2.1. Households

Assuming that there are a large number of continuous households indexed by $i \in [0,1]$ in the economy. In general equilibrium, households are homogeneous in consumption, investment and bond holdings, but heterogeneous in wages and labor. Households maximize their utility by choosing consumption $C_{t}$ Footnote ¹ and labor supply $l_{t}\left (i\right )$ . A representative household’s utility function takes the form,

(1) \begin{align} E_{0}\sum \limits _{t=0}^{\infty }\beta ^{t}\left [\varepsilon _{c,t}\frac {\left (C_{t}-hS_{t-1}\right )^{1-\sigma _{c}}}{1-\sigma _{c}}-\frac {l_{t}(i)^{1+\sigma _{l}}}{1+\sigma _{l}}\right ] \end{align}

where ${\beta }$ is the household’s subjective discounting factor and $\sigma _{c}$ is the inverse of the consumption elasticity of substitution, which is also defined as coefficient of relative risk aversion. $\sigma _{l}$ is the inverse of the labor supply elasticity. $h$ measures the habit formation intensity of household effective consumption. The greater the habit formation intensity, the more residents tend to maintain the previous consumption level, and $h\in \left [0,1\right ]$ . $S_{t}$ denotes the stock of consumption habit and $S_{t}=\rho _{s}S_{t-1}+\left (1-\rho _{s}\right )C_{t}$ , which references Ravn et al. (Reference Ravn, Schmitt-Grohé and Uribe2006)’s setting. Here $\rho _{s}$ measures the extent to which the stock of habit is dependent on itself, i.e., it reflects the reliance of habit on the aggregate of past consumption. $S_{t}$ is normal (superficial) habit when $\rho _{s}=0$ and external deep habit when $\rho _{s}\in \left (0,1\right )$ . $\varepsilon _{c,t}$ is a consumer demand shock that follows a first-order autoregressive process.

The household budget constraint is given by

(2) \begin{align} \left (1{ +}\tau _{t}^{c}\right )C_{t}+I_{t}+b_{t}\le \left (1-\tau _{t}^{l}\right )w_{t}(i)l_{t}(i)+\left (\left (1-\tau _{t}^{k}\right )r_{t}^{k}u_{t}-\phi _{t}(u_{t})\right )k_{t-1}+\frac {R_{t-1}b_{t-1}}{\pi _{t}}+G_{TR,t} \end{align}

On the left side of the equation, $\tau _{t}^{c}$ is the consumption tax rate, $I_{t}$ is the private investment, and $b_{t}$ is the current government bond balance, which can also be seen as government debt. On the right side, $\tau _{t}^{l}$ denotes the labor tax rate, $w_{t}$ is the real wage, $\tau _{t}^{k}$ denotes the capital tax rate, $r_{t}^{k}$ is the capital return rate, and $k_{t}$ is the private capital stock. Refer to Kliem and Kriwoluzky (Reference Kliem and Kriwoluzky2014) to set the regulatory private capital utilization rate $u_{t}$ , and the cost of capital utilization is expressed in $\phi (u_{t})=\frac {\bar {r}^{k}\left (1-\bar {\tau }^{k}\right )}{\sigma _{u}}\exp\! \left (\sigma _{u}(u_{t}-1)-1\right )$ . In addition, $R_{t}$ is the nominal interest rate, $\pi _{t}$ is the inflation rate, and $\pi _{t}=P_{t}/P_{t-1}$ , where $P_t$ is nominal price. $G_{TR,t}$ denotes the government transfer payments.

We incorporate endogenous investor confidence variables into the model. Many studies within the framework of New Keynesian models consider investment efficiency shocks or investment demand shocks, which are typically integrated into the capital accumulation equation and follow an exogenous process (Justiniano et al. Reference Justiniano, Primiceri and Tambalotti2011; Kliem and Kriwoluzky, Reference Kliem and Kriwoluzky2014). Intuitively, investment demand and investment efficiency are associated with the investor confidence or beliefs of micro agents. The investor confidence of micro agents influences their investment preferences or trends, shapes the capital market, and affects the macroeconomy by distorting investment adjustment costs (Christiano et al. Reference Christiano, Eichenbaum and Evans2005; Bassanin et al. Reference Bassanin, Faia and Patella2021). Therefore, we introduce investor confidence into the capital accumulation equation, expressed as,

(3) \begin{align} k_{t}=\left (1-\delta \right )k_{t-1}+\left (\eta _{i,t}\right )^{\omega }\left [1-v\left (\frac {I_{t}}{I_{t-1}}\right )\right ]I_{t} \end{align}

where ${\delta }$ is the capital depreciation rate. Private investment is adjusted by the investment adjustment cost $v\left (\frac {I_{t}}{I_{t-1}}\right ){ =}\frac {\sigma _{I}}{2}\left (\frac {I_{t}}{I_{t-1}}-1\right )^{2}$ , with $\sigma _{I}$ representing the adjustment cost parameter. $\eta _{i,t}$ is the investor confidence, and $\omega$ is the investor confidence elasticity on the investment adjustment cost. Much of the literature sets up exogenous investment demand or investment efficiency shock for the same position of $\eta _{i,t}$ , and we endogenize such shock by assuming a confidence feedback mechanism. The setting of elasticity $\omega$ contributes to the estimation of the intensity of the confidence channel. In addition, considering that the interest rate spread $r_{t}^{sp}=r_{t}^{k}/R_{t}$ and the output volatility may have a greater impact on investor confidence in reality, we assume that the confidence feedback mechanism follows,

(4) \begin{align} \hat {\eta }_{i,t}{ =}\rho _{\eta }\hat {\eta }_{i,t-1}+\left (1-\rho _{\eta }\right )\left (\kappa _{y}\hat {y}_{t}^{GDP}+\kappa _{r}\hat {r}_{t}^{sp}\right )+\hat {\zeta }_{\eta, t} \end{align}

where $\rho _{\eta }$ is the investor confidence smoothing coefficient. The coefficient of investor confidence response to output ( $\kappa _{y}$ ) measures the intensity of investor confidence as affected by economic cycles, the coefficient of investor confidence response to spreads ( $\kappa _{r}$ ) measures the intensity of investor confidence as affected by investment spreads, and $\zeta _{\eta, t}$ is the investor confidence shock. In the next section, we provide empirical evidence for equation (4) and validate that $\kappa _{y},\kappa _{r}\gt 0$ which implies that the investor confidence fluctuates in the same direction as output and spreads.

