Total Factor Productivity as a Growth-Accounting Residual

Why Growth Accounting Leaves a Residual

When real output grows, some of the increase may be associated with using more labor or capital. Growth accounting assigns measured contributions to those inputs, then identifies the part of output growth not explained by them. This remaining part is called total factor productivity growth, or TFP growth. It is useful precisely because it summarizes what the measured inputs alone do not explain; it is not, by itself, a direct observation of technology.

The calculation uses the production-function framework from growth accounting. In the standard Cobb–Douglas case, output depends on total factor productivity, capital, and labor. The weights determine how much measured growth in each input contributes to output growth. In applications, the weights are often approximated using capital’s and labor’s shares of national income. The framework is aligned with the SOA VEE Economics (Validation by Educational Experience) syllabus.

Definition: Total factor productivity growth is the portion of output growth left after subtracting the weighted growth contributions of measured capital and labor. In the Cobb–Douglas model \(Y=AK^\alpha L^{1-\alpha}\), it is the growth rate of \(A\), the total factor productivity term.

The Residual Formula

Let \(Y\) denote real output, \(K\) the measured capital input, and \(L\) the measured labor input. Suppose production is described by the constant-returns-to-scale Cobb–Douglas function \(Y=AK^\alpha L^{1-\alpha}\), where \(A\) is total factor productivity and \(0<\alpha<1\). The exponent \(\alpha\) is capital’s output elasticity; \(1-\alpha\) is labor’s output elasticity. With competitive factor markets and inputs paid their marginal products, these elasticities correspond to capital’s and labor’s income shares. When those assumptions are not appropriate, income shares may be imperfect estimates of the relevant weights.

For a variable \(Z\), define its log growth over a time interval as the change in \(\ln Z\) divided by the interval length. Log growth is often reported as a percentage per year. For modest changes, it is close to the ordinary percentage growth rate; for larger changes, the two need not be identical. The log-growth version of the decomposition is exact for the stated production function.

\Delta \ln Y=\Delta \ln A+\alpha\Delta \ln K+(1-\alpha)\Delta \ln L

Rearranging isolates the residual:

\Delta \ln A=\Delta \ln Y-\alpha\Delta \ln K-(1-\alpha)\Delta \ln L

For annual growth rates, the same relationship applies with each change expressed per year. If growth is reported in percentage points, keep all terms in percentage points throughout the calculation. The input contributions are weighted: capital growth is multiplied by \(\alpha\), and labor growth by \(1-\alpha\). The residual is what remains after subtracting both contributions from output growth.

Why the Decomposition Holds

The formula follows directly by taking natural logarithms of the production function. The logarithm changes multiplication into addition, and a constant exponent becomes a multiplier.

Take logarithms.
Starting from \(Y=AK^\alpha L^{1-\alpha}\), take natural logs of both sides: \(\ln Y=\ln A+\alpha\ln K+(1-\alpha)\ln L\).

Take changes over time.
Subtract the earlier-period equation from the later-period equation. The result is \(\Delta\ln Y=\Delta\ln A+\alpha\Delta\ln K+(1-\alpha)\Delta\ln L\), since the weights are held fixed in this model.

Isolate the productivity term.
Subtract the capital and labor contributions from both sides. Thus \(\Delta\ln A=\Delta\ln Y-\alpha\Delta\ln K-(1-\alpha)\Delta\ln L\).

This derivation also clarifies the meaning of “residual.” Once output growth and the input weights and growth rates are specified, TFP growth is calculated as the remaining term; it is not independently measured by this equation. The result depends on the production-function model and on how output and inputs have been measured.

Worked Example: Calculating a Positive TFP Residual

An economy’s annual real output growth is 5.0%. Its measured capital input grows by 6.0%, and its measured labor input grows by 2.0%. Use the Cobb–Douglas growth-accounting model with a capital weight of \(\alpha=0.35\). Find TFP growth. Treat the stated rates as annual log-growth rates, expressed in percentage points.

Method. Apply the weighted residual formula. The labor weight is \(1-\alpha=1-0.35=0.65\).

Capital contribution. Multiply capital growth by its weight:

0.35(6.0\%)=2.10\%

Labor contribution. Multiply labor growth by its weight:

0.65(2.0\%)=1.30\%

Residual. Subtract both contributions from output growth:

\Delta\ln A=5.0\%-2.10\%-1.30\%=1.60\%

Check. The accounted-for input contributions sum to \(2.10\%+1.30\%=3.40\%\). Adding the residual gives \(3.40\%+1.60\%=5.00\%\), which matches output growth. A common error is to subtract the full 6.0% capital growth and 2.0% labor growth without weighting them; the weights are part of the model and cannot be omitted.

Income Shares as Weights

In many applications, the production elasticities are not directly observed. A common procedure uses factor income shares as estimates. Under the competitive Cobb–Douglas assumptions above, capital’s income share estimates \(\alpha\), and labor’s income share estimates \(1-\alpha\). The shares should sum to one in this simplified calculation. If the data or assumptions do not support that interpretation, the resulting residual should be described as conditional on the chosen weights.

