4.3 Second-order linear differential equations

Definition and Characteristics

A second-order linear differential equation describes how a function relates to its first and second derivatives. In economics, these equations let you model systems where the rate of change itself is changing, which is exactly what happens in business cycles, growth dynamics, and market adjustments.

General Form

The standard form is:

$a(x)y'' + b(x)y' + c(x)y = f(x)$

• $y''$ is the second derivative of $y$ with respect to $x$ (often time)
• $a(x)$, $b(x)$, and $c(x)$ are coefficient functions that shape the system's behavior
• $f(x)$ on the right-hand side represents external forces or inputs (like a policy shock or an exogenous demand shift)

In most intro-level economic models, the coefficients are constants rather than functions of $x$, which simplifies things considerably.

Order and Linearity

Order refers to the highest derivative present. Since $y''$ is the highest here, these are second-order.

Linearity means the equation satisfies the superposition principle: if $y_1$ and $y_2$ are both solutions, then any linear combination $c_1 y_1 + c_2 y_2$ is also a solution. This is a huge practical advantage because it lets you build complex solutions from simpler pieces.

Homogeneous vs. Non-Homogeneous

• Homogeneous: $f(x) = 0$. The system has no external input. Solutions describe the system's natural behavior (free oscillations, decay, or growth on its own).
• Non-homogeneous: $f(x) \neq 0$. Something external is driving the system, like government spending, a technology shock, or a periodic demand fluctuation.

This distinction matters because the two types require different solution strategies, and the economic interpretation changes depending on which one you're dealing with.

Solutions and Methods

Solving these equations means finding a function $y(x)$ that satisfies the equation for given initial conditions. The approach differs for homogeneous and non-homogeneous cases, but they connect in a clean way.

Characteristic Equation

For a homogeneous equation with constant coefficients $ay'' + by' + cy = 0$, you find solutions by guessing $y = e^{rx}$ and substituting in. This gives you the characteristic equation:

$ar^2 + br + c = 0$

This is just a quadratic in $r$. The roots tell you everything about the system's natural behavior: whether it grows, decays, oscillates, or some combination.

Complementary Function

The complementary function $y_c$ is the general solution to the homogeneous equation. It takes the form:

$y_c = c_1 y_1 + c_2 y_2$

where $y_1$ and $y_2$ are two linearly independent solutions (found from the characteristic equation roots). This represents the system's natural response with no external forcing.

Particular Solution

The particular solution $y_p$ is any single solution that satisfies the full non-homogeneous equation. It captures how the system responds to the external input $f(x)$. Methods for finding it include undetermined coefficients and variation of parameters (covered below).

General Solution

The complete solution to a non-homogeneous equation combines both pieces:

$y = y_c + y_p$

The constants $c_1$ and $c_2$ in $y_c$ get pinned down by initial conditions (e.g., the initial value of output and its initial growth rate). Together, $y_c + y_p$ gives you the full picture of how the economic variable evolves.

Types of Solutions

The roots of the characteristic equation $ar^2 + br + c = 0$ fall into three cases, each producing a different solution shape with a distinct economic interpretation.

Real Distinct Roots

When the discriminant $b^2 - 4ac > 0$, you get two different real roots $r_1$ and $r_2$:

$y = c_1 e^{r_1 x} + c_2 e^{r_2 x}$

Each term grows or decays exponentially. If both roots are negative, the system decays to equilibrium. If one is positive, the system is unstable. This pattern shows up in models of capital accumulation or market penetration where variables move monotonically toward or away from a steady state.

Complex Conjugate Roots

When $b^2 - 4ac < 0$, the roots are complex: $r = a \pm bi$. The solution becomes:

$y = e^{ax}(c_1 \cos(bx) + c_2 \sin(bx))$

The $\cos$ and $\sin$ terms produce oscillations, while $e^{ax}$ controls whether those oscillations grow ($a > 0$), shrink ($a < 0$), or stay constant ($a = 0$). This is the mathematical backbone of business cycle models, where economic activity fluctuates around a trend.

Repeated Roots

When $b^2 - 4ac = 0$, there's one repeated root $r$:

$y = (c_1 + c_2 x) e^{rx}$

This is the boundary case between oscillation and pure exponential behavior. In economics, it represents a critically damped system: the variable returns to equilibrium as fast as possible without overshooting. Think of a market that adjusts to a new price level smoothly, without bouncing around.

Solution Techniques

Method of Undetermined Coefficients

Use this when $f(x)$ takes a "nice" form: polynomial, exponential, sine/cosine, or combinations of these.

1. Look at the form of $f(x)$ and guess a solution $y_p$ with the same form but unknown coefficients.
2. Substitute $y_p$ into the differential equation.
3. Match coefficients on both sides to solve for the unknowns.

For example, if $f(x) = 5e^{3x}$, you'd guess $y_p = Ae^{3x}$ and solve for $A$. This method is efficient for the kinds of forcing functions that come up most often in economic models (constant policy inputs, exponential growth trends).

Variation of Parameters

This is a more general method that works for any $f(x)$, not just the nice forms above. The idea: take the complementary function $y_c = c_1 y_1 + c_2 y_2$ and replace the constants with unknown functions $u_1(x)$ and $u_2(x)$. Then solve for those functions using a system of equations involving the Wronskian. It's more work, but it handles cases where undetermined coefficients can't.

Reduction of Order

If you already know one solution $y_1$ to the homogeneous equation, you can find a second independent solution by substituting $y = v(x) \cdot y_1$ into the equation. This reduces the problem to a first-order equation in $v'$. It's useful when partial information about the system is available from economic reasoning or prior analysis.