2.2. Firms

2.2.1 Final good manufacturers

Assuming that the final good manufacturers in the economy face perfect competition, and use the Dixit-Stiglitz Aggregator production function to produce the final good $y_{t}$ :

(5) \begin{align} y_{t}=\left [\int _{0}^{1}y_{t}(j)^{\frac {\theta _{p}-1}{\theta _{p}}}dj\right ]^{\frac {\theta _{p}}{\theta _{p}-1}} \end{align}

where $y_{t}(j)$ is the intermediate goods produced by intermediate product manufacturer $j$ , and $\theta _{p}$ is the substitution elasticity between different intermediate products. The final good manufacturer decides the amount of intermediate goods input to maximize its profit $P_{t}y_{t}-\int _{0}^{1}P_{t}(j)y_{t}(j)dj$ , where $P_{t}(j)$ is the price level of intermediate good $y_{t}(j)$ . By solving for the above profit maximization, the demand function for intermediate good can be obtained as $y_{t}(j)=\left (\frac {P_{t}(j)}{P_{t}}\right )^{-\theta _{p}}y_{t}$ , and price level of the final good is $P_{t}=\left (\int _{0}^{1}P_{t}(j)^{1-\theta _{p}}dj\right )^{1-\theta _{p}}$ .

2.2.2 Intermediate good manufacturers

Assuming that a continuum of intermediate good manufacturers, indexed by $j\in \left [0,1\right ]$ , face monopolistic competition in the economy, and produce intermediate products utilizing public capital provided by the government along with private capital and labor supplied by the household sector, the production function can be expressed as:

(6) \begin{align} y_{t} (j)=\left (k_{g,t-1} (j)\right )^{\alpha _{g}}\left (u_{t}k_{t-1} (j)\right )^{\alpha _{p}}\left (l_{t}(j)\varepsilon _{z,t}\right )^{1-\alpha _{g}-\alpha _{p}}-\Omega \end{align}

where $k_{g,t-1}(j)$ , $k_{t-1}(j)$ , $l_{t}(j)$ are respectively the public capital demand, private capital demand, and labor demand of intermediate product manufacturer $j$ . $\alpha _{g}$ is the elasticity of public capital to output, $\alpha _{p}$ is the elasticity of private capital to output, and $(1-\alpha _{g}-\alpha _{p})$ is the elasticity of labor supply to output, $\Omega$ representing the fixed cost of production. $\varepsilon _{z,t}$ is the labor productivity shock that obeys the first-order autoregressive process. In addition, public capital is accumulated by government investment through:

(7) \begin{align} k_{g,t}=\left (1-\delta \right )k_{g,t-1}+\left [1-v\left (\frac {G_{I,t}}{G_{I,t-1}}\right )\right ]G_{I,t} \end{align}

Referring to Calvo (Reference Calvo1983)’s staggered pricing mechanism, we introduce price stickiness for intermediate good manufacturers. That is, rather than being able to freely set or change prices solely based on inflation, each intermediate goods manufacturer has only a probability of $\left (1-{{\gamma _{p}}}\right )$ of adjusting its prices in each period.

2.3. Government

The government finances spending through the labor tax $x_{l,t}$ , capital tax $x_{k,t}$ , consumption tax $x_{c,t}$ , and government bonds $b_{t}$ , with the budget constraint as follows,

(8) \begin{align} G_{C,t}+G_{I,t}+G_{TR,t}+\frac {R_{t-1}b_{t-1}}{\pi _{t}}=x_{l,t}+x_{k,t}+x_{c,t}+b_{t} \end{align}

where $x_{l,t}=\tau _{t}^{l}w_{t}l_{t}$ , $x_{k,t}=\tau _{t}^{k}r_{t}^{k}u_{t}k_{t-1}$ and $x_{c,t}=\tau _{t}^{c}C_{t}$ . The government bonds $b_t$ is endogenously determined by the government budget constraint. Referring to Leeper et al. (Reference Leeper, Traum and Walker2017), the tax rate feedback rules are as follows,

(9) \begin{align} \hat {\tau }_{t}^{l}=\rho _{l}\hat {\tau }_{t-1}^{l}+\left (1-\rho _{l}\right )\kappa _{b}^{l}\hat {b}_{t-1}+\hat {\zeta }_{\tau ^{l},t} \end{align}

(10) \begin{align} \hat {\tau }_{t}^{k}=\rho _{k}\hat {\tau }_{t-1}^{k}+\left (1-\rho _{k}\right )\kappa _{b}^{k}\hat {b}_{t-1}+\hat {\zeta }_{\tau ^{k},t} \\[-20pt] \nonumber \end{align}

(11) \begin{align} \hat {\tau }_{t}^{c}=\rho _{c}\hat {\tau }_{t-1}^{c}+\left (1-\rho _{c}\right )\kappa _{b}^{c}\hat {b}_{t-1}+\hat {\zeta }_{\tau ^{c},t} \\[-8pt] \nonumber \end{align}

where ${\rho _{l}}$ , ${\rho _{k}}$ and ${\rho _{c}}$ denote the smoothing parameter of labor tax, capital tax and consumption tax, respectively. We assume that tax rates will be adjusted according to the economic cycle and government debt, thus $\kappa _{y}^{l}$ , $\kappa _{y}^{k}$ , and $\kappa _{y}^{c}$ denote three types of response coefficients to output, while $\kappa _{b}^{l}$ , $\kappa _{b}^{k}$ , and $\kappa _{b}^{c}$ denote three types of response coefficients to government debt. The labor tax rate shock $\hat {\zeta }_{\tau ^{l},t}$ , capital tax rate shock $\hat {\zeta }_{\tau ^{k},t}$ and consumption tax rate shock $\hat {\zeta }_{\tau ^{c},t}$ are independent and identically distributed.

In addition, discretionary fiscal policy includes government consumption, government investment and transfer payments. The three fiscal policies all follow the exogenous first-order autoregressive process evolution,

(12) \begin{align} {\hat {G}_{C,t}=\rho _{gc}\hat {G}_{C,t-1}+\hat {\zeta }_{gc,t}} \\[-20pt] \nonumber \end{align}

(13) \begin{align} {\hat {G}_{I,t}=\rho _{gi}\hat {G}_{I,t-1}+\hat {\zeta }_{gi,t}} \\[-20pt] \nonumber \end{align}

(14) \begin{align} \hat {G}_{TR,t}=\rho _{tr}\hat {G}_{TR,t-1}+\hat {\zeta }_{tr,t} \\[-8pt] \nonumber \end{align}

where ${\rho _{gc}}$ , ${\rho _{gi}}$ and ${\rho _{tr}}$ denote the smoothing parameters of government consumption, government investment and transfer payments, respectively. The three types of policy shocks $\hat {\zeta }_{gc,t}$ , $\hat {\zeta }_{gi,t}$ and $\hat {\zeta }_{tr,t}$ are independent and identically distributed.

