Let E i, denote the expectation operator conditioned on the information of the decision-maker of firm i in period Let x t, z t and v t denote the following vectors x t = z t = v t = ³ ³ ³ ˆp it ˆlit ˆli(J )t, (4) a t a it c t c Jt w t w Jt, (42) x t zt (43) Let v m,t denote the mth element of v t Suppose that there exist two constants δ<(/β) and A R such that, for each period t and for all m and n, E i, v m,t v n,t <δ t A (44) 3

14 Then = " # " X # X E i, β t f (xt,z t ) E i, β t f (x t,z t ) X β t E i, 2 (x t x t ) H (x t x t ), (45) where the matrix H is given by H = C γ j WL i ϑ α + α α ϑ 2 ηj ηj ηj ηj ηj ηj ηj 2 ηj, (46) and the vector x t is given by: ˆp α α it = ϑ JX J + α α j= c jt + + α α ϑ J JX j= w jt α + α α ϑ (a t + a it ), (47) and Proof See Appendix A ˆl ijt = η w jt J JX j= w jt (48) After the log-quadratic approximation to the real profit function, the profit-maximizing price in period t is given by equation (47) and the profit-maximizing factor mix in period t is given by equation (48) Furthermore, after the log-quadratic approximation to the real profit function, the loss in profits in period t in the case of a deviation from the profitmaximizing decisions (ie x t 6= x t ) is given by the quadratic form in expression (45) The upper-left element of the matrix H determines the profit loss in the case of a sub-optimal price setting decision The profit loss in the case of a sub-optimal price setting decision is increasing in the price elasticity of demand, ϑ, and increasing in the degree of decreasing returns-to-scale, (/α) The lower-right block of the matrix H determines the profit lossin the case of a sub-optimal factor mix decision The profit loss in the case of a sub-optimal factor mix decision is decreasing in the elasticity of substitution between types of labor, η, 4

15 and depends on the number of types of labor, J Note that the diagonal elements of H determine the profit loss in the case of a deviation in a single variable, while the off-diagonal elements of H determine how a deviation in one variable affects the loss in profits due to a deviation in another variable Finally, condition (44) ensures that, in the expression for the expected discounted sum of profits, after the log-quadratic approximation to the real profit function, one can change the order of integration and summation and the infinite sum converges It is worth pointing out that the profit-maximizing decision vector (47)-(48) and the expected discounted sum of profit losses (45) depend only to a limited extent on the function λ appearing in the discount factor (35) The profit-maximizing decision vector (47)-(48) does not depend at all on the function λ because the profit-maximizing price and the profit-maximizing factor mix are the solution to a static decision problem in the economy described in Section 2 The expected discounted sum of profit losses (45) depends only on the steady-state value of the function λ because of the log-quadratic approximation to the real profit function So far we have only derived an expression for the expected discounted sum of profits in the economy described in Section 2, but from this expression one can already see to some extent how a decision-maker in a firm who cannot attend perfectly to all available information will allocate his/her attention For example, the attention devoted to the price setting decision will depend on the profit lossthatthefirm incurs in the case of a price setting mistake (ie, a deviation of the actual price from the profit-maximizing price) Thus, the attention devoted to the price setting decision will depend on the upper-left element of the matrix H Furthermore, the decision-maker will track closely those changes in the environment that in expectation cause most of the variation in the profit-maximizing decisions As one can see from equations (47)-(48), which changes in the environment in expectation cause most of the variation in the profit-maximizing decisions depends on the technology parameters α and η, the calibration of the exogenous processes as well as the behavior of the other agents in the economy 5

