You are viewing the official metadata. View crowdsourced contributions.
Variable Name | tfp4 |
Top Access Level | released |
Label | 4-factor TFP index 1997=1.000 |
Codebook | NBER-CES Manufacturing Industry Database (2009) [NAICS] |
Concept | |
Type | numeric |
Files
naics5809.dta
http://www.nber.org/data/nberces5809.html
(
Stata
)
naics5809.sas7bdat
http://www.nber.org/data/nberces5809.html
(
SAS
)
naics5809.xls
http://www.nber.org/data/nberces5809.html
(
Excel spreadsheet
)
naics5809.csv
http://www.nber.org/data/nberces5809.html
(
CSV
)
Full Description
There are two versions of TFP in the NBER-CES Manufacturing Database: 4-factor and 5-factor. The 5- factor version separates out energy from non-energy materials; the 4-factor uses a single materials input (which includes energy). The TFP calculation requires definitions of the cost shares, the factors, the factor changes, and the output changes. The five cost shares (alpha_i) vary by industry by year, defined using the variable names in the dataset:
- (alpha_1) Non-production workers: (pay-prodw)/vship [i.e., their pay divided by shipments]
- (alpha_2) Production workers: prodw/vship [i.e., their pay divided by shipments]
- (alpha_3) Energy: energy/vship [i.e., energy expenditure divided by shipments]
- (alpha_4) Materials: (matcost-energy)/vship [i.e., non-energy materials divided by shipments]
- (alpha_5) Capital: 1 - (sum of the above shares) [i.e., the residual]
In calculating TFP growth from one year to the next, we use the average of the two years' cost shares: bar alpha_{it} = 0.5alpha_{it} + 0.5alpha_{it-1} The 5 factors (X_i) are defined as follows, using the variable names in the dataset:
- (X_1) Non-production workers: (emp-prode) [i.e., the number of non-production workers]
- (X_2) Production workers: prodh [i.e., production worker hours, not employees]
- (X_3) Energy: (energy/pien) [i.e., real energy expenditures]
- (X_4) Materials: ((matcost/pimat) - (energy/pien)) [i.e., real non-energy materials]
- (X_5) Capital: cap [i.e., total capital stock, already in real terms]
The change in factor usage between one year and the next is defined as the change in natural logs, (for example): dX_{it} = ln(X_{it}) - ln(X_{it-1})
We also need the change in real output: (Q) Real output: vship/piship
As with factor usage, we express output change in terms of natural logs, hence: dQ_t = ln(Q_t) - ln(Q_{t-1})
The change in 5-factor TFP (dTFP5) between this year and last is thus defined as: dTFP5_t = dQ_t - sum_i(alpha'_{it}dX_{it}), i = 1,..., 5
Given the series of dTFP5 values, one can then "roll up" these changes to form a TFP index (TFP5), by setting the index equal to 1.0 in some initial year t and then growing the index forward by the following equation: TFP5_{t+1} = exp[ ln(TFP5_t) + (dTFP5_{t+1})]
The values of 4-factor TFP growth (dTFP4) and the corresponding TFP index (TFP4) are calculated similarly, but using total materials cost spending rather than separating it into energy and non-energy materials.
Summary Statistics
Valid values | 24167 |
Invalid values | 429 |
Minimum | 0.011 |
Maximum | 13.193 |
Mean | 0.936 |
Standard deviation | 0.257 |
Value Ranges
Value Range
Range: [ 0.0109999999403954 , 13.1929998397827 ]