Assuming that the monetary department of the government adopts a price-based monetary policy, with the nominal interest rate serving as the means off regulation, this policy is embodied in the standard Taylor rule, which is articulated as follows:

(15) \begin{align} \hat {R}_{t} = \rho _{R}\hat {R}_{t-1} +\left (1-\rho _{R}\right )\left (\rho _{\pi }\hat {\pi }_{t}+\rho _{y}\hat {y}_{t}^{GDP}\right )+\hat {\zeta }_{m,t} \end{align}

where $\rho _R$ denotes the nominal interest rate smoothing parameter, ${\rho _{\pi }}$ and ${\rho _{y}}$ denote the response coefficients of nominal interest rate to inflation and output, respectively. $\hat {\zeta }_{m,t}$ denotes the monetary policy shock, which obeys independent and identical distribution.

2.4. Market clearing and shocks

In equilibrium, the market clearing condition is given by

(16) \begin{align} y_{t}=C_{t}+I_{t}+G_{C,t}+ G_{I,t} + \phi \left (u_{t}\right )k_{t-1} \end{align}

The GDP adjusted by the cost of capital utilization is expressed as $y_{t}^{GDP}$ , which follows

(17) \begin{align} {y_{t}^{GDP}=y_{t}-\phi \left (u_{t}\right )k_{t-1}} \end{align}

Moreover, the consumption demand shock ( $\varepsilon _{c,t}$ ) and labor productivity shock ( $\varepsilon _{z,t}$ ) follow the exogenous first-order autoregressive process,

(18) \begin{align} \hat {\varepsilon }_{c,t}=\rho _{cd}\hat {\varepsilon }_{c,t-1}+\hat {\zeta }_{cd,t} \\[-20pt] \nonumber \end{align}

(19) \begin{align} \hat {\varepsilon }_{z,t}=\rho _{z}\hat {\varepsilon }_{z,t-1}+\hat {\zeta }_{z,t} \\[-8pt] \nonumber \end{align}

where $\rho _{cd}$ is the consumption demand shock smoothing parameter, $\rho _{z}$ is the productivity shock smoothing parameter. $\hat {\zeta }_{cd,t}$ and $\hat {\zeta }_{z,t}$ obey independent and identical distribution.

2.5. Fiscal multiplier definition

This paper refers to the method of Leeper et al. (Reference Leeper, Traum and Walker2017) to describe the present value multipliers of the five types of fiscal policies: government consumption, government investment, labor tax, capital tax, and consumption tax. The formula for calculating the present value multiplier is as follows,

(20) \begin{align} \psi _{k} =\frac {E_{t}\Sigma _{j=0}^{k}\left (\Pi _{i=0}^{j}\left (R_{t+i}^{r}\right )^{-1}\right )\Delta Y_{t+j}}{E_{t}\Sigma _{j=0}^{k}\left (\Pi _{i=0}^{j}\left (R_{t+i}^{r}\right )^{-1}\right )\Delta G_{t+j}} \end{align}

where $\psi _{k}$ is the present value multiplier in period $k$ , $R_{t}^{r}$ is the real interest rate, ${\Delta Y_{t + j}}$ is the deviation value of the $j$ -period output from a given fiscal policy shock, and $\Delta G_{t+j}$ is the deviation value of the $j$ -period fiscal policy impact on the same unit.

3.1. Empirics for investor confidence feedback mechanism

We begin by providing empirical evidence on the investor confidence channel. The investor confidence feedback mechanism in equation (4) can be written in the following form for empirical estimation:

(21) \begin{align} \hat {\eta }_{i,t}{ =}\rho _{\eta }\hat {\eta }_{i,t-1}+\left (1-\rho _{\eta }\right )\left (\kappa _{y}\hat {y}_{t}^{GDP}+\kappa _{r}\left (\hat {r}_{t}-\hat {R}_{t}\right )\right )+\hat {\epsilon }_{\eta, t} \end{align}

where $\hat {\epsilon }_{\eta, t}$ is the error term, and the other notations have the same meaning as in equation (4). Given the strong endogeneity problem in the above equation, we select two methods, GMM and Two-Stage Least Squares (2SLS), to eliminate the effect of endogeneity on model estimation. In terms of data selection, considering that all private investment by households in the model serves for production, investor confidence is in fact production confidence as well. Thus the investment confident volatility ( $\hat {\eta }_{i,t}$ ) is measured by manufacturing production tendency in Business Tendency and Consumer Opinion Surveys from the OECD, and this quarterly data is expressed in units of deviation rate. The investment return rate is measured by the quarterly return rate of Dow Jones Industrial Average. The nominal interest rate is measured by the effective federal funds rate (EFFR) from the Federal Reserve Board. We convert this monthly data to quarterly data by taking the mean. The GDP is the quarterly gross domestic product from NIPA.

In this case, instrumental variables are selected to model explanatory variables lagged from 1 to 6 periods. The GMM model is based on the Heteroskedasticity and Autocorrelation Consistent Weighting Matrix to obtain robust standard errors for the parameters (see Table 1). From the results of the Hansen J-statistic test, it can be seen that the P-values corresponding to the two methods can’t reject the original hypothesis at the 5% significance level which indicates that there is no over-identification of instrumental variables, therefore, the selection of instrumental variables in this paper is reasonable and effective. Meanwhile, the D.W. statistics of the two methods are close to 2, indicating that there is no autocorrelation in the residuals of the models and the estimation results are robust and reliable.

Table 1.

Investor confidence feedback mechanism

Note: The robust standard errors are in parentheses $(\cdot )$ . P-values are in the parentheses $[\cdot ]$ . Instrumental variables: explanatory variables lagged from period 1 to 6. The notation $^{***}$ indicates rejection of the null hypothesis at the 1% significance level.

