Economics · Model
The Canonical RBC Model
The canonical Real Business Cycle (RBC) model describes a competitive economy with three building blocks. An infinitely lived representative household chooses how much to consume, work and save; a representative firm hires labour and rents capital to produce a single good, with a technology whose productivity evolves exogenously and stochastically; and perfectly competitive markets for the good, labour and capital clear in every period. Business cycles are the optimal reactions of households and firms to exogenous technology shocks: there is no other source of fluctuations. The model presented here is the basic neoclassical model of King and Rebelo (1999), with labour-augmenting trend growth. This page sets out the model, calibrates it, feeds it with the technology shocks measured in the US data, and compares the business cycles it generates with those observed.
The Model
Households
The household’s problem
An infinitely lived representative household values consumption, \(c_t\), and leisure, \(1-h_t\), where \(h_t\) is hours worked and the time endowment is normalised to one. It discounts the future at the rate \(\beta\) and forms rational expectations. It owns the capital stock, \(K_t\), which it rents to firms at the rate \(r_t\), supplies hours at the wage \(w_t\), and divides its income between consumption and investment, \(I_t\), which adds to the capital stock as capital depreciates at the rate \(\delta\):
Its choices satisfy two optimality conditions: labour supply, which equates the marginal utility cost of an extra hour of work to its benefit, and the Euler equation for saving:
Firms
Production technology and the firm’s problem
A representative firm rents capital and hires hours in competitive markets and produces with a Cobb–Douglas technology, in which \(\alpha\) is capital’s share of income:
Technology has two parts. Total factor productivity, \(A_t\), fluctuates around its long-run level, \(\bar{A}\), with persistence \(\rho\), and labour-augmenting technical progress, \(X_t\), grows at the constant rate \(g_X\) each quarter:
The innovations to total factor productivity, the technology shocks \(\varepsilon^A_t \sim \mathcal{N}(0, \sigma^2_{\varepsilon^A})\), are the only source of fluctuations, while \(X_t\) makes each hour of work steadily more productive.
Placing trend growth on labour is what gives the economy a balanced growth path (Uzawa, 1961; see Forms of technical progress). Along that path, output, consumption, investment, capital and the wage all grow at the rate \(g_X\), while hours, the interest rate and the shares of capital and labour in income stay constant, as in the long-run US data (Kaldor, 1961). Variables divided by \(X_t\) are stationary, so the model has a steady state, and the business cycle is measured as deviations from the balanced growth path.
Firms maximise profits by choosing capital and hours, given the market prices of the inputs, the wage, \(w_t\), and the rental rate of capital, \(r_t\), with output given by the production function:
The first-order conditions equate the wage and the rental rate to the marginal products of labour and capital:
Because the firm rents its capital, its problem is static: it chooses capital and hours period by period, without regard to the future, and the dynamic decisions, saving and investment, are made by households.
General Equilibrium
Definition of competitive equilibrium
Given an initial capital stock, \(K_0\), an initial level of technology, \(A_0\), and the stochastic process for technology, a competitive equilibrium consists of state-contingent sequences of household choices (consumption, labour supply and capital), firm choices (capital and labour demand) and prices (the wage and the rental rate of capital), in which every variable dated \(t\) depends only on the history of shocks up to \(t\), such that:
- Household optimisation. Given prices and the stochastic process for technology, households maximise lifetime expected utility subject to their sequence of budget constraints, starting from \(K_0\).
- Firm optimisation. Given prices and the current level of technology, firms maximise profits in every period.
- Market clearing. In every period and for every history of shocks, the demand for labour equals its supply, the demand for capital equals its supply, and the goods market clears: consumption plus investment equals output, \(c_t + I_t = Y_t\).
On this page, \(K_0\) and \(A_0\) are their values in the US data in 1947 Q1.
Methodology
Simulating the model takes three steps. The parameters are calibrated, so that the model’s steady state matches long-run features of the US economy. The model is then log-linearised around its steady state and solved, which gives the rules by which households and firms respond to technology (see “Solution” in Definitions, Methods and Solutions). Finally, the technology shocks measured in the US data are fed into the solved model, quarter by quarter. The settings below choose the technology series and the shocks used everywhere on the page.
These settings apply to every chart and table below. Linear trend: the shocks of King and Rebelo (1999), with the model’s series HP filtered afterwards. HP-filtered: the shortcut, with no filtering of the model’s series. One comparison at a time.