The next example uses factor shares rather than providing the capital weight directly. The arithmetic is the same: establish both weights first, calculate each weighted contribution, and then subtract their sum from output growth.

Worked Example: Inferring Weights From Factor Shares

A hypothetical economy has annual real output growth of 2.6%. Capital income is 35% of total factor income and labor income is 65%. Measured capital input grows by 5.0%, while measured labor input grows by 1.0%. Estimate TFP growth using the factor shares as Cobb–Douglas weights.

Method. The capital share gives \(\alpha=0.35\), and the labor share gives \(1-\alpha=0.65\). The shares are consistent because \(0.35+0.65=1.00\).

Capital contribution.

0.35(5.0\%)=1.75\%

Labor contribution.

0.65(1.0\%)=0.65\%

Residual.

\Delta\ln A=2.6\%-1.75\%-0.65\%=0.20\%

Check and common mistake. The two input contributions total \(1.75\%+0.65\%=2.40\%\); adding the residual \(0.20\%\) gives the stated 2.60% output growth. An unweighted average of input growth rates would be \((5.0\%+1.0\%)/2=3.0\%\), leading to the incorrect residual \(2.6\%-3.0\%=-0.4\%\). The correction is to use each input’s own income-share weight, not an arbitrary average.

Negative Residuals and the Meaning of TFP

A positive TFP residual means that output grew faster than the weighted growth of the measured inputs. A negative residual means that output grew more slowly than those weighted contributions. Neither sign, on its own, proves that technology improved or deteriorated. In this framework, “TFP” names the residual term in the model, so its interpretation depends on the completeness and quality of the measured inputs and output.

The residual can reflect technological change, but it can also incorporate other influences not separately represented in the calculation. Examples include changes in how intensively existing inputs are used, omitted or imperfectly measured inputs, changes in the composition of measured output, and errors in the data or in the chosen weights. These are reasons to interpret the residual as a broad accounting measure rather than a direct technology index.

Worked Example: Interpreting a Negative Residual

In a hypothetical production sector, annual real output growth is 1.0%. Capital income’s share is 40%, labor income’s share is 60%, measured capital input grows by 4.0%, and measured labor input grows by 2.0%. Calculate the residual and state what it does—and does not—show.

Method. Use the factor shares as weights: \(\alpha=0.40\) and \(1-\alpha=0.60\).

Capital contribution.

0.40(4.0\%)=1.60\%

Labor contribution.

0.60(2.0\%)=1.20\%

Residual.

\Delta\ln A=1.0\%-1.60\%-1.20\%=-1.80\%

Check and interpretation. The weighted input contributions total \(2.80\%\), and \(2.80\%+(-1.80\%)=1.00\%\), matching output growth. The negative residual says that measured output growth fell 1.80 percentage points short of the weighted input contributions under this model. It does not establish that technology declined by exactly 1.80%; measurement issues or omitted influences may also contribute.

Common Exam Traps

Leaving out a weight. Do not subtract all of capital growth and labor growth from output growth. Multiply each input growth rate by its corresponding weight first.
Using the same weight twice. In the two-input constant-returns Cobb–Douglas model, the labor weight is \(1-\alpha\), not \(\alpha\). Check that the weights sum to one.
Mixing time units. If output growth is annual but input growth covers a multi-year interval, convert the rates to a common interval before calculating the residual.
Confusing ordinary growth with log growth. The additive decomposition is exact in log changes under the stated model. Ordinary percentage changes provide an approximation for small changes; do not mix the two conventions within one calculation.
Calling the residual “technology growth” without qualification. The residual may include influences beyond technological change, so state what the model identifies and what it cannot identify separately.

Key takeaway: Calculate TFP growth by subtracting the weighted growth of measured capital and labor from output growth. Treat the result as a model-dependent accounting residual—not as a direct, complete measure of technological change.

Check Your Understanding

For each problem, use the stated growth-accounting convention and keep all rates on the same time basis.

An economy’s annual log growth rates are 3.4% for output, 4.0% for capital, and 2.0% for labor. If \(\alpha=0.25\), calculate TFP growth. (Answer: 0.90% per year)
Output grows by 1.8% per year, capital declines by 1.0% per year, and labor grows by 0.5% per year. If \(\alpha=0.40\), calculate the TFP residual. (Answer: 1.90% per year)
A hypothetical economy reports capital and labor income shares of 0.30 and 0.70. Output grows by 2.2%, capital input by 3.0%, and labor input by 1.0%. Find TFP growth using these shares as weights. (Answer: 0.60% per year)
Over a three-year interval, output rises by 15%, capital input by 9%, and labor input by 3%. Using \(\alpha=0.35\), calculate annual TFP log growth. Use \(\ln(1.15)/3\), \(\ln(1.09)/3\), and \(\ln(1.03)/3\) for the annual log-growth rates, then round the final result to two decimal places. (Answer: 3.01% per year)
A sector’s output grows by 4.0% per year, capital input by 2.0%, and labor input by 3.0%. Capital’s income share is 0.40 and labor’s is 0.60. Calculate the residual and give one reason it should not automatically be interpreted as technology growth. (Answer: 1.40% per year; the residual can also reflect omitted influences or measurement error.)

What You'll Learn