From the estimated parameters for the whole samples from 1960 to 2023, all the estimated parameters reject the null hypothesis of zero at a significance level of 1%. For the GMM and 2SLS models, the estimated parameters of rho are significantly to be 0.7033 and 0.7151, respectively, indicating that investor confidence has strong smoothness. The parameters for targeting to investor excess returns are significantly to be 0.0146 and 0.0192, while the parameters for targeting to economic cycles are significantly to be 0.1820 and 0.1940, respectively. It can be seen that investor confidence has a strong feedback mechanism for targeting to economic cycles and excess returns. In addition, the conclusions of the samples from 1960 to 2008 do not significantly change the estimated results compared to the whole sample, showing that the setting of the investor confidence feedback mechanism is reasonable.

3.2. Calibration

For the parameters commonly used in the model, this paper refers to the existing literature or estimates based on actual data. We refer to McManus et al. (Reference McManus, Ozkan and Trzeciakiewicz2021) and set the US household discount factor $\beta$ to 0.995. We firstly assume that normal (superficial) habit consistent with Kliem and Kriwoluzky (Reference Kliem and Kriwoluzky2014) and Leeper et al. (Reference Leeper, Traum and Walker2017), thus setting $\rho _{s}$ to 0, and changed $\rho _{s}$ in subsequent study to examine deep habit similar to Ravn et al. (Reference Ravn, Schmitt-Grohé and Uribe2006). Referring to Leeper et al. (Reference Leeper, Traum and Walker2017) and McManus et al. (Reference McManus, Ozkan and Trzeciakiewicz2021), the capital output elasticity ${\alpha _{p}}$ is set to 0.33, and the public capital output elasticity ${\alpha _{g}}$ is set to 0.02. Referring to Kliem and Kriwoluzky (Reference Kliem and Kriwoluzky2014), we set the price markup ${\theta _{p}}$ to 6, the wage markup ${\theta _{w}}$ to 11, and the capital depreciation rate ${\delta }$ to 0.025.

We estimate the three coefficients ( $\rho _{\eta }$ , $\kappa _{r}$ and $\kappa _{y}$ ) for the investor feedback mechanism by using methods GMM and 2SLS in the previous subsection, and we now take the average of the estimates from the two methods as a calibration value for the DSGE model. With respect to the steady-state value, we use the U.S. data from 1960 to 2023 (Model I) and from 1960 to 2008 (Model II) to calculate the average quarterly nominal interest rate $\bar {R}$ , the share of government consumption to output ${\bar {G}_C}/\bar {y}$ , the share of government investment to output ${\bar {G}_I}/\bar {y}$ , and the share of transfer payments to output ${\bar {G}_{TR}}/\bar {y}$ , the labor tax rate $\bar {\tau }^{l}$ , the capital tax rate $\bar {\tau }^{k}$ , and the consumption tax rate $\bar {\tau }^{c}$ , respectively. Table 2 summarizes the calibration results of the above parameters.

Table 2.

Calibration of some parameters

3.3. Bayesian estimation

We estimate remaining dynamic parameters by Bayesian estimation method, and we refer to previous studies to set the prior distribution (Leeper et al. Reference Leeper, Plante and Traum2010; Kliem and Kriwoluzky, Reference Kliem and Kriwoluzky2014; Leeper et al. Reference Leeper, Traum and Walker2017; McManus et al. Reference McManus, Ozkan and Trzeciakiewicz2021). Specifically, the prior setting of habit formation intensity, inverse of the consumption elasticity, and inverse of the labor supply elasticity are referring to Leeper et al. (Reference Leeper, Plante and Traum2010). The priors of investment adjustment cost and capital utilization cost then follow Kliem and Kriwoluzky (Reference Kliem and Kriwoluzky2014). For setting of price stickiness and the Taylor rule, we refer to Leeper et al. (Reference Leeper, Traum and Walker2017)’s priors as well as estimated intervals. We set the prior mean of confidence channel intensity to be 1 based on the assumptions of the model, and we further assume that its distribution and standard deviation are Gamma and 0.1. In the literatures, the prior mean and estimated levels of the tax rate response to debt range between 0 and 0.5 (Leeper et al. Reference Leeper, Plante and Traum2010; Kliem and Kriwoluzky, Reference Kliem and Kriwoluzky2014; McManus et al. Reference McManus, Ozkan and Trzeciakiewicz2021). We therefore set the prior mean of our three types of tax rate response to debt to 0.25, and their distribution and standard deviations are consistent with Kliem and Kriwoluzky (Reference Kliem and Kriwoluzky2014). Finally, the remaining priors of the AR parameters and the exogenous shocks also follow Kliem and Kriwoluzky (Reference Kliem and Kriwoluzky2014).

The actual observation data used in Bayesian estimation mainly include the output, government consumption, government investment, private investment, labor tax, capital tax, and consumption tax. In Bayesian estimation, all observations are logged, first-order differenced, and then the mean is subtracted. The Metropolis-Hastings algorithm is used to randomly sample 40,000 times. Table 3 shows the prior setting and Bayesian estimation results.

Table 3.

The prior setting and posterior results of Bayesian estimation

4.1. Multipliers in the benchmark model

We firstly measure five types of fiscal multipliers for 1960–2023 (Model I). An excessively long horizon weakens the significance of the study and the reference value of the policy, thus we report multipliers over 80 quarters as in Leeper et al. (Reference Leeper, Traum and Walker2017). The results are presented in Table 4, which also show the effects of fiscal policy on private consumption $\Delta C$ and private investment $\Delta I$ in order to illustrate and compare the fiscal effects more clearly.

Table 4.

Present value multipliers of fiscal policies in Model I: 1960–2023

Firstly, we found that the government consumption multiplier of GDP showed a decline after the shock, indicating a more pronounced contribution of government consumption to output in the short run. As a result of changes in the real interest rate and higher distortionary taxes under expansionary government spending, then the boost to income from government consumption fails to outweigh the extent to which it stimulates aggregate demand, thereby crowding out households’ spending. In the long run, the effect of government consumption is diluted by the crowding-out effect of private spending, and the boost to the economy declines significantly.

Secondly, the government investment multiplier of GDP is a U-shaped curve that first declines and then rises. This is because an increase in government investment may push up the real interest rate and, at the same time, prompt an increase in distortionary taxes, both of which together crowd out the level of private spending. Then the policy effect of government investment diminishes slightly in the short term. In addition to directly affecting aggregate demand, government investment also acts on aggregate supply through the accumulation of public capital, thereby the cumulative effect of positive shocks to government investment gradually emerging over time to enhance output. The results corroborate the view of Bachmann and Sims (Reference Bachmann and Sims2012) that government investment growth is conducive to future productivity growth and is more persistent in its effect on output. In addition, the productive government investment multiplier is close to the unproductive government consumption multiplier in the early stages, but is much larger than the government consumption multiplier in the later stages.