16 5 Derivation of the households objective In this section, we derive a log-quadratic approximation to the expected discounted sum of period utility We will use this expression below when we assume that households choose the allocation of attention so as to maximize the expected discounted sum of period utility To derive this expression, we proceed in three steps: (i) we make a guess concerning the demand function for a particular type of labor, (ii) we substitute the labor demand function, the consumption aggregator and the flow budget constraint into the period utility function to obtain a period utility function that incorporates those constraints, and (iii) we compute a log-quadratic approximation to the expected discounted sum of period utility First, we guess that the demand function for type j labor has the form L jt = ξ µ Wjt W t ζ L t, (49) where L t is aggregate composite labor input, W t is a wage index satisfying the following equation for some function h that is homogenous of degree one, symmetric and continuously differentiable W t = h (W t,,w Jt ), (5) and ζ> and ξ> are undetermined coefficients satisfying ξŵ ζ j = Ŵ η j (5) When we solve the model for alternative assumptions about the information structure below, we will always verify that this guess concerning the labor demand function is correct 6 Second, we substitute the consumption aggregator (2), the flow budget constraint (3) and the labor demand function (49) into the period utility function () to obtain a period utility function that incorporates those constraints We begin by rewriting the flow budget constraint (3) as à IX! C jt P it Ĉ ijt + B jt = R t B jt +(+τ w ) W jt L jt + D t J T t J, i= 6 To give the simplest example, when firms have perfect information then W t is given by equation (2), ζ = η and ξ = 6

17 where Ĉijt =(C ijt /C jt ) is relative consumption of good i by household j The term in brackets on the left-hand side is consumption expenditure per unit of composite consumption Rearranging yields C jt = R t B jt B jt +(+τ w ) W jt L jt + D t J X I P itĉijt i= Dividing the numerator and the denominator on the right-hand side by P t,wherep t is some price index, yields T t J C jt = R t Π t Bjt B jt +(+τ w ) W jt L jt + D t X I i= J T t J ˆP it Ĉ ijt, (52) where B jt =(B jt /P t ) are real bond holdings by the household, Dt =(D t /P t ) are real aggregate profits, Tt =(T t /P t ) are real lump-sum taxes, and Π t =(P t /P t ) is inflation Rearranging the consumption aggregator (2) yields = IX i= Ĉ θ θ ijt (53) Substituting the flow budget constraint (52), the consumption aggregator (53) and the labor demand function (49) into the period utility function () yields a period utility function that incorporates those constraints: γ R t Π t γ ϕ +ψ Bjt B jt +(+τ w ) W ³ ζ W jt ξ jt Lt + D t W t J T t J XI i= ξ Ã Wjt W t ˆP it Ĉ ijt + ˆP XI It Ã! ζ L t +ψ i= Ĉ θ θ ijt! θ θ γ (54) Expressing the period utility function (54) in terms of log-deviations from the non-stochastic steady state and using equations (), (2)-(23), (26), (5) and L j = Ŵ η j L yields our final 7

18 period utility function: C γ ω j Bβ e r t π t + b jt ω B e b jt + ζ ζ ω W e ( ζ) w jt+ζ w t +l t + ω D e d t ω T e t t γ Ã! X XI θ θ eˆp it+ĉ ijt + I eˆp It I e θ θ ĉijt I I i= γ C γ j +ψ ω W e ζ(+ψ)( w jt w t)+(+ψ)l t, (55) where ω B, ω W, ω D and ω T denote the following steady-state ratios: ³ ³ D T ω B ω W ω D ω T = Bj Wj L j J J (56) C j C j C j C j Third, we compute a log-quadratic approximation to the expected discounted sum of period utility around the non-stochastic steady state i= γ Proposition 2 (Expected discounted sum of period utility) Let g denote the functional that is obtained by multiplying the period utility function (55) by β t and summing over all t from zero to infinity Let g denote the second-order Taylor approximation to g at the nonstochastic steady state Let E j, denote the expectation operator conditioned on information of household j in period Letx t, z t and v t denote the following vectors x t = z t = v t = ³ bjt w jt ĉ jt ĉ I jt, (57) ³ r t π t w t l t dt t t ˆp t ˆp It, (58) ³ x t zt (59) Let v m,t denote the mth element of v t Suppose that i E j, h b2 j, <, (6) and, for all m, E j, bj, v m, < (6) Furthermore, suppose that there exist two constants δ<(/β) and A R such that, for each period t, forτ =, and for all m and n, E j, v m,t v n,t+τ <δ t A (62) 8

