Why the Source of Output Growth Matters
An increase in production does not, by itself, show that an economy or firm has become more productive. Output may rise because more labor or capital is being used, because the same inputs produce more, or because both have occurred. Separating these sources is useful when interpreting economic growth, comparing production periods, or assessing a hypothetical insurer’s claims-processing operation.
The key distinction is between a movement along a production function, caused by changing input quantities while the production relationship is held fixed, and a change in productivity, represented by more output from a given bundle of inputs. The distinction is conceptually clear, although measured productivity can reflect many influences and is not automatically a measure of improved technology alone.
A convenient model writes output as \(Y=A F(K,L)\), where \(F(K,L)\) is the production function and \(A\) is a productivity factor. In this form, changing \(K\) or \(L\) while \(A\) and \(F\) remain fixed is input growth. Raising \(A\) while keeping \(K\) and \(L\) fixed is productivity growth. This is a model, not a claim that every real-world change can be cleanly observed or assigned to one cause.
A simple comparison uses the same production function in two periods. First calculate the output the new inputs would produce if productivity had not changed. The difference from initial output is the input contribution under that counterfactual. Then compare actual final output with that counterfactual; the remaining difference is the productivity contribution under this ordering.
Worked Example: Separating Input Growth From a Productivity Increase
A hypothetical workshop produces components according to \(Y=A(2K+3L)\), where \(K\) is machine-hours, \(L\) is labor-hours, and \(Y\) is components per week. Initially, \(K=20\), \(L=10\), and \(A=1\). In the next period, \(K=24\), \(L=12\), and \(A=1.1\). Decompose the output increase by first changing the inputs and then changing productivity.
Method and assumptions. Use the stated linear production function, with the coefficients held fixed between periods. Treat \(A\) as a dimensionless productivity multiplier. The specified order is inputs first, productivity second.
Initial output. The input-based output is \(2(20)+3(10)=40+30=70\). With \(A=1\), initial output is \(Y_0=1(70)=70\) components per week.
Output with new inputs but initial productivity. The new input-based output is \(2(24)+3(12)=48+36=84\). Holding \(A=1\), this counterfactual output is \(84\). The input contribution is \(84-70=14\) components.
Final output with new productivity. Apply \(A=1.1\) to the new input-based output: \(Y_1=1.1(84)=92.4\). The productivity contribution after the input change is \(92.4-84=8.4\) components.
Check. Total output growth is \(92.4-70=22.4\). The two contributions sum to \(14+8.4=22.4\), as required. The common mistake is to call the entire increase of \(22.4\) “productivity growth”; \(14\) of it is accounted for by using more inputs under the stated model and ordering.
Labor Productivity Is Not the Same as Productivity Growth
Labor productivity is output per unit of labor input. Depending on the data, labor input may be measured as workers, worker-hours, or another defined labor quantity. For hourly data, the measure is
This ratio can rise even when the productivity factor \(A\) does not change. For example, workers may have more capital equipment available, which can increase output per worker-hour. Changes in the mix or quality of labor inputs can also affect measured output per hour. Therefore, labor productivity is a useful measure, but it is not automatically a direct measure of a shift in the production function.
Worked Example: Decomposing Output Growth Using Hours and Output per Hour
A hypothetical document-processing team completes 600 files in a week using 120 labor-hours. The following week it completes 780 files using 130 labor-hours. Treat files as comparable units of output and assume the hours and output refer to the same production periods. Find the change in hourly labor productivity and decompose the output increase by adding hours first.
Method. Write output as labor-hours multiplied by output per labor-hour. This identity allows a sequential decomposition; it does not establish that the productivity change was caused by technology.
Initial productivity. The initial rate is \(600/120=5\) files per labor-hour.
Final productivity. The final rate is \(780/130=6\) files per labor-hour. The rate has increased by \(6-5=1\) file per labor-hour, or \(1/5=0.20=20\%\) relative to its initial level.
Output if hours increased but productivity stayed at 5. At 130 hours and 5 files per hour, output would be \(130(5)=650\) files. The added-hours contribution is \(650-600=50\) files.
Productivity contribution. Actual final output is 780 files, so the additional contribution after the hours increase is \(780-650=130\) files. The total increase is \(50+130=180\) files.
Check using growth rates. Hours increased by \(10/120=8.3333\%\), and output per hour increased by \(20\%\). Their multiplicative combination is \((130/120)(6/5)=1.083333\ldots(1.2)=1.30\), a 30% output increase. Directly, \(780/600=1.30\). The common mistake is to add 8.3333% and 20% and report 28.3333%; the factors multiply, so the interaction also affects total growth.
A Cobb-Douglas Model and the Role of Capital per Worker
A Cobb-Douglas production function is a common way to represent output produced using capital and labor. In \(Y=A K^\alpha L^\beta\), the exponents describe how output responds to proportional changes in each input, subject to the model’s assumptions. With constant returns to scale, \(\alpha+\beta=1\): multiplying both inputs by the same factor multiplies output by that factor when \(A\) is unchanged.
Dividing output by labor gives a direct explanation of why output per worker can rise without a productivity shift. Under constant returns to scale, the equation reduces to a function of capital per worker and \(A\).
The last equality follows because \(\beta-1=-\alpha\), so \(K^\alpha L^{\beta-1}=K^\alpha L^{-\alpha}=(K/L)^\alpha\). Thus, with \(A\) fixed, a rise in capital per worker can raise labor productivity. A rise in \(A\), by contrast, raises output per worker for a given capital-labor ratio in this model.
