Introduction to Data

In modern economics, the term Statistics is employed in two distinct senses: the plural sense (statistical data) and the singular sense (statistical methods).

• Statistics in the Plural Sense (Numerical Facts): As defined by Horace Secrist, statistics refers to aggregates of facts, affected to a marked extent by a multiplicity of causes, numerically expressed, enumerated or estimated according to reasonable standards of accuracy, collected in a systematic manner for a predetermined purpose, and placed in relation to each other. Under this definition, an isolated numerical figure (e.g., "Ramesh earns ₹25,000 per month") is not statistics, whereas a systematic record of average household incomes across rural districts of West Bengal constitutes statistical data because it represents an aggregate susceptible to comparison.
• Statistics in the Singular Sense (Statistical Methods): Refers to the scientific body of methods and principles developed for the collection, organisation, presentation, analysis, and interpretation of quantitative data. It is the operational toolkit by which raw figures are converted into actionable economic intelligence.

Economics is essentially an empirical science that seeks to explain how scarce resources are allocated to satisfy unlimited human wants. Statistics serves as the indispensable bridge between abstract economic theories and observed economic realities:

While statistical analysis is powerful, students of economics must be acutely aware of its inherent constraints to avoid fallacious deductions:

• Deals Only with Quantitative Characteristics: Statistics cannot directly measure qualitative attributes such as honesty, intelligence, patriotism, or consumer welfare unless they are transformed into quantitative surrogate scales or ranks.
• Deals with Aggregates, Not Isolated Individuals: Statistical laws apply to mass observations. The per capita income of India (e.g., ₹1,80,000) does not reveal whether a specific individual in a Purulia village is destitute or affluent.
• Statistical Laws are True Only on Average: Unlike deterministic laws in physics (e.g., Newton's laws), economic relationships expressed statistically (such as the law of demand) hold true only probabilistically and in the aggregate, subject to the ceteris paribus assumption.
• Vulnerability to Misuse and Bias: Statistics can be deliberately manipulated by cherry-picking sample periods, utilizing inappropriate base years, or drawing ungrounded causal inferences from mere correlations. As the aphorism notes, "Statistics are like clay, of which you can make a God or a Devil as you please."

In economic research, Data (the plural of datum) denotes raw or processed facts, figures, and numerical observations collected systematically for inquiry. Data represent the empirical evidence that grounds theoretical economics in observable reality.

Economic phenomena manifest in two primary dimensions:

• Qualitative Data (Attributes): Observations that describe non-numerical qualities, characteristics, or categorical states. Examples include gender, employment status (employed, unemployed, self-employed), caste category, educational level (primary, secondary, tertiary), and occupational sector (primary, secondary, tertiary). Although non-numerical by nature, qualitative attributes can be coded (e.g., Male = 0, Female = 1) or converted into frequency counts and percentage distributions for statistical analysis.
• Quantitative Data (Variables): Observations that represent measurable quantities expressed in definite numerical units (e.g., rupees, tonnes, kilometers, years). A quantitative characteristic that varies from one individual or observation to another is termed a Variable (e.g., monthly income, crop yield per hectare, price of rice).

Quantitative variables are subdivided based on the mathematical continuum of values they can assume:

In empirical economics, data structures are rigorously classified according to their temporal and cross-unit dimensions:

• Time Series Data: A sequence of numerical observations on one or more variables recorded at successive, uniform intervals of time (e.g., daily, monthly, quarterly, or annually).
  Example: India's Annual Real GDP Growth Rate from 2010 to 2024; West Bengal's annual paddy production over the last 20 years. Time series data capture trends, cyclical movements, seasonal variations, and structural economic shifts.
• Cross-Sectional Data: Data collected on multiple economic entities (such as individuals, households, firms, districts, or countries) at a single specific point in time or during a single reference period.
  Example: Per capita income, literacy rate, and poverty ratio of all 28 Indian states recorded in the financial year 2022-23. Cross-sectional data reveal inter-regional, inter-household, or inter-firm disparities.
• Pooled / Panel Data (Longitudinal Data): A composite dataset combining cross-sectional and time-series dimensions, where the same cross-sectional units (e.g., a fixed cohort of 500 rural households) are surveyed repeatedly across multiple successive time periods (e.g., every year for 10 years). This enables economists to track dynamic adjustments and behavioral changes over time.