Calibration
To simulate the model, each of its parameters must be given a value. The calibration follows King and Rebelo (1999), updated to current empirical targets.
Technology shocks
The technology shocks that drive the model are measured in the US data, following King and Rebelo (1999): the logarithm of technology is split into a linear trend and a deviation from it, whose innovations are the shocks.
The logarithm of technology is split into a linear trend and a stochastic component, which follows a first-order autoregressive process; the technology shock is the innovation of that process, as in King and Rebelo (1999). Both are estimated over the full history of each series; see Extracting Technology Shocks for each step. Hicks-neutral technology multiplies the whole production function, \(Y = A\,F(K, h)\), and is total factor productivity itself, as in King and Rebelo. Harrod-neutral (labour-augmenting) technology multiplies labour, \(Y = F(K, Zh)\); with a Cobb–Douglas production function its growth is TFP growth divided by labour’s share of income, \(1-\alpha\), using Fernald’s capital share \(\alpha\) for each quarter, so its shocks are larger by about one half and equally persistent.
Model Simulations vs the Data
The model is fed with the technology shocks above, quarter by quarter, and its series are compared with the HP cyclical components of the US data shown on the Real Business Cycles page. The model has no other shocks: every movement in its series comes from technology.
Data: HP cyclical components (λ = 1600) from the Real Business Cycles page, where the sources are listed. Model: author’s calculations.
Business-Cycle Statistics: Model and Data
The statistics of King and Rebelo (1999, Tables 1 and 3), for the US economy and for the model over the same quarters.
Correlation with the data
How much of the US history the model reproduces: the volatility of each model series relative to the data, and the correlation between the two, quarter by quarter.
Impulse Responses to One Positive Technology Shock
Consider a positive technology shock, \(\varepsilon^A_1\), of 1 percent hitting an economy on its balanced growth path in quarter 1, with no further shocks.
Output rises on impact, both because technology is higher and because the higher wage draws more hours. Households smooth consumption: they save most of the extra income, so investment jumps, while consumption rises gradually, in a hump, as the capital stock builds up. Capital peaks after several years and then returns slowly to its path. Hours return to normal faster than technology, and fall slightly below it after a few years, when the wealth effect of a larger capital stock outweighs the substitution effect of a higher wage. The responses follow the technology series selected above; the type of shocks does not affect them.
Definitions, Methods and Solutions
Equilibrium conditions and balanced growth
Substituting the firm’s first-order conditions for the wage and the rental rate into the household’s optimality conditions, and using market clearing, the equilibrium is summarised by three conditions, one for every period: labour supply, the Euler equation and the resource constraint of the economy,
together with the process for technology and the initial conditions \(K_0\) and \(A_0\).
Balanced growth. Because trend growth is labour-augmenting and utility is logarithmic in consumption, the economy grows along a balanced growth path (King, Plosser and Rebelo, 1988). Dividing each growing variable by \(X_t\), as in \(\tilde{c}_t = c_t/X_t\), \(\tilde{k}_t = K_t/X_t\) and \(\tilde{y}_t = Y_t/X_t\), gives a stationary economy. Its conditions are those of the economy without growth, except that growth appears in the accumulation of capital and in the discount factor:
In the steady state of the stationary economy, the marginal product of capital is \(\alpha\, \tilde{y}/\tilde{k} = (1+g_X)/\beta - 1 + \delta\): faster growth raises the interest rate and lowers the capital–output ratio. The model’s series are percent deviations from the balanced growth path, so that, in logarithms, \(\ln Y_t = \ln \tilde{y} + \ln X_t + \hat{y}_t\): a constant, a linear trend and the cycle. The HP filter removes a linear trend exactly, so trend growth affects the model’s cycles only through the steady state and the dynamics, not through the trend itself.
Technology. With Hicks-neutral technology, \(A_t\) is total factor productivity, as in King and Rebelo. With Harrod-neutral technology, \(Y_t = K_t^{\alpha} (Z_t X_t h_t)^{1-\alpha}\), so that \(A_t = Z_t^{1-\alpha}\). In both cases the trend of technology estimated in the data fixes \(g_X\): the trend growth of TFP is \(1-\alpha\) times the growth of \(X_t\).