19 Then = i h i E j, h g ³ bj,,x,z,x,z, E j, g ³ bj,,x,z,x,z, X β t E j, 2 (x t x t ) H (x t x t )+(x t x t ) H xt+ x t+ (63) Here the matrix H is given by ³ γω 2 B + β γω B ζω W γω B ζω W ζω W (γζω W ++ζψ) H = C γ 2 j θi θi 2 θi θi, (64) and the matrix H is given by H = C γ j γω 2 B γω Bζω W (65) Furthermore, the process {x t } is defined by the following two requirements: (i) the vector v t with x t = x t satisfies conditions (6)-(62), and (ii) in each period t, " Ã! # c jt = E t r t π t+ IX (ˆp it+ ˆp it ) + c jt+, (66) γ I i= Ã! w jt γ = +ζψ c jt + ψ +ζψ (ζ w t + l t )+ IX ˆp it, (67) +ζψ I i= Ã! ĉ ijt = θ ˆp it IX ˆp it, (68) I where c jt is defined by i= c jt = ω ³ B r t π t + β b jt ω B b jt + ζ ζ ω W ( ζ) w jt + ζ w t + l t Ã! IX +ω D dt ω T t t ˆp it, (69) I i= 9

20 and E t denotes the expectation operator conditioned on the entire history of the economy up to and including period t Proof See Appendix B After the log-quadratic approximation to the expected discounted sum of period utility, stochastic processes for real bond holdings, the real wage rate and the consumption mix that satisfy conditions (6)-(62) can be ranked using equation (63) Equations (66)-(69) characterize the optimal behavior under perfect information (ie the decisions the household wouldtakeifineachperiodt the household knew the entire history of the economy up to and including period t) Equation (63) gives the loss in expected lifetime utility in the case of deviations from the optimal behavior under perfect information The upper-left block of the matrix H and the upper-left block of the matrix H determine the loss in expected lifetime utility in the case of sub-optimal real bond holdings and wage setting A single percentage deviation in real bond holdings from optimal bond holdings causes a larger utility loss the larger γ, ω B and (R/Π) =(/β) See the (,) element of the matrix H A single percentage deviation in the real wage rate from the optimal wage rate causes a larger utility loss the larger γ, ψ, ω W and ζ See the (2,2) element of the matrix H Furthermore, the off-diagonal elements of H show that a bond deviation in period t affects the utility cost of a wage deviation in period t, andthefirst row of H shows that a bond deviation in period t affects both the utility cost of a bond deviation in period t +and the utility cost of a wage deviation in period t + The lower-right block of the matrix H determines the utility loss in the case of a sub-optimal consumption mix The loss is decreasing in the elasticity of substitution between consumption goods, θ, and depends on the number of consumption goods, I Finally, conditions (6)-(62) ensure that, in the expression for the expected discounted sum of period utility, after the log-quadratic approximation to expected lifetime utility, one can change the order of integration and summation and all infinite sums converge From Proposition 2 one can already see how some parameters will matter for the optimal allocation of attention by a household that cannot attend perfectly to all available information For example, consider the role of γ Increasing γ raises the utility loss caused by a deviation of real bond holdings from optimal bond holdings At the same time, increasing γ 2

21 lowers the response of optimal bond holdings to the real interest rate The relative strength of these two effects will determine whether for a household with a higher γ it is more or less important to be aware of movements in the real interest rate 6 Aggregation In this section, we describe issues related to aggregation In the following, we will work with log-linearized equations for all aggregate variables Log-linearizing the equations for aggregate output (8), for aggregate composite consumption (3) and for aggregate composite labor input (4) yields y t = IX (ˆp it + y it ), (7) I i= c t = J and l t = I JX c jt, (7) j= IX l it (72) Log-linearizing the equations for the price index (29) and for the wage index (5) yields IX = ˆp it, (73) i= i= and JX = ŵ jt (74) j= Note that the last two equations can also be stated as p t = I and w t = J IX p it, (75) i= JX w jt (76) Furthermore, we will work with log-linearized equations when we aggregate the demands for a particular consumption good or for a particular type of labor Formally, j= c it = JX c ijt, (77) J 2 j=

