From Economic Concept to Recorded Variable
Economic questions often refer to concepts that are not directly recorded as a single number. “Employment,” “income,” and “productivity” sound concrete, but each can refer to several different things. Does employment mean having a job at a particular moment, working any hours during a week, or being employed for most of a year? Does income mean wages before taxes, all income received, or income per household member? A measure is useful only when its definition makes these choices explicit.
This process is called operationalization: specifying how an abstract economic concept will be represented by observations in a data set. Operationalization does not establish that one definition is universally correct. It makes the chosen definition precise enough to assess against the empirical question and, where appropriate, reproduce.
For example, “employment” is a construct. A rule that classifies an adult as employed if they performed any paid work during a specified reference week is an operational definition. The resulting variable might record 1 for employed and 0 for not employed. Those values describe the variable; they do not, by themselves, communicate the full rule. A reader also needs to know the population, period, and coding.
What a Reproducible Operational Definition Specifies
A useful operational definition makes the path from question to recorded value clear. Before coding, state the choices that determine what a value means:
- Construct and target: what concept is being represented, and for which units or population?
- Observational unit: is each record about a person, household, worker, firm, job, or another unit?
- Reference period: when, or over what duration, is the characteristic measured?
- Units: is the value expressed in dollars, dollars per month, hours, units per hour, or another scale?
- Inclusion and calculation rules: which components count, and how are they combined?
- Coding rules: how are categories represented, and what does each code mean?
- Data source: where does the recorded information come from, such as a survey question, payroll record, or firm production log?
The source is part of the description because sources may record different aspects of a construct. A payroll record, for example, can document payments processed by an employer, while a survey may ask about work performed during a reference week. Stating the source does not settle whether the measure is a good representation; it lets the reader understand what was actually recorded.
A compact variable specification could read: “For each sampled adult, employment status is 1 if the person performed any paid work or had a paid job from which they were temporarily absent during the seven days ending on the interview date; otherwise it is 0. Source: respondent report.” This is more informative than “employment: binary.”
Worked Example: Defining Employment Status
Worked Example: Defining Employment Status
A hypothetical survey asks six adults about their activity during the seven days ending on the interview date. The empirical question concerns the share of these adults who had paid employment during that week. For this exercise, define “employed” as having performed any paid work during the week or having a paid job but being temporarily absent. Include self-employment. Do not count unpaid household work as paid employment. Define the denominator as all six surveyed adults.
This rule makes the construct and reference period explicit. It also makes the classification reproducible: a person who worked for one hour for pay qualifies, while a person who did only unpaid household work does not. The survey report is the stated source for the classifications; this example does not assume the rule is the only reasonable definition of employment.
| Adult | Hypothetical reported activity during the reference week | Employment code |
|---|---|---|
| A | Worked 35 paid hours | 1 |
| B | Self-employed; completed paid work | 1 |
| C | Had a paid job; temporarily absent | 1 |
| D | Looked for work; did no paid work and had no job | 0 |
| E | Did unpaid household work only | 0 |
| F | Did no paid work and had no job | 0 |
With \(E_i\) denoting the employment code for adult \(i\), the share employed under this definition is the sum of the codes divided by the six adults:
Under the stated rule, 50.0% of these six surveyed adults are classified as employed in the reference week. This is a descriptive result for the records shown, not a claim about a broader population. Changing the reference period, denominator, or definition could change the resulting variable and share.
Amounts Need Units and Time Periods
An amount without a unit or period is difficult to interpret. “Income = 3,850” is incomplete: it could mean dollars in a month, dollars in a year, or another currency and period. “Wage” also needs care. A wage rate is an amount per unit of labor time, such as dollars per hour; earnings are an amount received over a period. They are related, but they are not the same variable.
A labor-income measure also requires a rule about which receipts count. For example, a study of monthly labor income might include wages and commissions received during a calendar month, while excluding investment income. A different study could define total household income, include other sources, and use the household rather than the person as the observational unit. These are different operationalizations, not interchangeable labels for the same recorded value.
Worked Example: Recording Monthly Labor Income
A hypothetical study asks how much labor income each adult received during April. Define the variable as the total amount in U.S. dollars received during April from wages and commissions, before deductions. Exclude investment income. Use each adult as one observational unit, and record zero if the adult received no labor income in that month. These choices make the measure a monthly amount, not an hourly wage or an annual income.
Suppose the hypothetical April records are: Ana received $3,200 in wages and $250 in commissions; Ben received $0 in wages and $0 in commissions; and Chris received $1,800 in wages and $400 in commissions. The calculation rule is to add the included components for each person:
For Ana, \(Y_{\text{Ana}}=\$3{,}200+\$250=\$3{,}450\). For Ben, \(Y_{\text{Ben}}=\$0+\$0=\$0\). For Chris, \(Y_{\text{Chris}}=\$1{,}800+\$400=\$2{,}200\). The resulting values are therefore $3,450, $0, and $2,200 per person for April.
