Clive Granger

Clive Granger: The Man Who Broke Econometrics to Fix It

In 1974 Clive Granger and Paul Newbold published a result that should have been humiliating for their profession. Applied to the kind of data economists actually work with, the standard regression methods everyone was using could report strong, statistically convincing relationships between variables that had no relationship at all. As the Nobel committee would later put it, conventional approaches "could yield wholly misleading results when applied to analysis of nonstationary data." A great deal of published economics was, in effect, measuring nothing.

The Boy the Teacher Wrote Off

Clive William John Granger was born in Swansea on 4 September 1934, into a family that moved frequently through his childhood. A primary school teacher delivered a discouraging verdict on his prospects — the sort of judgement that usually settles the matter. It did not: encouraged by his father and by friends, he went to university anyway.

He read mathematics at the University of Nottingham, taking his degree in 1955, and stayed on for doctoral work under Harry Pitt. His thesis, completed in 1959, was called "Testing for Non-stationarity," and the choice of subject was strategic rather than accidental. Time series analysis was a field he thought needed development, and he went into it deliberately. He had secured a junior lectureship at Nottingham at twenty-three.

Princeton and the Spectral Detour

A postdoctoral year at Princeton from 1959 to 1960 put him alongside John Tukey, one of the twentieth century's great statisticians, working on spectral analysis — the decomposition of a signal into its component frequencies. It is an engineer's toolkit rather than an economist's, and importing it into economics is a good early example of what Granger habitually did: take a technique from a field with better mathematics and point it at economic data.

He returned to Nottingham, where he spent more than two decades as professor of applied statistics and econometrics, before moving to the University of California, San Diego, where he would remain for three decades and collaborate with Robert Engle.

Granger Causality

His 1969 paper in *Econometrica* gave the field the idea that still carries his name. The question it addresses is old and slippery: what does it mean to say one economic variable drives another? Granger's answer was operational rather than philosophical. If knowing the past of series X improves your prediction of series Y beyond what Y's own past already tells you, then X is said to Granger-cause Y. It is a definition built entirely out of predictive content, and its virtue is that it can actually be tested against data rather than argued about.

The Spurious Regression Problem

The 1974 paper with Paul Newbold was the demolition. Economic series — prices, output, money supply — typically trend. They are non-stationary: their statistical properties change over time. Feed two such series into a conventional regression and the machinery will frequently announce a strong relationship where none exists, because both are drifting and the test is not built to tell drift from dependence. Granger, in the obituary's phrase, "uncovered the phenomenon of spurious regression," and in doing so cast doubt on a large body of existing empirical work.

Plenty of careers have been built on that kind of destructive result. Granger's distinction is that he did not stop there.

Cointegration

The constructive answer arrived in 1987, in an *Econometrica* paper written with Robert Engle. The insight is that two non-stationary series may nonetheless be tied to one another, wandering individually but never drifting far apart — sharing a common trend, so that some combination of them is stable. Granger called such series cointegrated, and the paper supplied the machinery to test for the condition and to model the relationship properly when it holds.

This solved the problem the 1974 paper had created. Where before the choice had been between a method that lied and no method at all, economists could now distinguish a genuine long-run relationship from a coincidence of trends. It turned out to apply well beyond finance: Granger demonstrated its relevance to commodity prices, interest rates, electricity demand, deforestation, river flooding and sunspots.

In 2003 he and Engle shared the Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel, in recognition of their contributions to the analysis of time series data; his half was awarded specifically "for methods of analyzing economic time series with common trends (cointegration)." He was knighted in 2005. Nottingham established the Granger Centre for Time Series Econometrics in 2006. He died in San Diego on 27 May 2009, aged seventy-four, survived by his wife Patricia and two children.

Why Clive Is Called a Genius

The particular faculty here is easy to state and rare to find: Granger noticed that a universally trusted procedure was broken, and then built the thing that fixed it. Either half alone would make a substantial career. Spotting spurious regression required a mathematician's suspicion of a result everyone else found comfortable — the willingness to ask whether a test that keeps returning significant findings is detecting truth or manufacturing it. Building cointegration required the opposite temperament: patient constructive work, thirteen years later, on the specific problem his own critique had exposed. Very few people who break a field stay to repair it.

There is also the matter of what he chose to work on. At twenty-three he picked time series analysis because he judged the field underdeveloped, and spent fifty years importing tools from better-equipped disciplines — Tukey's spectral methods among them — into an area with messy data and weak mathematics. That is a strategic intelligence about where effort will pay, which is not the same as raw brilliance and is arguably scarcer.

The counter-case is real and Granger made most of it himself. In his Nobel memoir he described himself as "naturally lucky," with a gift for turning up useful information — not the self-portrait of a man who thought he was a genius. The word does not appear in the citations; the director of the centre named after him reached for "enormous talent" and "generosity of spirit," which is praise of a different kind. Every one of his signature results is shared: spurious regression with Newbold, cointegration with Engle, the Nobel itself split down the middle. And his contributions are methodological. Granger did not produce a theory of how economies work; he produced instruments for finding out, which means his influence is enormous but entirely mediated by the people who used them well or badly.

Legacy

Granger's tools are now so embedded in empirical economics that using them is unremarkable, which is the strongest form of influence available to a methodologist. Cointegration and error-correction modelling are standard equipment for anyone studying interest rates, exchange rates or consumption; Granger causality tests appear in fields far from economics, including neuroscience and climate science. He also left the discipline a permanent lesson in intellectual hygiene: that a statistical procedure returning confident answers is not thereby returning correct ones, and that the burden falls on the researcher to know which. Nottingham's Granger Centre carries the name of a man a schoolteacher once judged unlikely to amount to much.

Achievements

• Prize in Economic Sciences in Memory of Alfred Nobel — 2003
• Affiliated with University of California, San Diego and University of Nottingham
• Educated at University of Nottingham
• Worked as economist, statistician and university teacher