Thirdly, the labor and capital tax multipliers of GDP are negative and significantly decreasing in the short to medium term, suggesting that an increase in private income taxation has a contractionary effect on the economy and that the effect is more pronounced in the long term. Both labor and capital taxes crowd out disposable income. Labor taxes dampen the willingness of households to work and contract the labor market. Capital taxes, on the other hand, significantly discourage investment, and the reduction in private investment leads to a gradual decline in the level of capital accumulation, thus presenting an increasingly pronounced dampening effect on output.

Finally, the consumption tax multiplier of GDP is also negative and tends to fall and then rise in the 20-quarter perspective, indicating that the policy effect of the consumption tax increases in the short term but weakens in the medium to long term. This is because the rise in the consumption tax rate cuts the level of private consumption, but due to the presence of habit formation, private consumption does not fall much at the moment of the tax shock, and continues to decay after the shock has occurred, so that the contractionary effect of the consumption tax on the economy becomes more pronounced in the short term. As households adjust their spending decisions to shift some of the reduced consumption to investment, private capital increases and mitigates the degree of output decline, weakening the negative effect of consumption tax in the medium to long term.

We then measure five types of fiscal multipliers for 1960–2008 (Model II) and present them in Table 5. Compared to Table 4, it can be seen that the two government spending multipliers decline to a faster extent, probably due to the more active financing properties of the distortionary taxes during 1960–2008. Although the government investment multiplier cannot rise in the long run, it declines much less than the government consumption multiplier, which also suggests that the effect of productive government spending is more durable. It also clearly shows that three distortionary taxes play a more prominent role as automatic stabilizers in 1960–2008, and thus their multipliers converge to positive values; that is, tax financing has positive benefits in the forward period.

Table 5.

Present value multipliers of fiscal policies in Model II: 1960−2008

In addition, we carefully compared our multiplier results with previous studies, such as Ramey (Reference Ramey2019) et.al, which is organized as Table C.1. Much of the literature demonstrates changes in multipliers over a 20-quarter or 40-quarter period, as too long a horizon would make the study less meaningful. We therefore show, in the last sample in Table C.1, the range of our benchmark multipliers within 40 quarters. From previous studies, the range of the multipliers is strongly influenced by the model setting, with government spending multipliers ranging from 0.05 to 1.5, tending to be larger in the initial period and lower in the long run. At the same time, some literature distinguishes between unproductive government consumption and productive government investment (or infrastructure investment), and the result is generally that the government investment multiplier is expected to be higher than the government consumption multiplier (Ramey, Reference Ramey2019; McManus et al. Reference McManus, Ozkan and Trzeciakiewicz2021). Our estimates of the two government spending multipliers support this conclusion and are within a reasonable range. On the other hand, the existing literature has a wide range of estimates on tax multipliers, from −1 to 0.6. Most of the literature measures negative short-term tax multipliers. However, in a debt- or deficit-financed environment, taxes provide a source of revenue for the government, mitigating the effects of debt and deficits in the long run, and therefore the tax multiplier may turn positive at a later stage. Similarly, our estimated tax multipliers are consistent with the level of estimates in the existing literature. In summary, the multipliers in this paper are reasonable and credible compared to the existing literature.

4.2. Multipliers under the liquidity trap

The fiscal impact on household expenditures is obviously related to interest rate, which affects the multipliers to a greater extent. Therefore, we try to close the monetary policy channel in order to further investigate the importance of the interest rate transmission effect for fiscal policy. The presence of a liquidity trap in a zero lower bound (ZLB) constraint invalidates the central bank’s ability to regulate the economy through traditional monetary policy, i.e., micro agents’ preferences affect the effectiveness of the monetary policy (Eggertsson and Woodford, Reference Eggertsson and Woodford2003; Woodford, Reference Woodford2003). Based on this, we extend the study of fiscal multipliers based on the dimension of the liquidity trap. Referring to Guerrieri and Iacoviello (Reference Guerrieri and Iacoviello2015) and Caramp and Silva (Reference Caramp and Silva2023), we consider that the equilibrium economy suffers from a liquidity trap lasting for $T^{*}$ periods. When $T^{*}=0$ , there is no ZLB constraint in the economy. And when $T^{*}\gt 0$ , the interest rate rule in equation (15) is replaced by:

(22) \begin{align} {\hat {R}_{T}{ =}0;\qquad T=0,1,\ldots, T^{*}-1} \\[-20pt] \nonumber \end{align}

(23) \begin{align} \hat {R}_{t+T^{*}}{ =}\rho _{R}\hat {R}_{t+T^{*}-1}{ +}\left (1-\rho _{R}\right )\left (\rho _{\pi }\hat {\pi }_{t}+\rho _{y}\hat {y}_{t}^{GDP}\right )+\hat {\zeta }_{m,t} \\[-8pt] \nonumber \end{align}

Given that the interest rates in the U.S. have fallen to near zero since the first quarter of 2009 (which implies the occurrence of ZLB), we estimate the model based on the 1960–2008 sample period (Model II) to exclude the realistic effect of ZLB on the model measurement. Getting rid of the sample period after 2009 helps to better conjecture the economic status before the ZLB and more accurately simulate the movement of fiscal multipliers when the ZLB occurs in that environment. We then compare the multipliers for the liquidity trap maintained for 0 (no ZLB), 4 quarters, and 8 quarters in Model II, as shown in Figure 1.

Figure 1.

Multipliers under ZLB in Model II’s estimated level.

It can be seen that the multiplier effects of all fiscal instruments are dramatically amplified, both within and after the ZLB. The results are similar to the findings of existing studies (Christiano et al. Reference Christiano, Eichenbaum and Rebelo2011; Miyamoto et al. Reference Miyamoto, Nguyen and Sergeyev2018; Kilponen et al. Reference Kilponen, Pisani and Schmidt2019) that the ZLB leads to an expansion of the fiscal effect. Specifically, the government consumption multiplier is less than 1 at the baseline, but is substantially greater than 1 within the ZLB, which is also in line with Christiano et al. (Reference Christiano, Eichenbaum and Rebelo2011)’s conclusion. The reason is that, due to the failure of the Taylor rule under ZLB, the nominal interest rate is unable to rise along with inflation induced by the government consumption shock, resulting in a sharp fall in the real interest rate. Since households earn less interest on their savings, they reduce savings and choose to spend more. Consequently, the crowding-out effect of government consumption on household spending is substantially reduced or transformed into a crowding-in effect. This greatly amplifies the government consumption multiplier. Our specifications of fiscal instruments do not affect this conclusion, as both government investment, which is productive in nature, and the three types of taxes, which have negative multipliers, have substantially amplified their original effects in the ZLB. Thus, the ZLB changes private spending preferences by distorting the interest rate transmission mechanism of fiscal policy, ultimately amplifying the fiscal shocks, which turns out to be very robust, at least for the estimation based on the data of 1960–2008.