Solving and simulating the model
Log-linearisation. Each condition of the stationary economy is approximated around its steady state in percent deviations, written with a hat (\(\hat{y}_t = 100 \ln (\tilde{y}_t/\tilde{y})\), the percent deviation of output from its balanced growth path). With \(\Upsilon = \alpha + h/(1-h)\) and \(\Omega = 1 - \beta(1-\delta)/(1+g_X)\):
Eliminating hours with the labour-supply condition leaves three equations in technology and capital, which are predetermined, and consumption, which is free to jump.
Solution. The system is solved by the eigenvalue decomposition of Blanchard and Kahn (1980). It has a unique stable solution when the number of eigenvalues outside the unit circle equals the number of jump variables, here one; the Solution table reports the eigenvalues. Consumption then follows the saddle path
and every other variable follows from technology, capital and consumption: hours from the labour-supply condition, output from the production function, investment from the resource constraint, labour productivity and the real wage from \(\hat{y}_t - \hat{h}_t\) (the two are equal with a Cobb–Douglas production function), and the real interest rate from the Euler equation, \(\hat{r}_t = 4\,(E_t \hat{c}_{t+1} - \hat{c}_t)\), in percentage points a year.
Simulation. The model starts in the first quarter of the technology data (1947 Q1), with technology and capital at their values in the data, and is then driven by the measured shocks, quarter by quarter. Both starting values are measured in the same way as the shocks that follow them: with the linear-trend shocks, technology and capital per capita start at their deviations from their linear trends; with the HP-filtered shocks, at their HP cyclical components, the cycle of capital being the one on the Real Business Cycles page. Capital is Fernald’s capital input, which measures the services of the capital stock. The Solution table reports the starting values; the first three years are not shown.
Two ways of comparing the model with the data. The model’s series are deviations from the balanced growth path, while the data’s are HP cyclical components, from which slow movements around the trend have also been removed. King and Rebelo (1999, section 4.3.1) therefore HP filter the model’s series too (λ = 1600) before comparing them with the data: this is the “Linear trend” setting. The “HP-filtered” setting uses the shortcut described on the Extracting Technology Shocks page: the model is driven by the HP-filtered shocks, and its series are already HP cyclical components. Because the model is linear, the two coincide except near the ends of the sample; the comparison setting shows both.
The data. As in King and Rebelo, the model is compared with per-capita data: output, consumption and investment per person, hours per person aged 16 and over (nonfarm business sector), and Fernald’s capital input of the business sector per person. The real wage is real hourly compensation and labour productivity is output per hour, both in the nonfarm business sector. The real interest rate is the ex-post real three-month Treasury-bill rate; King and Rebelo use an expected real rate built from inflation forecasts (Stock and Watson, 1999).
Solution
Technology shocks, step by step
Following King and Rebelo (1999, section 4.1), the logarithm of total factor productivity is split into a linear trend and a deviation from that trend. The deviation is regressed on its value one quarter earlier, which gives its persistence, \(\rho\); the technology shock is the part of the deviation that last quarter does not predict, and its standard deviation, \(\sigma\), measures the size of the shocks. The model’s technology then follows
where \(\hat{A}_t\) is the percent deviation of technology from its trend and \(\varepsilon^A_t\) the technology shock (\(\varepsilon^{A^u}_t\) for utilization-adjusted TFP).
Forms of technical progress: Hicks, Harrod and Solow
Technical progress can enter the production function \(Y = F(K, h)\) in different places. The classic definitions say which ratio it leaves unchanged; they are called “neutral” because they do not change the distribution of income between capital and labour at a given value of that ratio.
- Hicks-neutral (Hicks, 1932): \(Y = A\,F(K, h)\). Technology multiplies the whole production function, raising the marginal products of capital and labour in the same proportion, so the ratio of marginal products is unchanged at a given capital–labour ratio. \(A\) is total factor productivity; King and Rebelo (1999) use it for the technology shocks.
- Harrod-neutral or labour-augmenting (Harrod, 1948): \(Y = F(K, Z h)\). Technology makes each hour of work more productive, as if there were more workers; it leaves the capital–output ratio unchanged at a given interest rate. It is the only form consistent with balanced growth for a general production function (Uzawa, 1961), which is why growth models, and the trend in King and Rebelo’s model, use it.
- Solow-neutral or capital-augmenting: \(Y = F(B K, h)\). Technology makes each machine more productive, as if there were more capital; it leaves the labour–output ratio unchanged at a given wage.
Which form of technical progress should a model use? The long-run evidence favours Harrod neutrality. Kaldor (1961) summarised the facts of growth: output per worker grows at a roughly constant rate, while the capital–output ratio, the real return on capital and the shares of capital and labour in income stay roughly constant. These “stylised facts” describe balanced growth, and Uzawa (1961) showed that, for a general production function, balanced growth is possible only if technical progress is labour-augmenting. Harrod neutrality is therefore the natural modelling assumption, and the one used for trend growth in most growth and business-cycle models, including King and Rebelo’s. Kaldor (1957) himself went further and rejected the separation between moving along a production function and shifting it, replacing it with a “technical progress function” linking the growth of output per worker to the growth of capital per worker.
With a Cobb–Douglas production function, however, the choice does not matter for the data: the three forms are equivalent. If \(Y = A K^{\alpha} h^{1-\alpha}\), with \(\alpha\) the capital share, then
The data cannot tell them apart, but the size of the shocks depends on where technology is placed: in logarithms, \(\Delta \ln Z = \Delta \ln A / (1-\alpha)\), so labour-augmenting shocks are about one and a half times as large as TFP shocks, with the same persistence. The “Technical progress” setting above the technology chart shows the Hicks- and Harrod-neutral versions; the one to use is the one that matches where technology enters the model’s production function.
References
- Basu, S., J. G. Fernald and M. S. Kimball (2006). “Are Technology Improvements Contractionary?” American Economic Review 96(5): 1418–1448.
- Blanchard, O. J. and C. M. Kahn (1980). “The Solution of Linear Difference Models under Rational Expectations.” Econometrica 48(5): 1305–1311.
- Fernald, J. G. (2014). “A Quarterly, Utilization-Adjusted Series on Total Factor Productivity.” Federal Reserve Bank of San Francisco Working Paper 2012-19 (updated March 2014).
- Harrod, R. F. (1948). Towards a Dynamic Economics. London: Macmillan.
- Hicks, J. R. (1932). The Theory of Wages. London: Macmillan.
- Kaldor, N. (1957). “A Model of Economic Growth.” Economic Journal 67(268): 591–624.
- Kaldor, N. (1961). “Capital Accumulation and Economic Growth.” In F. A. Lutz and D. C. Hague (eds.), The Theory of Capital: 177–222. London: Macmillan.
- King, R. G., C. I. Plosser and S. T. Rebelo (1988). “Production, Growth and Business Cycles: I. The Basic Neoclassical Model.” Journal of Monetary Economics 21(2–3): 195–232.
- King, R. G. and S. T. Rebelo (1999). “Resuscitating Real Business Cycles.” In J. B. Taylor and M. Woodford (eds.), Handbook of Macroeconomics, Vol. 1B, Chapter 14: 927–1007. Amsterdam: Elsevier.
- Kydland, F. E. and E. C. Prescott (1982). “Time to Build and Aggregate Fluctuations.” Econometrica 50(6): 1345–1370.
- Long, J. B. and C. I. Plosser (1983). “Real Business Cycles.” Journal of Political Economy 91(1): 39–69.
- Prescott, E. C. (1986). “Theory Ahead of Business Cycle Measurement.” Federal Reserve Bank of Minneapolis Quarterly Review 10(4): 9–22.
- Solow, R. M. (1957). “Technical Change and the Aggregate Production Function.” Review of Economics and Statistics 39(3): 312–320.
- Stock, J. H. and M. W. Watson (1999). “Business Cycle Fluctuations in US Macroeconomic Time Series.” In J. B. Taylor and M. Woodford (eds.), Handbook of Macroeconomics, Vol. 1A, Chapter 1: 3–64. Amsterdam: Elsevier.
- Uzawa, H. (1961). “Neutral Inventions and the Stability of Growth Equilibrium.” Review of Economic Studies 28(2): 117–124.
Page created on 06/11/2026.
© Arthur Galichère.
- Errors: Every effort has been made to ensure the data and calculations are accurate; any remaining errors are mine. If you believe you have found a mistake, or have a comment or suggestion, please let me know.
- Reuse and citation: Total factor productivity, capital input and capital’s share of income are from John Fernald at the Federal Reserve Bank of San Francisco; the other data are from the U.S. Bureau of Economic Analysis, the U.S. Bureau of Labor Statistics and the Federal Reserve, via FRED. The model, estimates, charts and text are the author’s; the source data remain subject to the terms of their providers. When reusing these figures, please cite the original data providers and this website: Arthur Galichère, The Canonical RBC Model.