Introduction
A slowdown in output growth is not, by itself, evidence that productivity growth has weakened. Output can grow more slowly because capital services or labor input are growing more slowly, because measured productivity growth has declined, or because the underlying data do not capture output and inputs consistently. Growth accounting provides a structured way to separate these possibilities.
This workshop uses a production-accounting framework to interpret short datasets. The goal is not to identify a single cause from a residual. It is to make a disciplined diagnosis: calculate the input contributions, find the part of output growth they do not account for, and then ask what the data can and cannot establish. This approach is aligned with the SOA VEE Economics (Validation by Educational Experience) syllabus.
The Growth-Accounting Framework
Use the production relationship \(Y=A K^\alpha L^{1-\alpha}\), where \(Y\) is real output, \(K\) is the flow of capital services, \(L\) is labor input, \(A\) is measured productivity, and \(\alpha\) is the output elasticity of capital. The model assumes that the production relationship and its weights are stable over the period being compared. Under constant returns to scale, the labor elasticity is \(1-\alpha\). When factor markets are competitive, factor income shares can be used as approximate estimates of these elasticities; that interpretation requires the additional assumption that factor payments reflect marginal products.
Taking logarithms and comparing two dates gives an exact accounting identity in log changes. If growth rates are instead reported as ordinary percentage changes, the familiar weighted-sum formula is an approximation, especially for larger changes. Throughout the examples, the supplied annual growth rates are small and are used as percentage-point growth-accounting inputs.
The equation says that output growth is divided into measured productivity growth and the weighted growth of measured inputs. Rearranging gives the residual:
Labor input should reflect both hours and labor composition when the data allow it. If \(H\) is hours worked and \(Q_L\) is an index of labor composition, define \(L=H Q_L\). Then \(\Delta\ln L=\Delta\ln H+\Delta\ln Q_L\). More hours raise labor input; a shift toward workers with greater measured skills or experience can also raise labor input even if total hours do not change.
Similarly, \(K\) should represent productive capital services rather than merely the number or book value of assets. Changes in asset quality, capacity, or utilization can make a simple capital-stock measure a poor proxy for the services actually used in production.
Why the Residual Has This Form
The residual formula follows directly from the production relationship, rather than being an independent estimate of a particular cause. Taking natural logarithms of \(Y=A K^\alpha L^{1-\alpha}\) gives \(\ln Y=\ln A+\alpha\ln K+(1-\alpha)\ln L\). Subtract the expression for the earlier date from the expression for the later date. Each term becomes its change over the period, so the output change equals the productivity change plus the two weighted input changes. Rearranging isolates \(\Delta\ln A\).
If growth is reported in ordinary percentage terms rather than log changes, analysts commonly approximate log changes by percentage changes. This makes the weighted decomposition convenient but not exact for large changes. A residual is also sensitive to errors in any of its inputs: an output price index that is too high, for example, can distort measured real output growth and therefore the residual.
Worked Diagnoses
Worked Example: Output Slows While Productivity Growth Holds Steady
A hypothetical economy reports the following annual growth rates for two periods. Treat all rates as ordinary percentage changes and use the growth-accounting approximation. Assume a stable production relationship with \(\alpha=0.35\), and that the capital and labor input indices measure the relevant services accurately.
| Period | Real output growth | Capital-services growth | Labor-input growth |
|---|---|---|---|
| Earlier period | 3.0% | 4.0% | 2.0% |
| Later period | 2.0% | 3.0% | 1.0% |
Method. Calculate the capital and labor contributions in each period, then subtract their sum from output growth. The common mistake to avoid is to label any decline in output growth a productivity slowdown before checking input growth.
In the earlier period, the capital contribution is \(0.35\times4.0\%=1.40\) percentage points. The labor weight is \(1-0.35=0.65\), so the labor contribution is \(0.65\times2.0\%=1.30\) percentage points. Total input contribution is \(1.40+1.30=2.70\) percentage points. The residual is \(3.00-2.70=0.30\%\).