22 and l jt = I IX l ijt (78) Note that the production function (4) and the monetary policy rule (7) are already log-linear: y it = a t + a it + αl it, (79) and r t = ρ R r t +( ρ R ) φ π π t + φ y y t + ε R t (8) i= 7 Case : Perfect information In this section, we present the solution of the model under perfect information This solution will serve as a benchmark We define the solution of the model under perfect information as follows: In each period t, all agents know the entire history of the economy up to and including period t; firms choose the profit-maximizing price and factor mix; households choose the utility-maximizing consumption vector and nominal wage rate; the government sets the nominal interest rate according to the monetary policy rule, pays subsidies so as to correct the distortions due to market power and chooses a fiscal policy that satisfies the government budget constraint; aggregate variables are given by their respective equations; and households have rational expectations The following proposition characterizes real variables at the solution of the model under perfect information after the log-quadratic approximation to the real profit function (see Section 4), the log-quadratic approximation to the expected discounted sum of period utility (see Section 5) and the log-linearization of the equations for the aggregate variables (see Section 6) Proposition 3 (Solution of the model under perfect information) A solution to the system of equations (47)-(48), (66)-(69), (7)-(8), (2) and y it = c it withthesameinitialbond holdings for each household and a non-explosive bond sequence for each household (ie 22

23 lim s E t h β s+ ³ bj,t+s+ b j,t+s i =)satisfies: and y t = +ψ c t = α + αγ + ψ a t, (8) l t = γ α + αγ + ψ a t, (82) w t = γ + ψ α + αγ + ψ a t, (83) r t E t [π t+ ] = +ψ γ α + αγ + ψ E t [a t+ a t ], (84) ĉ ijt = θˆp it, (85) α ˆp it = + α α θ a it, (86) ˆlijt = ηŵ jt, (87) ŵ jt = (88) Proof See Appendix C Under perfect information, aggregate output, aggregate composite consumption, the aggregate composite labor input, the real wage index, and the real interest rate depend only on aggregate technology The relative price of good i and relative consumption of good i by household j depend only on firm-specific productivity of firm i The relative wage rate for type j labor and firm i s relative input of type j labor are constant Under perfect information, monetary policy has no real effects in this model Monetary policy does affect nominal variables The nominal interest rate and inflation follow from the monetary policy rule (8) and the real interest rate (84) Since ( ρ R ) φ π > and ( ρ R ) φ π + ρ R >, the equilibrium paths of the nominal interest rate and inflation are locally determinate 7 7 See Woodford (23), Chapter 2, Proposition 28 23

24 8 Case 2: Firms have limited attention and households have perfect information In this section, we solve the model assuming rational inattention by decision-makers in firms By rational inattention we mean that decision-makers have limited attention and that they decide how to allocate their attention For the moment, we continue to assume that households have perfect information to isolate the implications of rational inattention by decision-makers in firms 8 The firms attention problem Following Sims (23), we model attention as an information flow and we model limited attention as a constraint on information flow To take decisions that are on average close to the profit-maximizing decisions, the decision-maker in a firm has to be aware of changes in the economic environment that cause changes in the profit-maximizing decisions Being aware of stochastic changes in the environment requires information flow Decision-makers that have limited attention therefore face a trade-off: Tracking closely particular changes in the environment improves decision making but also uses up valuable information flow We formalize this trade-off by letting the decision-maker choose directly the stochastic process for the decision vector, subject to a constraint on information flow For example, the decision-maker could decide to respond swiftly and correctly with his/her price setting decision to changes in firm-specific productivity but this would require allocating attention to firm-specific productivity We assume that the decision-maker in a firm chooses the level and the allocation of information flow so as to maximize the expected discounted sum of profits net of the cost of information flow Formally, the attention problem of the decision-maker in firm i reads: ( X β t E i, 2 (x t x t ) H (x t x t ) max κ,b (L),,B 3 (L),C (L),,C 3 (L),ζ,χ ) μ β κ, (89) 24