Worked Example: Capital Deepening Versus Productivity Growth
A hypothetical producer has a constant-returns Cobb-Douglas production function \(Y=A K^{0.4}L^{0.6}\). Initially \(A=1\), \(K=100\), and \(L=100\). Later, \(K=200\) while \(L=100\); in a further comparison, productivity rises to \(A=1.1\) with those later input quantities unchanged. Calculate the output and output per worker at each stage.
Assumptions and method. Use the stated production function throughout, with exponents \(0.4\) and \(0.6\) fixed. The first change is capital deepening: capital rises while labor stays fixed. The second change raises \(A\) while both inputs stay fixed. Round output and output per worker to four decimal places only at the end.
Initial output. Since \(100^{0.4}100^{0.6}=100^{1}=100\), initial output is \(Y_0=1(100)=100\). Initial output per worker is \(100/100=1\).
Output after capital deepening with \(A=1\). The capital-labor ratio changes from \(100/100=1\) to \(200/100=2\). The output-per-worker formula gives \(Y/L=1(2)^{0.4}\). Since \(2^{0.4}=1.3195079\ldots\), output per worker is \(1.3195\), and total output is \(100(1.3195079\ldots)=131.9508\).
Output after the productivity increase. At the same \(K=200\) and \(L=100\), raise \(A\) to 1.1. Output per worker becomes \(1.1(2^{0.4})=1.4514587\ldots\), or \(1.4515\). Total output is \(100(1.4514587\ldots)=145.1459\).
Check and interpretation. Capital deepening raises output from \(100\) to \(131.9508\), an increase of \(31.9508\), without any change in \(A\). The separate productivity increase adds \(145.1459-131.9508=13.1951\). Together, \(31.9508+13.1951=45.1459\), equal to \(145.1459-100\). The common mistake is to label the entire rise in output per worker as productivity growth; here, the first \(31.9508\) of output growth comes from a higher capital-labor ratio in the model.
Growth Accounting and Careful Interpretation
For the Cobb-Douglas model \(Y=A K^\alpha L^\beta\), take logarithms to express proportional growth as a sum. The exact log-change identity is
This follows by taking logarithms of the production function: \(\ln Y=\ln A+\alpha\ln K+\beta\ln L\). Subtracting the initial-period equation from the final-period equation gives the identity. For small percentage changes, log changes are approximately percentage changes, which motivates the familiar growth-accounting approximation. The identity depends on the Cobb-Douglas specification and measured inputs; it is not a model-free decomposition.
The part not explained by measured input growth is often called a productivity residual. A residual is a remainder relative to the model and data, not proof of a particular cause. It may reflect technological improvement, but it can also capture omitted inputs, changing input quality, utilization, measurement error, or a production relationship that is not well represented by the assumed function.
Use comparable output units and define the relevant labor or capital input measures before calculating growth.
Specify which inputs and productivity terms are held fixed in the comparison. Do not infer a cause from an output ratio alone.
Apply the new input quantities to the original production relationship to isolate an input contribution under the stated ordering.
Compare actual output with the counterfactual, and describe the result as a productivity contribution within the model—not necessarily as pure technological progress.
The sequential decomposition depends on the order used when inputs and productivity both change. If productivity is changed first, the input contribution may be calculated using the new productivity level and therefore differ from the input contribution calculated using the old level. The total change remains the same. State the chosen order or use a specified growth-accounting convention rather than presenting one allocation as uniquely determined.
Check Your Understanding
Work each problem using the stated production relationship or productivity definition. Treat every scenario as hypothetical.
- A workshop’s output is \(Y=A(4K+2L)\). Initially \(A=1\), \(K=10\), and \(L=15\). Later, \(K=12\), \(L=18\), and \(A=1.05\). Decompose the output increase by changing inputs first and productivity second. (Answer: Initial output \(70\); input-only output \(84\); final output \(88.2\); input contribution \(14\), productivity contribution \(4.2\), total increase \(18.2\).)
- A hypothetical service team completes 960 cases in 160 labor-hours, then completes 1,200 cases in 180 labor-hours. Find productivity in each period and decompose output growth by increasing hours first. (Answer: Productivity rises from \(6\) to \(6.6667\) cases per hour; input contribution \(120\) cases, productivity contribution \(120\) cases, total increase \(240\) cases.)
- A firm uses \(Y=A K^{0.3}L^{0.7}\). Initially \(A=1\), \(K=100\), \(L=100\); later \(A\) is unchanged, \(K=400\), and \(L=100\). Find the output-per-worker multiplier and explain its source. (Answer: The capital-labor ratio rises from \(1\) to \(4\), so output per worker is multiplied by \(4^{0.3}\approx1.5157\), a \(51.57\%\) increase from capital deepening, with no change in \(A\).)
- Under \(Y=A K^{0.25}L^{0.75}\), capital and labor each rise by \(10\%\), while \(A\) rises by \(4\%\). Find the exact output multiplier and the separate sequential contributions as percentages of initial output, with inputs changed first. (Answer: Input-only multiplier \(1.10\), so input contribution is \(10\%\); final multiplier \(1.04(1.10)=1.144\), so productivity contribution is \(4.4\%\) and total output growth is \(14.4\%\).)
- Suppose measured output rises by \(12\%\) while measured labor-hours are unchanged. Can one conclude that technology improved by exactly \(12\%\)? State the conclusion that follows from the labor-productivity calculation and one limitation. (Answer: Output per labor-hour rises by \(12\%\); this alone does not prove technology improved by exactly \(12\%\), because omitted inputs, input quality, utilization, measurement error, or model choice may affect the result.)