Depending on the origin and method of acquisition, statistical data utilized in economic inquiries are universally divided into two primary classes: Primary Data and Secondary Data.

Primary data are statistical figures collected for the first time by an investigator, research team, or enumerator directly from the primary units of observation (informants) for a specific, predetermined purpose.

• Originality: Primary data are original in character; they resemble raw materials freshly extracted from the mine.
• Custom Alignment: Because the inquiry is designed specifically for the researcher's stated objective, the definitions, concepts, and scope match the research questions precisely.
• Resource Intensity: Gathering primary data requires substantial financial expenditure, large teams of trained personnel, and extensive time.
• Example: A team of university researchers visiting handloom weavers in Shantipur, Nadia to record their exact daily earnings, working hours, and debt levels through personal interviews.

Secondary data are statistical figures that have already been collected, compiled, processed, and published (or preserved in unpublished records) by another individual, institution, or agency for some other objective, which are subsequently utilized by an investigator for their own analysis.

• Pre-Existing Origin: Secondary data are secondary in character; they resemble finished goods manufactured by one agency that serve as intermediate inputs for another researcher's analysis.
• Cost and Time Efficiency: Acquiring secondary data is vastly cheaper, faster, and less labor-intensive than conducting a primary census or sample survey.
• Risk of Misalignment: The original data may have utilized different concepts, definitions, geographic coverage, or reference periods, requiring careful adjustments and verification before use.
• Example: An economist studying national unemployment trends by analyzing tables published in the Periodic Labour Force Survey (PLFS) report of the National Statistical Office (NSO).

When an economic investigator resolves to collect primary data, the choice of methodology is governed by the geographic scope of the inquiry, the available financial budget, the required degree of accuracy, the time horizon, and the literacy level of the target population. There are five classical methods:

The investigator meets informants face-to-face, poses questions directly, and records their verbal responses on the spot.

• Merits:
  • Highest Accuracy and Originality: The investigator can observe non-verbal reactions and probe ambiguous responses.
  • High Response Rate: Personal presence minimizes non-response or outright refusal.
  • Uniformity: Questions are interpreted consistently by the investigator.
  • Flexibility: The interviewer can adjust language and tone to suit the informant's educational level.
• Demerits:
  • High Cost and Time: Extremely expensive and time-consuming, rendering it unfeasible for vast nationwide inquiries.
  • Interviewer Bias: The personal opinions, prejudices, or tone of the investigator can subconsciously distort findings.
  • Limited Coverage: Unsuitable when the population is geographically dispersed across remote regions.
• Ideal Use-Case: Intensive, localized inquiries such as studying child nutrition in a specific block or investigating an industrial strike at a manufacturing plant.

When informants cannot be contacted directly or are reluctant to disclose sensitive or incriminating facts, the investigator interviews third-party witnesses (called informants or testifiers) who possess intimate, reliable knowledge regarding the subjects of inquiry.

• Merits: Covers a wider area with lower cost; highly effective for sensitive investigations (e.g., drug addiction, tax evasion, alcoholism, causes of industrial accidents).
• Demerits: Information is secondhand; witnesses may be biased, indifferent, or intentionally conceal truth to protect colleagues.
• Ideal Use-Case: Government commission inquiries, police investigations, or labor dispute fact-finding committees.

The investigating organization appoints local agents or correspondents across different geographic zones. These agents collect periodic reports using their personal judgment and dispatch regular statistical updates to the central headquarters.

• Merits: Extremely cheap and fast continuous monitoring across expansive geographic territories.
• Demerits: Lacks precision and statistical standardization; heavily influenced by the subjective estimates of individual correspondents.
• Ideal Use-Case: Agricultural market price reporting (wholesale Mandi prices), daily stock market commentary, crop yield preliminary forecasts, and journalistic news dispatches.