The zero for Ben means that, under the stated rule, no included labor income was received in April; it does not mean the person had no income of any kind. Likewise, a monthly labor-income amount does not reveal the hourly rate without additional information about hours worked. The interpretation follows from the operational definition, not from the variable name alone.
Productivity: Define Both the Output and the Input
A productivity measure is a ratio of output to an input. “Productivity” alone does not specify which output, which input, or the period over which either is measured. Labor productivity, for example, may be defined as units of output per labor-hour. A measure of output per worker instead uses a worker count in the denominator, and generally has different units and meaning.
Before using a ratio, ensure that its numerator and denominator refer to a compatible unit and period. If output is a monthly total but labor input is recorded for a different period, the ratio does not represent output per labor-hour for that month unless the periods are aligned. The output unit also matters: the number of items produced is meaningful as a common quantity when the items are comparable, while a value-based output measure requires a clearly stated value unit.
Worked Example: Constructing Labor Productivity
A hypothetical workshop records its production and labor input for May. Define labor productivity as finished units produced during May per paid labor-hour during May. Assume the workshop produces one type of comparable unit, its production log counts finished units, and its time records sum the hours worked by all employees on production. These assumptions align the numerator and denominator to the same activity and period.
The log records 1,200 finished units. Two workers contribute 160 and 140 paid production hours, respectively, so total labor input is \(160+140=300\) labor-hours. The operational definition and calculation are:
Under this definition and these records, the workshop produced 4.00 finished units per production labor-hour in May. The result is not units per worker, and it does not by itself explain why productivity took this value or establish that one input caused another outcome.
A Short Justification: Aggregate Ratios Weight by Input
The ratio of total output to total labor-hours is not generally the same as the simple average of individual workers’ output per hour. The distinction follows directly from the definitions. Let worker \(i\) produce \(q_i\) units in \(h_i\) hours, and let total hours be \(H=\sum_i h_i\). The aggregate productivity measure is:
Thus aggregate productivity is a weighted average of workers’ output-per-hour rates, with weights \(h_i/H\). The weights sum to one because \(\sum_i h_i/H=H/H=1\). Workers who contribute more hours receive more weight in the aggregate ratio.
For example, if one worker produces 100 units in 10 hours and another produces 30 units in 10 hours, their rates are 10 and 3 units per hour. Since both contribute the same hours, the aggregate rate is \((100+30)/(10+10)=6.50\) units per hour, equal to their simple average \((10+3)/2=6.50\). If the second worker instead contributes 30 hours at the same rate, aggregate productivity is \((100+90)/(10+30)=4.75\) units per hour. The simple average of the two worker rates remains 6.50, but it no longer represents output divided by total hours. Which summary is appropriate depends on whether the question concerns the typical worker’s rate or output per total labor input.
Check That the Measure Matches the Question
An operational definition is a substantive decision, not just a formatting step. A careful reader should be able to move from the empirical question to the construct, then from the construct to a precise recorded variable. Small changes in the definition can alter what the resulting value represents.
For instance, a question about a worker’s current hourly compensation calls for a rate with currency-per-hour units. A question about the amount of labor income received in a month calls for a period total. A question about a firm’s labor productivity calls for a specified output measure divided by a specified labor input. Using one of these measures to answer a different question can create an interpretation mismatch even when every recorded value is accurate.
The definition should also be sufficiently detailed that another analyst could apply the same rules to the same records. For a categorical variable, state what each code means. For a quantitative variable, state its units and period, along with the components included in its calculation. If the intended construct has several plausible representations, explain why the selected one fits the empirical question rather than implying that the measure is the construct itself.
Common Mistake: Treating a Variable Name as a Definition
Wrong reasoning: “The data set has a variable called income, so it measures income; its values can be compared without further information.”
Correction: A label does not identify the observational unit, currency, reference period, included income sources, or whether the values are amounts or rates. A more informative statement is: “For each adult, the variable is total wages and commissions received in U.S. dollars during April, before deductions.” The operational rule supports a meaningful interpretation; the label alone does not.
Check Your Understanding
Use the definitions and examples above to assess what a measured variable represents.
- A study records whether each adult performed any paid work during the seven days before an interview. Name two additional details that would make the employment variable’s operational definition clearer.
- A researcher records $2,400 for a person’s earnings. What information is needed before the amount can be interpreted as a measured variable?
- Explain why a monthly labor-income amount and an hourly wage rate are not interchangeable, even if both concern paid work.
- A firm produces 900 finished units using 150 labor-hours in a month. Under a units-per-labor-hour definition, calculate the productivity measure and state its units.
- Why might aggregate output divided by total labor-hours differ from the simple average of workers’ output-per-hour rates?