25 where x t x t = p it ˆlit p it ˆl it, (9) ˆli(J )t ˆl i(j )t subject to the equations characterizing the profit-maximizing decisions p it = A (L) ε A t + A {z } 2 (L) ε R t {z } p A it p R it + A 3 (L) ε I it {z } p I it (9) ˆl ijt = ηŵ jt, (92) the equations characterizing the actual decisions p it = B (L) ε A t + C (L) ν A it + B {z } 2 (L) ε R t + C 2 (L) ν R it {z } p A it p R it + B 3 (L) ε I it + C 3 (L) ν I it {z } p I it µ ˆlijt = ζ ŵ jt + Var(ŵ jt) ν L ijt, (94) χ and the constraint on information flow I (93) ³n p A it,p R it,p I it, ˆl o it,,ˆl i(j )t ; np A it,p R it,p I it, ˆl o it,,ˆl i(j )t κ (95) Here A (L) to A 3 (L), B (L) to B 3 (L), andc (L) to C 3 (L) are infinite-order lag polynomials The noise terms ν A it, νr it, νi it and νl ijt appearing in the actual decisions are assumed to follow unit-variance Gaussian white noise processes that are: (i) independent of all other stochastic processes in the economy, (ii) firm-specific, and (iii) independent of each other The operator I measures the amount of information that the actual decisions contain about the profit-maximizing decisions The operator I is formally defined below Proposition states that, after the log-quadratic approximation to the real profit function, the profit-maximizing decisions of firm i in period t are given by equations (47)-(48) and the expected profit loss due to suboptimal decisions is given by equation (45) Objective (89) therefore states that the decision-maker of firm i chooses the level and the allocation of information flow so as to maximize the expected discounted sum of profits net of the cost of information flow 8 The variable κ is the information flow devoted to the price setting 8 A more negative value of expression (45) means a larger expected profit loss due to suboptimal decisions 25

26 and factor mix decisions The parameter μ is the per-period marginal cost of information flow This marginal cost of information flow can be interpreted as an opportunity cost (ie the cost of devoting less attention to some other activity) or a monetary cost (eg an extra wage payment to a manager to improve decision-making) 9 Equation (9) characterizes the profit-maximizing price setting decision Here we guess that the profit-maximizing price given by equation (47) has the representation (9) after using equations (7), (76) and ˆp it = p it p t and after substituting in the equilibrium processes for p t, c t, w t, a t and a it We will verify this guess Equation (92) characterizes the profit-maximizing factor mix decision Here we have simply rewritten the equation for the profit-maximizing factor mix (48) using equations (76) and ŵ jt = w jt w t Equation (93) characterizes the actual price setting decision By choosing the lag polynomials B (L) and C (L) to B 3 (L) and C 3 (L), the decision-maker chooses the joint distribution of the profit-maximizing price and the actual price For example, if B (L) =A (L) and C (L) =, the price set by the decision-maker responds perfectly to aggregate technology shocks Similarly, if B 2 (L) =A 2 (L) and C 2 (L) =, the price set by the decision-maker responds perfectly to monetary policy shocks Equation (94) characterizes the actual factormixdecision Bychoosingthecoefficients ζ and the signal-to-noise ratio χ, the decision-maker chooses the joint distribution of the profit-maximizing factor mix and the actual factor mix The fact that the decision-maker can only choose two coefficients in equation (94) may seem restrictive compared to equation (93), but we will show below that the firm cannot do better with a less restrictive choice in equation (94) The information flow constraint (95) restricts the amount of information that the actual decisions contain about the profit-maximizing decisions We follow Sims (23) and a large literature in information theory by quantifying information by reduction in uncertainty, where uncertainty is measured by entropy Let H (X) denote the entropy of the random vector X = (X,,X N ) Entropy is a measure of uncertainty Let H (X Y ) denote the conditional entropy of the random vector X =(X,,X N ) given knowledge of Y = (Y,,Y N ) Conditional entropy is a measure of conditional uncertainty The reduction 9 In equation (9), we use the fact that ˆp it ˆp it = p it p it 26

27 in uncertainty H (X) H (X Y ) is a measure of the amount of information that Y contains about X The operator I in the information flow constraint (95) is defined as I ({X t } ; {Y t })= lim T T + [H (X,,X T ) H (X,,X T Y,,Y T )] (96) The operator I measures the average per-period amount of information that the stochastic process {Y t } contains about the stochastic process {X t} Thus, the information flow constraint (95) states that the average per-period amount of information that the actual decisions contain about the profit-maximizing decisions cannot exceed the value of κ Note that we have assumed that the actual decisions follow a Gaussian process One can show that a Gaussian process for the actual decisions is optimal because the objective (89) is quadratic and the profit-maximizing decisions (9)-(92) follow a Gaussian process We have also assumed that the noise appearing in the actual decisions is firm-specific This assumption accords well with the idea that the friction is the decision-maker s limited attention rather than the availability of information Finally, we have assumed that the noise terms ν A it, νr it, νi it and νl ijt are independent of each other This assumption captures the idea that paying attention to the state of aggregate technology, paying attention to monetary policy disturbances, paying attention to firm-specific productivity and paying attention to relative wage rates are independent activities work in progress Relaxing this assumption is Two remarks are in place before we present solutions of the problem (89)-(95) First, when we solve the problem (89)-(95) numerically, we turn this infinite-dimensional problem into a finite-dimensional problem by parameterizing each infinite-order lag polynomial B (L) to B 3 (L) and C (L) to C 3 (L) as a lag-polynomial of an ARMA(p,q) process where p and q are finite Second, when a variable appearing in the information flow constraint (95) is (or may be) non-stationary, we replace the original variable by its first difference in the information flow constraint to ensure that entropy is always well defined See Sims (26) or Section VIIA in Maćkowiak and Wiederholt (29) 27