A structured questionnaire containing printed questions is sent by post or electronic mail to informants across a wide geographical area, accompanied by a polite covering letter and a self-addressed stamped return envelope (or digital submission link). Informants read the questions, write down answers independently, and return the completed form.

• Merits:
  • Vast Geographic Coverage: Can span entire states or nations at minimal incremental cost.
  • Economy: Drastically lower cost compared to personal field interviews.
  • Absence of Interviewer Bias: Informants provide candid answers free from the psychological influence of an interviewer.
  • High Anonymity: Suitable for confidential personal queries (e.g., household budgeting, consumer satisfaction).
• Demerits:
  • Extremely High Non-Response Rate: Most recipients ignore or discard the questionnaire, inducing severe non-response bias.
  • Restricted to Literate Informants: Cannot be utilized in illiterate or semi-literate populations.
  • Rigidity and Misunderstanding: If an informant misinterprets a question, no clarification can be provided on the spot.
  • Incomplete and Careless Submissions: Informants often leave critical questions blank.

In this method, the questionnaire is termed a Schedule. Trained enumerators (field investigators) personally visit the informants, read out the standardized questions from the schedule, translate or clarify them if necessary, and record the informant's verbal answers directly on the form.

• Merits:
  • Near 100% Response Rate: Personal visits ensure complete schedules without missing entries.
  • Applicable to Illiterate Populations: Enumerators translate complex economic concepts into local vernacular dialects.
  • Superior Data Quality: Pre-trained enumerators avoid recording contradictory or nonsensical answers.
• Demerits:
  • Massive Financial Outlay: Requires colossal budgets to recruit, train, transport, and remunerate enumerators.
  • Enumerator Dishonesty: Inadequately supervised enumerators may fabricate answers without conducting real field visits.
• Ideal Use-Case: Decennial Census of India, NSSO nationwide socio-economic surveys, and large-scale agricultural censuses.

While both documents feature a structured list of questions designed to elicit empirical data, their administration method establishes a decisive methodological boundary:

The statistical validity of primary research is entirely dependent on the quality of the questionnaire. Statistical textbooks prescribe nine fundamental golden rules:

1. Polite and Purposeful Covering Letter: The questionnaire must begin with a cordial introduction clearly stating the research objective, the identity of the conducting institution, and an explicit legal guarantee of strict confidentiality.
2. Brevity (Optimal Number of Questions): The list of questions should be as short as possible. Overly long questionnaires induce respondent fatigue, leading to hasty, careless answers or mid-way abandonment.
3. Simplicity, Clarity and Unambiguous Phrasing: Questions must be framed in simple, everyday language, avoiding technical jargon, acronyms, or double negatives. (e.g., instead of "Are you not unsupportive of tariff de-escalation?", ask "Do you support lower import duties?").
4. Logical Sequence: Questions must progress systematically from general, easy-to-answer introductory queries (e.g., age, household size) to specific economic topics (income, expenditure, debt). Abrupt thematic jumps confuse informants.
5. Avoidance of Leading / Suggestive Questions: Questions must be completely neutral and must not hint at or nudge the respondent toward a particular answer.
   Defective: "Don't you agree that inflation has severely hurt poor households?" (Biased).
   Correct: "In your opinion, how has the recent price change affected your household budget?"
6. Avoidance of Sensitive, Intrusive or Offensive Queries: Direct inquiries into private marital disputes, criminal history, or exact wealth evoke suspicion or hostility. Where sensitive economic data (such as exact monthly income) are required, broad income class intervals should be provided rather than demanding an exact figure.
7. Avoidance of Questions Requiring Arithmetical Calculations: Never ask respondents to perform complex calculations (e.g., "What percentage of your annual aggregate expenditure was allocated to non-staple carbohydrates?"). Ask simple raw inputs and let the computer perform calculations.
8. Incorporation of Cross-Verification / Check Questions: Strategic check questions should be embedded to detect internal inconsistencies (e.g., asking "Date of Birth" in Question 3 and "Current Age" in Question 18).
9. Clear Instructions and Guidelines: Provide concise instructions beside questions (e.g., "Tick $(\checkmark)$ only one option" or "Specify amounts in Indian Rupees").