The above results raise another question: does the existence of ZLB surely strengthen the fiscal policy multiplier? From the existing literature and our benchmark analysis, it is significant that the ZLB amplifies the fiscal multiplier through the interest rate effect. But imagine if micro agents (households) had been able to anticipate the end of the ZLB: the central bank had quickly reverted to the Taylor rule, and the nominal interest rate had moved rapidly higher in line with inflation, somewhat leaving a contraction in the economy. In that case, micro agents may have scaled back their expenditures and shifted to more savings during the ZLB in order to withstand the post-ZLB economic situation and prevent the arrival of the nominal interest rate hikes. Micro agents’ expectations of the extent of future interest rate hikes can also influence their behavior preferences at the moment. This precautionary saving effect may be implicit in the impact of the ZLB on the multiplier, but is not perceived by the benchmark simulation in Figure 1. To check this hypothesis, we increase the coefficient of interest rate response to inflation $\rho _{\pi }$ from the estimation level 1.0047 to counterfactual 1.5 in equation (23), thereby increasing the extent of interest rate hikes after the end of the ZLB. The equation (23) confers the ability for future nominal interest rates to be able to affect the current economy through expected inflation and be perceived by micro agents. Our intuition comes from the fact that private expectations with a more active Taylor rule profoundly influence the occurrence of the liquidity trap (Benhabib et al. Reference Benhabib, Schmitt-Grohé and Uribe2002). We then display the multipliers results of the counterfactual raising ${\rho _{\pi }}$ in Figure 2.

Figure 2.

Multipliers under ZLB with counterfactual higher $\rho _{\pi }$ .

Under the counterfactual simulations in Figure 2, the expansionary effect of the ZLB on fiscal multipliers is dramatically weakened. For the government consumption, government investment, and labor tax, the multiplier effect contracts during the ZLB, which is contrary to existing research and Figure 1’s results. This validates the existence of a precautionary saving effect, that is, if micro agents anticipate the end of the ZLB and a rise in the nominal interest rate through policy inertia or by some means (e.g., the Federal Open Market Committee), then they have reduced spending and increased savings during the ZLB. When the post-ZLB Taylor rule is more active and can be perceived by micro agents, the effect of precautionary saving is much stronger and leads to a substantial weakening of the fiscal effect. As a result, the ZLB impacts on the fiscal multipliers is divided into two paths: (i) the interest rate transmission effect, which is able to reduce households’ savings through a lower real interest rate, ultimately encouraging private spending and amplifying the fiscal multipliers; (ii) and the precautionary savings effect, which influences private behavior preferences via expectations of interest rate hikes, finally increasing the level of private savings and weakening the multipliers. The tradeoff between the two paths together shapes the fiscal multipliers under the ZLB. In the baseline estimation, Figure 1, the interest rate transmission effect dominates, while precautionary saving is not significant, contributing to a dramatic expansion of the multiplier effect. However, in the Figure 2’s economic system with stronger policy inertia (more active Taylor rule), the precautionary saving effect caused by the ZLB may be larger, resulting in a contraction of the fiscal multipliers.

To explore the mechanism of the ZLB’s impact on the multipliers even further, Figure 3 shows the impulse response functions of a positive shock of government consumption, with a shock size of 1% from its steady-state. We compare different ZLB cases from a perspective of the government consumption shock. At the estimation level, the ZLB significantly dampens the crowding-out effect of government consumption shocks on private consumption and leads to a crowding-in of private investment, which further reduces private saving through the interest rate transmission effect, resulting in a stronger stimulus to output from government consumption. When the Taylor rule is more active ( $\rho _{\pi }=1.5$ ), the ZLB instead leads to a rise in private savings. In this case, private spending is crowded out more by government consumption, weakening the increase in output. At the same time, the economy experienced lower levels of inflation during the liquidity trap as in Mertens and Ravn (Reference Mertens and Ravn2014). Thus, the ZLB is able to contract the economy with aggregate demand shocks through the precautionary saving effect.

Figure 3.

The government consumption shock and IRFs under different ZLB cases.

Our results help explain two things about the real data (as in Figure C.1): the rapid rise in nominal interest rates after the ZLB, and the fact that the private saving ratio stayed high during the ZLB (especially in 2020). Overall, our study broadens the channels through which the ZLB affects the fiscal multiplier and provides new perspectives based on private preferences and precautionary saving motivation.

5.1 Habit and multipliers

5.1.1. Habit formation intensity

The habit formation intensity $h$ in this paper is estimated to be 0.5935 for Model I (1960–2023) and 0.6189 for Model I (1960–2008). According to Havranek et al. (Reference Havranek, Rusnak and Sokolova2017), the mean value of habit formation in the macro literature is 0.58, while the value estimated using the DSGE method is 0.67, and thus our habit formation estimates are in line with this range of estimates. To investigate the impact of changes in habit formation on the effects of fiscal policy, we counterfactually simulate the absence of habit formation ( $h=0$ ) or sufficiently large habit formation ( $h=0.9$ ) and analyze differences in fiscal multiplier curves for benchmark Model I. The simulation results are shown in Figure 4.

Figure 4.

Fiscal multipliers at different habit formation intensities.