28 82 Computing the equilibrium of the model We use an iterative procedure to solve for the equilibrium of the model First, we make a guess concerning the process for the profit-maximizing price (9) and a guess concerning the process for the relative wage rate in equation (92) Second, we solve the firms attention problem (89)-(95) Third, we aggregate the individual prices to obtain the aggregate price level: p t = I IX p it (97) Fourth, we compute the aggregate dynamics implied by those price level dynamics Recall that in this section we assume that households have perfect information The households optimality conditions (66)-(68), equations (7)-(8), equation (2), y it = c it and the assumption that aggregate technology follows a first-order autoregressive process imply that the following equations have to be satisfied in equilibrium: i= c t = E t γ (r t p t+ + p t )+c t+, (98) w t = γc t + ψl t, (99) y t = c t, () y t = a t + αl t, () a t = ρ A a t + ε A t, (2) r t = ρ R r t +( ρ R ) φ π (p t p t )+φ y y t + ε R t (3) Here E t denotes the expectation operator conditioned on the entire history of the economy up to and including period t We employ a standard solution method for linear rational expectations models to solve the system of equations containing the price level dynamics and those six equations We obtain the law of motion for (c t, w t,y t,l t,a t,r t ) implied by the price level dynamics Fifth, we compute the law of motion for the profit-maximizing price The firms optimality condition (47) and equations (7), (76) and ˆp it = p it p t imply that the profit-maximizing price is given by p it = p t + α α + α α ϑc t + + α α ϑ w t 28 α + α α ϑ (a t + a it ) (4)

29 By substituting the law of motion for p t, c t, w t, a t and a it into the last equation, we obtain the law of motion for the profit-maximizing price In the last equation, we set ϑ = θ because the households optimality condition (68) and equations (7), (73) and (77) imply that the demand function for good i has the form (28)-(3) with a price elasticity of demand ϑ = θ If the process for the profit-maximizing price differs from our guess, we update the guess until a fixed point is reached Finally, we derive the equilibrium relative wage rates and the equilibrium factor mix Suppose that firmschooseavalueforζ that exceeds and a value for χ that is strictly positive Then, each firm can attain the profit-maximizing factor mix without any information flow Thus, no firm has an incentive to deviate The argument is the following Equation (94) and equations (72) and (78) imply that the labor demand function for each type of labor has the form (49)-(5) Since all households have exactly the same decision problem, all households set the same wage rate It follows from equation (76) that w t = w jt, or equivalently ŵ jt = Thus, in equilibrium the profit-maximizing factor mix is constant (ˆl ijt =), implying that each firmcanattaintheprofit-maximizing factor mix (ˆl ijt =) without any information flow 83 Benchmark parameter values and solution In this section, we report the numerical solution of the model for the following parameter values We set β =99, γ =, ψ =, θ =4, α =2/3 and η =4 To set the parameters governing the process for aggregate technology, equation (2), we consider quarterly US data from 96 Q to 26 Q4 We first compute a time series for aggregate technology, a t, using equation () and measures of y t and l t Weusethelog of real output per person, detrended with a linear trend, as a measure of y t Weusethelog of hours worked per person, demeaned, as a measure of l t We then fit equation (2) to thetimeseriesfora t obtaining ρ A =96 and a standard deviation of the innovation equal to 85 In the benchmark economy, we set ρ A =95 and we set the standard deviation of ε A t equal to 85 We use data for the non-farm business sector The source of the data is the website of the Federal Reserve Bank of StLouis 29