An ideal questionnaire blends four structured question types:

• Dichotomous Questions (Two-Way Choice): Suitable for simple binary states. Poses two mutually exclusive options, such as Yes/No, True/False, or Male/Female.
• Multiple Choice Questions (MCQ): Provides a comprehensive list of mutually exclusive and collectively exhaustive categories. (e.g., Primary source of irrigation: [ ] Canal [ ] Tubewell [ ] Monsoon rain [ ] Tank [ ] Other).
• Rating Scale / Likert Scale Questions: Gauges the intensity of an economic attitude or satisfaction level along a calibrated scale (e.g., 1 = Highly Dissatisfied to 5 = Highly Satisfied).
• Open-Ended Questions: Permits the informant to express unstructured personal qualitative viewpoints in their own words. Should be used sparingly due to difficulties in statistical coding and tabulation.

Before launching a large-scale, costly survey across an entire district, state, or nation, the draft questionnaire must be field-tested on a small, representative group of informants. This preliminary reconnaissance trial is termed a Pilot Survey (or Pre-Test).

• Core Objectives and Utility of a Pilot Survey:
  • Uncovering Latent Ambiguities: Identifies confusing, misleading, or offensive questions that caused hesitation during trial interviews.
  • Testing Informant Reaction: Assesses whether informants are willing to answer sensitive questions regarding income, debt, and assets.
  • Estimating Time and Resource Requirements: Measures the exact average time required to complete one questionnaire, enabling accurate forecasting of total survey duration, field staff requirements, and budgetary costs.
  • Training and Evaluating Enumerators: Serves as practical hands-on field training for enumerators, testing their ability to record answers faithfully and handle difficult respondents.

Secondary data utilized in economic analysis are broadly bifurcated into Published Sources and Unpublished Sources:

• Published Sources:
  • Government Publications: Official reports released by central ministries and state departments (e.g., Economic Survey by Ministry of Finance, Agricultural Statistics at a Glance by Ministry of Agriculture, RBI Annual Report). These possess the highest authenticity.
  • Semi-Government Publications: Statistical bulletins issued by municipal corporations, district zilla parishads, and development authorities regarding vital statistics (births, deaths, sanitation).
  • International Publications: World Development Indicators (World Bank), World Economic Outlook (IMF), Human Development Report (UNDP), and statistical releases by ILO and WTO.
  • Trade and Industry Associations: Reports published by apex commercial chambers such as FICCI, CII, ASSOCHAM, and the Indian Jute Mills Association (IJMA).
  • Research Institutes and Academic Journals: Scholarly findings published by institutions like the Indian Statistical Institute (ISI), National Council of Applied Economic Research (NCAER), and journals like Economic and Political Weekly (EPW).
• Unpublished Sources: Statistical records compiled by private enterprises, internal hospital records, trade union registers, and unpublished doctoral research dissertations that are not released into the public domain for commercial or confidentiality reasons.

Professor A.L. Bowley famously observed that "Secondary data should never be accepted at their face value without rigorous scrutiny." An economist must examine four fundamental audit criteria before utilizing secondary data:

1. Reliability of the Collecting Agency: Who collected the data? Was the agency competent, experienced, objective, and free from commercial or political bias? Government bodies (CSO, NSSO, RBI) carry high credibility, whereas biased interest groups may distort figures.
2. Suitability to Current Objectives: Does the original data match the scope and definitions of the current research? For instance, if an investigator requires figures on agricultural laborers, does the secondary report include marginal and seasonal unpaid family workers under that heading?
3. Adequacy of the Data: Is the geographical coverage, sample size, and time span adequate for drawing valid generalized conclusions? A survey restricted to urban Kolkata cannot be used to generalize about rural Purulia.
4. Degree of Accuracy and Methodology: What sampling method was adopted? What was the margin of sampling and non-sampling error? Were definitions consistently applied throughout the survey period?

The Indian statistical architecture is globally renowned for its scale, rigor, and methodological sophistication. Four institutions represent its principal pillars: