Sampling Bias – The Mirror That Lies
Imagine you want to know the average height of students in your school. You stand outside the basketball court and measure everyone who walks out after practice. Your result: the average height is 6 feet. But is that true for the whole school? No — you only measured tall students. Your sample was biased.
That is sampling bias in a nutshell: the error that occurs when the group you actually observe is not representative of the group you want to draw conclusions about.
The Precise Meaning
In economics, we rarely study an entire population — it is too large and expensive. Instead, we take a sample (a subset) and use it to estimate something about the population (the whole group). Sampling bias creeps in when the method of selecting the sample systematically favours certain kinds of people, households, firms, or regions over others.
The result is that your estimate is systematically wrong — not just a little off by chance, but skewed in a particular direction. This is different from random sampling error, which averages out over many samples. Bias does not average out; it persists.
A large sample does NOT fix sampling bias. If your selection method is biased, a bigger sample just gives you a more precise wrong answer.
Why It Matters in Economics
Economic data drives policy. The government decides interest rates, tax slabs, welfare spending, and minimum wages based on surveys and sample data. If the sample is biased, the policy will miss its target.
Consider the Consumer Price Index (CPI) , which measures inflation. The government surveys households to find out what they buy and how much prices change. If the survey only covers urban households, it misses rural consumption patterns. The CPI will then understate or overstate the true inflation faced by the majority of Indians. That is sampling bias with real consequences.
CPI=∑(P0×Q0)∑(Pt×Q0)×100
Where Pt = current price, P0 = base year price, Q0 = base year quantity. The bias enters through Q0 — if the basket of goods is chosen from a biased sample, the index itself is distorted.
Common Types of Sampling Bias
Selection bias — You choose your sample in a way that excludes part of the population. Example: telephonic surveys during the day only reach people at home (the unemployed, retired, or very wealthy), missing working professionals.
Non-response bias — People who choose not to respond are different from those who do. In a survey about income, the very rich and the very poor often refuse to answer. Your data shows a middle-class picture that is false.
Survivorship bias — You only observe the 'survivors' and miss the failures. Studying successful startups to find the formula for success is useless if you ignore the thousands that failed with the same strategy.
A Diagram in Words …