Applications in Economics

Growth Models

Second-order equations appear in growth theory when models account for acceleration effects. The Solow-Swan model, for instance, can involve second-order dynamics when extended to include adjustment costs on capital. Endogenous growth models use these equations to capture how knowledge accumulation and innovation feed back into the growth process.

Business Cycle Analysis

Fluctuations in output, employment, and inflation over time are naturally modeled by oscillatory solutions (the complex roots case). Real Business Cycle (RBC) models use second-order equations to study how productivity shocks propagate through the economy, producing the boom-bust patterns you see in GDP data.

Market Equilibrium Dynamics

When a market is disturbed from equilibrium, the adjustment path can involve overshooting (the price goes past equilibrium before coming back) or smooth convergence. Second-order equations capture both possibilities. Exchange rate models, for example, use these equations to explain why currencies often overshoot their long-run values after a monetary policy change.

Stability Analysis

Stability analysis tells you whether an economic system will return to equilibrium after a shock or diverge away from it. This is often the most policy-relevant question you can ask about a model.

Equilibrium Points

An equilibrium (or steady state) is where all derivatives equal zero, so the system stays put. Set $y' = 0$ and $y'' = 0$ in your equation and solve. An equilibrium is stable if nearby trajectories are pulled toward it, and unstable if they're pushed away.

Phase Diagrams

Phase diagrams plot $y$ against $y'$ (the variable against its rate of change). Without solving the equation explicitly, you can see the qualitative behavior: spirals indicate oscillations, straight-line trajectories indicate exponential behavior, and the direction of arrows shows whether the system converges or diverges.

Stability Conditions

For a constant-coefficient equation $ay'' + by' + cy = 0$, stability requires that the real parts of both characteristic roots are negative. In terms of the coefficients: you need $b/a > 0$ and $c/a > 0$ (assuming $a > 0$). The Routh-Hurwitz criteria generalize these conditions to higher-order systems.

Numerical Methods

When an equation can't be solved analytically (which happens often with realistic economic models), numerical methods approximate the solution computationally.

Euler's Method

The simplest approach. At each time step, you approximate the next value using the current derivative:

$y_{n+1} = y_n + h \cdot y'_n$

where $h$ is the step size. It's fast and easy to implement but can accumulate significant error over many steps, so it's best for quick rough estimates.

Runge-Kutta Methods

The fourth-order Runge-Kutta (RK4) method is the workhorse of numerical economics. It evaluates the derivative at multiple points within each step to get a much more accurate approximation than Euler's method, without being too computationally expensive. Most economic simulation software uses some variant of this.

Finite Difference Schemes

These methods approximate derivatives by differences between nearby grid points (forward, backward, or central differences). They're especially useful for boundary value problems, where you know conditions at two different points (say, initial capital stock and a long-run steady state) rather than just initial conditions.

Systems of Differential Equations

Real economies involve many interacting variables. A single second-order equation can actually be rewritten as a system of two first-order equations, and multi-variable economic models naturally produce coupled systems.

Coupled Equations

In coupled systems, each variable's rate of change depends on the other variables. For example, consumption growth might depend on the capital stock, while capital accumulation depends on consumption. These feedback loops are what make economic dynamics interesting and complex.

Matrix Representation

A linear system can be written compactly as:

$\mathbf{y'} = A\mathbf{y} + \mathbf{f}(x)$

where $A$ is a matrix of coefficients and $\mathbf{y}$ is a vector of variables. This compact form lets you apply linear algebra tools directly to economic problems, including input-output analysis and multi-sector models.

Eigenvalue Analysis

The eigenvalues of matrix $A$ play the same role as the roots of the characteristic equation for a single equation. They tell you about growth rates, oscillation frequencies, and stability. Eigenvectors indicate the directions along which the system evolves most naturally. If all eigenvalues have negative real parts, the system is stable.

Economic Interpretations

Dynamic Multipliers

Dynamic multipliers measure how a one-time shock (like a fiscal stimulus) affects an economic variable over time. They come from the particular solution and show the cumulative, time-varying response. A multiplier that grows and then fades tells you the stimulus has a temporary effect; one that settles at a new level indicates a permanent shift.

Adjustment Processes

The complementary function captures how quickly and in what pattern an economy moves toward equilibrium after a disturbance. A fast-decaying complementary function means rapid adjustment; an oscillatory one means the economy bounces around before settling. This matters for understanding policy response lags.

Long-Run Equilibrium

As $x \to \infty$, the complementary function typically dies out (in a stable system), leaving only the particular solution. This particular solution is the long-run equilibrium. In growth models, it corresponds to the balanced growth path; in market models, it's the steady-state price or quantity.

Limitations and Extensions

Non-Linear Systems

Most real economic relationships are non-linear. Linear second-order equations are useful approximations near equilibrium, but they can't capture phenomena like market crashes, regime shifts, or tipping points. Non-linear dynamics require tools from bifurcation theory and chaos theory.

Higher-Order Equations

Some economic models produce third-order or higher equations, especially when multiple feedback mechanisms interact. These require the same basic approach (characteristic equations, stability analysis) but with more complex algebra and additional solution components.

Stochastic Differential Equations

Real economies are full of randomness. Stochastic differential equations add noise terms to capture uncertainty in things like asset prices, demand shocks, or policy changes. They're central to financial economics (option pricing, risk management) and provide a more realistic picture of economic variability than purely deterministic models.