Firstly, under a positive government consumption shock, an increase in the habit formation intensity leads to a rise in the multiplier of government consumption, implying a more pronounced stimulative effect of government consumption on GDP. It appears that government consumption shocks affect household expenditure through two channels: On the one hand, government consumption increases aggregate demand and raises inflation. Changes in the real interest rate, which is a measure of the risk-free rate of return for households, have crowding out or crowding in effects on private spending based on the activation of the Taylor rule. On the other hand, expansionary fiscal policy has triggered a rise in government debt, causing distortionary taxes to increase and crowding out household disposable income. Under both channels, the initial boost to output from government consumption is significant, but will diminish significantly as private spending falls, resulting in the multiplier showing a decline from the moment of the shock to the long term. In this process, the persistence of habit formation smooths out the effect of government consumption shocks on household consumption through the interest rate mechanism. The greater the habit formation intensity, the less households’ consumption will be affected by the shock. When the habit formation is sufficiently high ( $h=0.9$ ), the government consumption multiplier is higher than the benchmark estimate in all periods. Furthermore, at the time level, the enhancement of habit formation on the government consumption effect is more pronounced within 20 quarters, while after 40 quarters, the increasing habit formation only leads to a slight increase in the government consumption multiplier. This is because habit formation cannot smooth out the crowding effect of the interest rate transmission mechanism on private investment. Therefore, the presence of habit formation in the short run can affect the government consumption multiplier to a large extent, while in the long run, habit formation has a weaker effect on the government consumption multiplier.

Secondly, the government investment multiplier is also positively correlated with the strength of habit formation. In addition to causing an increase in aggregate demand, productive government investment also directly affects aggregate supply through public accumulation capital. The presence of habit formation also weakens the crowding-out effect of government investment on private consumption. As a result, habit formation in the early period substantially enhances the policy effect of positive government investment shocks. This difference in government investment multipliers due to habit formation diminishes over time because habit formation discourages households from adjusting their consumption levels too much in the face of shocks. Instead, changes in private expenditure are realized more by adjusting private investment; that is, the smoothing of household consumption by habit formation is converted into a high sensitivity of household investment to shocks. As private investment adjusts more, the spread in the government investment multiplier induced by habit gradually narrows until the economy returns to a steady state. Again, the results indicate the importance of habit formation as a measure of the short-term effects of government investment.

Thirdly, for labor taxes, the occurrence of a positive tax rate shock will inevitably lead to a decline in disposable labor income and reduce the willingness of private spending, resulting in a significant reduction in the demand for consumption and investment in the economy, which in turn will lower the level of output. After the fall in labor disposability, even if the households want to earn more capital returns by investing, they have no source of funds to compensate for this, and besides, capital returns are inherently lagging. Therefore, the reduction in labor income caused by the labor tax rate shock is difficult to compensate for, and the effect on private consumption is significant. Since the presence of habit formation mitigates the dampening effects of declining labor income on consumption, higher habit formation helps to mitigate the initial negativity of the labor tax multiplier. Similarly, the effect of habit formation on the labor tax multiplier diminishes over time.

Fourthly, for capital taxes, the multiplier trend is similar to the labor tax multiplier, but the negative effect is greater. Clearly, the very small effect of habit formation on the capital tax multiplier suggests a relatively weak effect of capital taxes on private consumption. It basically only affects private investment and is almost independent of habit formation.

Finally, in contrast to capital taxes, the consumption tax effect is significantly influenced by habit formation. Habit formation effectively weakens the consumption tax multiplier effect and maintains this effect in the long run. The reason for this is that, unlike labor and capital taxes, the size of consumption tax revenues depends directly on the households’ consumption. When habit formation is absent ( $h=0$ ), the elasticity of household consumption in response to the consumption tax is very high, and households will counteract the impact of consumption tax rate shock by shifting their spending to investment, which rises much less than the reduction in consumption, and eventually the consumption tax multiplier will be reflected in a negative value. The increase in habit formation intensity irons out the volatility of the consumption tax for changes in the expenditure side of the population and attenuates the policy effect of changes in the consumption tax rate. When the intensity of habit formation is sufficiently high ( $h=0.9$ ), the multiplier of the consumption tax is positive after 30 quarters.

5.1.2. Deep habit

In the benchmark model, we set $\rho _s=0$ and assume a normal (superficial) habit consistent with Kliem and Kriwoluzky (Reference Kliem and Kriwoluzky2014) and Leeper et al. (Reference Leeper, Traum and Walker2017). Here, we also consider a external and deep habit, and set $\rho _{s}=0.85$ with reference to Ravn et al. (Reference Ravn, Schmitt-Grohé and Uribe2006) to compare the effect of different types of habit on multipliers. The results are shown in Figure 5.

Figure 5.

Multipliers under different types of habits.

It is clear that the multipliers of government consumption, government investment and consumption tax are generally smoother in the deep habit compared to the superficial habit, suggesting that deep habits lead to further flattening of households’ consumption fluctuations. Theoretically, deep habits, based on the stock of all past consumption, have a stronger smoothing effect on private consumption than superficial habits, which would consequently make the multipliers smoother. However, the transformation from superficial to deep habits does not change the nature of habit and the overall values of five types of multipliers.

Moreover, Figure 6 provides the impulse response functions of a 1%-size positive government consumption shocks with different types of habit. In terms of impulse response, the deep habits do significantly reduce the crowding-out effect of private consumption from government consumption shocks. At the same time, government consumption crowds out private investment more, which largely offsets the effect of deep habits on private consumption. Since households offset the deep habit effect through portfolio rebalancing, the overall multipliers do not change significantly.

Figure 6.

The government consumption shock and IRFs with different types of habits.

The above results suggest that habit formation amplifies the multiplier effect of both types of government spending, while it weakens the policy effect of labor and consumption taxation, but hardly affects the output effect of capital tax. Moreover, habit formation has different mechanisms and maintenance periods for different fiscal policies. Compared to the normal (superficial) habits, the external and deep habits lead to flatter multipliers. However, due to the portfolio rebalancing of households, the difference in multiplier values due to changes in the habit types is modest overall. Totally, the importance of habit formation in an economy cannot be ignored, and its existence and intensity are closely relevant to the measurement of fiscal multipliers.

5.2. Investor confidence and multipliers

In addition to consumption, households also play the role of investors in the economy, and their investment confidence is inevitably related to economic fluctuations and policy effects. Based on the endogenous system of investor confidence influenced by spreads and economic cycles incorporated in the model, we simulate changes in the relevant parameters to capture the policy effects and better enrich the micro-foundations of macroeconomic policies.

5.2.1. The role of investor confidence channel

Investor confidence shocks can drive longer-term isotropic fluctuations in output (see Appendix C.3). The investor confidence channel intensity $\omega$ in equation (3) measures the activity of the investor confidence channel. The estimated results of the benchmark in Model I show that $\omega =1.0287$ . When $\omega = 0$ , the investor confidence channel is completely closed and has no influence to economy. To investigate the effect of the confidence channel on the multipliers, we consider three intensity of the channel: no influence ( $\omega =0$ ), the baseline ( $\omega =1.0278$ ), and a greater influence ( $\omega =2$ ). Figure 7 shows the difference in fiscal multipliers resulting from changes in this coefficient.