In the later period, the capital contribution is \(0.35\times3.0\%=1.05\) percentage points. The labor contribution is \(0.65\times1.0\%=0.65\) percentage points. Total input contribution is \(1.05+0.65=1.70\) percentage points. The residual is \(2.00-1.70=0.30\%\).
Check. Earlier-period contributions plus residual equal \(2.70+0.30=3.00\%\); later-period contributions plus residual equal \(1.70+0.30=2.00\%\). Output growth fell by 1.00 percentage point, and the measured input contribution also fell by 1.00 point. On these data, measured productivity growth did not slow.
Worked Example: Keep Precision When Calculating the Residual
A hypothetical industry reports annual real output growth of 9.531020%. Assume \(\alpha=0.40\), capital-services growth of 5.314875%, and labor-input growth of 6.250000%. Use the growth-accounting approximation and report the residual to three decimal places.
Method. Weight each input growth rate, retain the unrounded contributions through the subtraction, and round only the final residual. The trap is to claim that early rounding necessarily changes the final reported answer.
The capital contribution is \(0.40\times5.314875\%=2.125950\) percentage points. The labor weight is \(1-0.40=0.60\), so the labor contribution is \(0.60\times6.250000\%=3.750000\) percentage points. The total input contribution is \(2.125950+3.750000=5.875950\) percentage points.
The unrounded productivity residual is \(9.531020-5.875950=3.655070\%\). Rounded to three decimal places, this is 3.655%.
If the total input contribution is rounded first to 5.876, the subtraction gives \(9.531020-5.876=3.655020\%\). That is a different unrounded residual, but it also rounds to 3.655% to three decimal places. Therefore, in this example, premature rounding changes the unrounded residual from 3.655070% to 3.655020% but does not change the reported answer at the requested precision.
Check. The unrounded components add back to output growth: \(5.875950+3.655070=9.531020\%\).
Worked Example: Labor Composition Changes the Diagnosis
A hypothetical service sector has annual real output growth of 2.80%, capital-services growth of 4.00%, and hours growth of 1.00%. A labor-composition index grows by 0.80%. Assume \(\alpha=0.30\), a stable production relationship, and accurate measures of output and capital services. Use the growth-accounting approximation.
Method. First combine hours and labor composition to measure labor input; then calculate the residual. The common mistake is to treat hours alone as total labor input when the problem supplies a composition measure.
Labor-input growth is \(1.00\%+0.80\%=1.80\%\), because \(L=H Q_L\) and log growth adds. The capital contribution is \(0.30\times4.00\%=1.20\) percentage points. The labor weight is \(1-0.30=0.70\), so the labor contribution is \(0.70\times1.80\%=1.26\) percentage points. Total input contribution is \(1.20+1.26=2.46\) percentage points.
Measured productivity growth is \(2.80-2.46=0.34\%\). As a comparison, ignoring composition would produce a labor contribution of \(0.70\times1.00\%=0.70\) percentage points and a total input contribution of \(1.20+0.70=1.90\) points. That would give a residual of \(2.80-1.90=0.90\%\), overstating the residual by \(0.90-0.34=0.56\) percentage point.
Check. With composition included, \(1.20+1.26+0.34=2.80\%\). The 0.56-point difference equals the labor weight times composition growth: \(0.70\times0.80\%=0.56\) percentage point.
Worked Example: A Price-Index Error Can Inflate Measured Productivity
A hypothetical producer reports nominal output growth of 5.00%. Its output-price index rises by 2.00%. Assume the reported nominal growth and price change are ordinary percentage changes. Capital-services growth is 3.00%, hours growth is 1.00%, labor-composition growth is 0.50%, and \(\alpha=0.35\). Assume the input measures are accurate.
Method. Convert nominal output growth into real output growth before calculating the residual. The trap is to use nominal output growth as though it measured the quantity of production.
