A proxy in science is a measurable variable used in place of something that cannot be measured directly. Scientists rely on proxies when the thing they actually want to study is too far away, too far in the past, too expensive, or too complex to observe or measure on its own. The concept appears across nearly every scientific discipline, from climate science to medicine to astronomy, and understanding how proxies work helps make sense of how researchers draw conclusions about things they can never directly see.
How Proxies Work
The core idea is straightforward: if you can’t measure X, find something else (Y) that reliably tracks with X and measure that instead. Oxford’s Dictionary of Statistics defines a proxy variable as “a measurable variable that is used in place of a variable that cannot be measured.” The relationship between the proxy and the real variable of interest has to be strong and well-understood, or the conclusions drawn from it will be unreliable.
A simple everyday example: if a survey researcher can’t interview someone directly, they might ask that person’s spouse instead, since people in the same household often share similar views. The spouse’s answer is a proxy for the absent person’s answer. It’s not the real thing, but it’s close enough to be useful. Scientific proxies follow the same logic, just with more rigorous validation behind them.
Climate Proxies: Reading Earth’s Past
Climate science is where most people first encounter the word “proxy,” because there were no thermometers thousands of years ago. To reconstruct past temperatures, rainfall, and atmospheric composition, researchers turn to natural records that preserve climate information in physical or chemical form.
Tree rings are one of the most intuitive examples. Each year a tree grows a new ring, and the thickness of that ring reflects growing conditions. In years with optimal temperature and rainfall for that species, the ring is thicker. In harsh years, it’s thinner. By measuring ring widths across very old trees or preserved timber, scientists can infer temperature and precipitation patterns stretching back centuries. Scars and burn marks in the wood can also reveal the timing of past wildfires.
Ice cores go even further back. Deep cores drilled from glaciers in Antarctica and Greenland contain layers of compressed snow that fell tens or hundreds of thousands of years ago. Tiny air bubbles trapped in the ice preserve samples of the ancient atmosphere, letting researchers measure past levels of carbon dioxide and other gases. The chemical makeup of the ice itself, specifically the ratio of different forms of oxygen and hydrogen atoms, reveals the temperature at the time the snow originally fell. Pollen grains trapped in these layers show what plants were growing nearby, which adds another line of evidence about past conditions.
Ocean sediment cores work similarly. Layers of mud on the seafloor contain the shells of microscopic organisms like diatoms and foraminifera. The chemistry and species composition of these tiny fossils change with ocean temperature, salinity, and acidity. By analyzing sediment cores layer by layer, researchers can build climate records going back millions of years.
Medical Proxies: Surrogate Endpoints
In medicine, proxies go by a different name: surrogate endpoints. The FDA defines a surrogate endpoint as “a laboratory measurement or physical sign that is used in therapeutic trials as a substitute for a clinically meaningful endpoint that is a direct measure of how a patient feels, functions, or survives.” In plain terms, it’s a number doctors can check now that predicts what will happen to a patient later.
Blood pressure is a classic example. What researchers really care about is whether a treatment prevents strokes, heart attacks, and kidney failure. But those events take years to develop, and waiting that long to evaluate a drug would be impractical. Blood pressure serves as a validated proxy because decades of data confirm that lowering it reliably reduces the risk of those outcomes. Other validated surrogates include LDL cholesterol (for cardiovascular disease risk) and HbA1c, a blood marker that reflects average blood sugar over several months (for diabetes management).
Clinical trials also use functional proxies. The six-minute walk distance, which is simply how far a patient can walk in six minutes, serves as a proxy for overall cardiovascular fitness and disease progression in heart failure studies. Depression scores on standardized questionnaires stand in for the harder-to-quantify concept of mental well-being.
Astronomy: Measuring What You Can’t Reach
Astronomers face perhaps the most extreme version of the proxy problem. You can’t travel to a distant galaxy and measure how far away it is with a ruler. Instead, astronomers use objects called “standard candles,” which are stars or events with a known intrinsic brightness.
Cepheid variable stars are the most famous example. These stars pulse in brightness at a rate that’s directly tied to how luminous they actually are, a relationship first established by Henrietta Leavitt in 1908. By measuring how fast a Cepheid pulses, astronomers know its true brightness. Comparing that to how bright it appears from Earth reveals its distance. Harlow Shapley used this technique to map the Milky Way’s globular clusters, and later researchers used it to prove that galaxies like Andromeda were far beyond our own. Type Ia supernovae serve a similar role at even greater distances, acting as proxies that let astronomers measure the expansion rate of the universe.
Even the pixel-to-pixel brightness variations in telescope images can serve as a distance proxy. A nearby galaxy’s image will show more variation between pixels because each pixel captures fewer individual stars, while a distant galaxy appears smoother. This method, called surface brightness fluctuations, extends distance measurements out to galaxy clusters tens of millions of light-years away.
Social Science Proxies
Social scientists frequently need to measure things like poverty, quality of life, or socioeconomic status, but these concepts are broad and difficult to capture with a single number. In developing countries especially, detailed data on household income or spending is rarely collected in demographic surveys. Researchers instead rely on proxy indicators: years of education, type of housing, access to clean water, or ownership of assets like a radio or bicycle.
These proxies are convenient but imperfect. Research comparing proxy indicators against actual consumption data (the preferred measure of living standards) has found that common proxies are “very weak predictors of consumption per adult.” They can still detect broad patterns, such as whether education level is associated with health outcomes within a population, but they miss a lot of the variation between individual households. This is a useful reminder that a proxy’s value depends entirely on how tightly it tracks the thing it’s supposed to represent.
Why Proxies Can Go Wrong
The biggest risk with any proxy is that the relationship between the proxy and the real variable breaks down without the researcher realizing it. This can happen in several ways.
Confounding variables are the most common problem. A confounding factor is something that influences both the proxy and the outcome, creating the illusion of a direct connection where none exists, or hiding a real one. In clinical research, failing to account for confounders “can bias your study results and lead to erroneous conclusions.” For example, if a study uses income as a proxy for access to healthcare, it might miss that geography is the real driver: people in rural areas may have both lower incomes and fewer hospitals, but the lack of hospitals is the actual barrier, not income.
Proxies can also decouple from their target over time. A tree ring proxy calibrated against modern temperature records might not behave the same way during periods with very different atmospheric CO2 levels. A medical surrogate that predicts outcomes for one class of drugs might fail for a drug that works through a completely different mechanism. This is why scientists validate proxies using multiple independent lines of evidence rather than relying on a single one.
How Scientists Validate Proxies
Before a proxy is accepted, researchers test how well it agrees with direct measurements wherever overlap exists. In clinical research, this involves comparing the proxy measure against the gold-standard measure using statistical tools. Common approaches include comparing average scores on both measures, calculating correlation coefficients to see how tightly they track each other, and assessing sensitivity (how often the proxy correctly identifies a positive case) and positive predictive value (how often a positive proxy result turns out to be correct).
In climate science, validation works by comparing proxy reconstructions against the instrumental record during the period where both exist, typically the last 150 years. If tree ring data accurately reproduces known temperature variations during this overlap window, researchers gain confidence in using the same tree ring data to infer temperatures from centuries earlier. Using multiple independent proxies that converge on the same answer, such as ice cores, tree rings, and coral records all pointing to the same warming event, strengthens the case further.
No proxy is perfect, and scientists generally don’t treat them as though they are. The goal is to understand the uncertainty, quantify it, and use proxies within the bounds of what they can reliably tell you.