Figure 7.

The investor confidence influence and fiscal multipliers.

The results suggest that the investor confidence channel amplifies the five types of multipliers, and therefore ignoring the role of investor confidence in the economy may underestimate the fiscal effectiveness. Unlike habit formation, the effect of investor confidence on multipliers is initially weak and increasingly significant in the long run. Fiscal shocks trigger endogenous investor confidence fluctuations in the same direction as output and interest rate spreads. This effect is not obvious in the initial period because of the lag in private capital accumulation. However, through the positive cycle of output, confidence, and investment rising together, investor confidence will continue to provide a multiplier expansionary effect in the medium to long term. In the meantime, the nominal interest rate will rise to reduce spreads and weaken confidence to avoid a herd behavior in the economy. From this point of view, the Taylor rule can also countercyclically regulate the economy through the investor confidence channel. As a result, a more active investor confidence channel contributes to the increased effectiveness of long-term fiscal policy, but the increase is limited due to the Taylor rule.

5.2.2. Investor confidence feedback mechanism

In Section 3.1, we provide empirical evidence of the investor confidence feedback mechanism in equation (4). We now explore the effect of parameters $\kappa _{y}$ and $\kappa _{r}$ on the multipliers through a counterfactual simulation. This helps us understand how endogenous investor confidence transmits its impact to the multipliers. We recall from Table 2 that the estimation of $\kappa _{y}$ and $\kappa _{r}$ for model I is 0.1880 and 0.0169. Then Figure 8 illustrates three scenarios regarding $\kappa _{y}$ : no response ( $\kappa _{y}=0$ ), estimated level ( $\kappa _{y}=0.1880$ ), and twice the estimated level ( $\kappa _{y}=2\times 0.1880$ ). Figure 9 presents a similar simulation for $\kappa _{r}$ .

Figure 8.

The investor confidence response to output and fiscal multipliers.

First, Figure 8 shows that the confidence-to-output response mechanism significantly expands the five types of multipliers. The positive channel through which confidence acts on the output is intuitive based on the implications of the output multipliers. At the same time, the empirical estimates suggest a larger $\kappa _{y}$ (compared to $\kappa _{r}$ ), so that the confidence-to-output response mechanism is particularly active and has a significant impact on the multiplier.

Figure 9.

The investor confidence response to spreads and fiscal multipliers.

Second, Figure 9 reveals that the confidence-to-spread response mechanism on the overall multiplier is almost negligible due to the small estimate of $\kappa _r$ . Within the five types of multipliers, capital tax is relatively more affected. Specifically, an increase in $\kappa _r$ causes the capital tax multiplier to contract, which is the opposite effect of an increased $\kappa _y$ . The reason is that capital taxes dampen households’ willingness to invest, leading to a higher marginal cost of capital for firms, higher return on capital, and wider interest rate spreads. The larger the $\kappa _r$ , the stronger the boost to investor confidence from widening spreads, which in turn mitigates the negative impact of capital taxes on private investment. Therefore, the confidence-to-spread response mechanism can weaken the negative effect of capital tax. As a whole, the investor confidence channel primarily increases the fiscal effect in the medium to long term by identifying output fluctuations.

5.3. The micro-foundation and liquidity trap on multipliers

From Section 4.2, we know that the multiplier effect expands in an economy that moves from no ZLB to a certain period ZLB at the estimated level of Model II. We define the multiplier difference between the two states as the multiplier movement; that is

(24) \begin{align} \mathrm {Multiplier}\,\,\mathrm {Movement=Multiplier_{ZLB}-Multiplier_{no \, ZLB}} \end{align}

We first compare the multiplier movements of 4 and 8 quarters ZLB constraint, with the habit formation as the estimated level ( $h=0.6189$ , in Model II) versus non-existent ( $h=0$ ), respectively, to analyze the impact of habit formation on fiscal effect under ZLB. Figure 10 shows the results.

Figure 10.

Multiplier movements in the ZLB based on habit formation perspective.

It can be seen that the liquidity trap leads to the positive multiplier movements of five types of fiscal policies. And the multiplier movements are notably larger as the ZLB period increases, as in Figure 1. The presence of habit formation significantly shrinks the multiplier movements during the ZLB period compared to an economy without habit formation. In particular, for both types of government spending, habit formation, while expanding the government spending multipliers (as shown in Figure 1), weakens the amplifying effect of the liquidity trap on government spending multipliers. Thus in systems with high intensity of habit formation, the economic stimulus effect of government spending policies on the onset of a liquidity trap may be overestimated. Moreover, after the end of the ZLB, as the Taylor rule comes back into effect, the nominal interest rate rises rapidly, restarting the crowding-out effect on private spending, and hence the multiplier movement declines. At this point, the difference in multiplier movements due to changes in habit formation also contracts. This indicates that the effect of habit formation on multipliers with ZLB is more pronounced in the early period, but weakens in the later period.

Similarly, we next compare the multiplier movements based on investor confidence perspective in Figure 11. We simulate two cases of confidence channel: the estimated level ( $\omega =1.0720$ , in Model II) and closing the channel ( $\omega =0$ ), for 4 and 8 quarters ZLB’s multiplier movements, respectively.

Figure 11.

Multiplier movements in the ZLB based on investor confidence perspective.

It is worth noting, on the one hand, that the overall impact of the investor confidence channel on multiplier movements is smaller compared to habit formation. Its impact is mainly realized in the medium to long term, rather than during the period when the ZLB occurs. On the other hand, the investor confidence channel amplifies the multiplier movements. Since the nominal interest rate cannot move, fiscal policies in the ZLB allow for greater volatility of interest rate spread. At the same time, the confidence-to-output mechanism is also able to amplify the multiplier movements caused by the interest rate transmission effect of the ZLB.

In summary, the two micro-foundations of private behavior preference have opposite effects on the multiplier under the liquidity trap. Habit formation, based on a mechanism that smoothes out fluctuations in micro agents’ consumption, contracts the multiplier movements under the liquidity trap. Investor confidence, by contrast, is based on the output and spread response mechanism, which slightly amplifies the multiplier movement. The effect of habit is considerable and more pronounced in the initial period, while the investor confidence continues to have an impact in the long run due to the nature of capital accumulation. From this perspective, government decisions based on a modern fiscal and monetary policy framework should focus on the role of micro-foundations in the policy transmission mechanism to enhance the effectiveness and stability of policy coordination.