The exact real-output growth rate implied by the nominal output and price index is \((1.05/1.02)-1\). The quotient is \(1.02941176\), so real output growth is \(0.02941176\), or approximately 2.941176%. This conversion uses the stated ordinary growth rates; simply subtracting the price growth would give the approximation \(5.00\%-2.00\%=3.00\%\), not the exact rate.
Labor-input growth is \(1.00\%+0.50\%=1.50\%\). The capital contribution is \(0.35\times3.00\%=1.05\) percentage points. The labor weight is \(1-0.35=0.65\), so the labor contribution is \(0.65\times1.50\%=0.975\) percentage points. Total input contribution is \(1.05+0.975=2.025\) percentage points.
Using the real-output rate, the residual is approximately \(2.941176-2.025=0.916176\%\). If nominal growth were mistakenly used, the residual would instead be \(5.00-2.025=2.975\%\), an overstatement of approximately \(2.975-0.916176=2.058824\) percentage points.
Check. Correctly measured real output growth is approximately the sum of its input contribution and residual: \(2.025+0.916176=2.941176\%\). The large difference in the two residuals arises from using nominal rather than real output, not from a change in the production model.
Interpreting the Evidence Carefully
A falling residual is evidence of weaker measured productivity growth under the chosen accounting framework. It does not, by itself, prove that innovation slowed or that technology deteriorated. The residual can also reflect unmeasured changes in capacity utilization, capital quality, labor skills, product quality, or the allocation of resources. Data revisions and differences in timing can matter as well.
A useful diagnosis proceeds in stages:
Use real output rather than nominal expenditure, and check that the price index corresponds to the output being measured.
Check whether capital quality and utilization, hours, and labor composition are reflected in the input measures.
Apply the stated weights consistently, retain precision during calculation, and verify that contributions plus residual equal output growth.
State whether the slowdown comes from slower measured input growth, a lower residual, or both, and identify measurement limits that could change the interpretation.
A practical exam response should distinguish three statements: output growth slowed; input growth contributed less; and measured productivity growth changed. The first does not automatically establish the second or third. The decomposition quantifies the accounting contributions, while the interpretation depends on the assumptions and data quality.
Check Your Understanding
For each problem, use the stated production weights and growth-accounting approximation unless the question asks for an exact conversion. Keep intermediate values long enough to avoid unnecessary rounding.
- A hypothetical plant reports output growth of 2.4%, capital-services growth of 2.0%, and labor-input growth of 1.0%. If \(\alpha=0.40\), calculate the two input contributions and the residual. (Answer: Capital contribution 0.80 percentage points; labor contribution 0.60 percentage points; residual 1.00%.)
- In a second period, output grows 2.0%, capital services 1.0%, and labor input 2.0%. With \(\alpha=0.40\), compare the residual with that in Question 1. Does the lower output growth indicate lower measured productivity growth? (Answer: Capital contribution 0.40 points; labor contribution 1.20 points; residual 0.40%. Yes, the residual is lower by 0.60 percentage point, even though output growth is lower by only 0.40 point.)
- A sector reports hours growth of 0.5% and labor-composition growth of 1.0%. If \(\alpha=0.25\), what is labor-input growth and its contribution to output growth? What would the labor contribution be if composition were ignored? (Answer: Labor-input growth 1.5%; adjusted contribution 1.125 percentage points; hours-only contribution 0.375 percentage points.)
- Nominal output grows 4.0% and its price index grows 1.0%. Find exact real-output growth using the ratio of gross growth factors. If total weighted input contribution is 1.8%, calculate the residual. (Answer: Real-output growth \(1.04/1.01-1=0.02970297\), or approximately 2.9703%; residual approximately 1.1703%.)
- A reported residual declines after hours data are revised upward, while the output and capital data remain unchanged. Explain the accounting direction of the change and give one reason the revision alone cannot establish a true decline in technology growth. (Answer: Higher measured labor-input growth raises the labor contribution and, with output growth fixed, lowers the residual; the revision may reflect corrected measurement or timing rather than a change in technology.)