What are "0-kilometer used cars"?
"0-kilometer used cars" refer to vehicles that are registered but have not actually been driven, yet are marketed and resold as used cars. In recent years, this phenomenon has attracted widespread attention in China's automotive market.
According to a Reuters report, in May 2025, China's Ministry of Commerce convened a meeting with automakers such as BYD and Dongfeng Motor, along with the China Automobile Dealers Association, to discuss the issue of "0-kilometer used car" sales. These vehicles, although registered and licensed, have never been actually driven. They are resold as used cars, a strategy adopted by some automakers to meet sales targets. This practice affects the authenticity of sales data and has drawn high attention from both the market and regulators.
The reason the topic of "0-kilometer used cars" is so sensitive is that it essentially touches upon the most core "key variable" in an industry's data system—actual sales volume.
In the new energy vehicle industry chain, no matter how many times a vehicle completes "transactions" through the chain or is wrapped in capital narratives, if there is no real end consumer making the purchase, all of it lacks substance. Like the relationship between noise and signal in statistical analysis, actual sales volume is the "signal" that cannot be ignored.
What is a key variable?
The essence of statistics is "extracting order from chaos," and a "key variable" is the most important factor driving changes in outcomes.
In Daniel Kahneman's Thinking, Fast and Slow, two cases are particularly suitable for explaining this:
- Wine Price Prediction: A study found that different wine experts' predictions of future wine prices were like blind guesses or monkeys throwing darts. However, when researchers used a simple linear regression model with "weather conditions" (such as spring rainfall, summer average temperature) as input variables, the results significantly outperformed expert judgments. The reason is simple: weather variables are the true key variables determining grape quality and price; the rest of the expert opinions are merely background noise.
- Divorce Probability Prediction: Another example is even more insightful—whether a marriage is stable does not depend on whether the couple shares common interests or goes on holiday trips, but on a nearly simple indicator: "whether the frequency of making love exceeds the frequency of arguments." Although simple, this variable has strong predictive power. The reason is the same: it captures the quality of the core relationship, not superficial phenomena.
One involves variable selection under logical modeling, and the other involves variable compression based on intuitive experience. Both reveal the same truth—there aren't many truly important variables, but they are extremely decisive.
The Statistical Significance of "0-Kilometer Used Cars" and Sales Volume
Returning to the electric vehicle industry.
The essence of the current "0-kilometer used car" phenomenon is that companies "cook the books" by moving inventory vehicles through internal transactions, secondary market platforms, etc., to create an appearance of sales. This behavior affects precisely the core variable in market analysis—end-user sales volume (or "effective deliveries").
- From a logical analysis perspective, if an industry's sales volume does not truly reflect consumer purchasing behavior, then the entire foundation for judgments about user demand, market share, and cash flow models becomes shaky.
- From a data structure perspective, in a set of industry analyses, we can observe many variables such as shipment volume, inventory turnover rate, loan notes, used car platform listings, and customer complaints. However, most of these are "covariates" or "secondary indicators." The one that truly constitutes the system's signal is still sales volume.
In other words: Sales volume is a decisive explanatory variable in the entire model. When it is compromised, the entire model's predictions become invalid.
The Cost of Distorting Key Variables
Why did the "0-kilometer used car" incident quickly attract market attention?
Because in an industry driven by narratives to achieve capital premiums, if the key variable of sales volume cannot be verified, the story itself may collapse. Once the core variable in a statistical model is distorted, the confidence interval of the results widens, or even loses its explanatory power.
In recent years, some noteworthy data phenomena have emerged in the new energy vehicle industry, reminding us to be particularly vigilant about potential statistical biases when observing key variables:
- Financial phenomena reflecting payment pressure, such as lengthening accounts receivable cycles, are gradually increasing.
- The frequency of using certain financial instruments (such as bill mortgages, etc.) is rising.
- In some used car markets, listings for vehicles in nearly new condition appear relatively concentrated.
Although these signals themselves cannot directly prove sales data fraud, they constitute external observational clues regarding the reliability of key variables and warrant further scrutiny.
These variables might be "weak signals" in the model. But as long as the core variable—actual sales volume—cannot be confirmed or is tampered with, the entire system enters the statistical uncertainty interval of data distortion.
However, what is truly unsettling is that this distortion has a "counterintuitive" characteristic.
In statistical analysis, we often rely on so-called "hard data" to judge trends—such as revenue in financial statements or sales volume in industry reports. These data are widely cited because they appear quantifiable, verifiable, and comparable. They are not as subjectively influenced as "soft variables" like sentiment indices or consumer confidence, nor are they as difficult to falsify as certain model assumptions.
Yet, the "0-kilometer used car" phenomenon shatters a cognitive illusion:
The most transparent data can also be the most easily contaminated.
Sales data appears transparent and is even considered the primary indicator of industry health. But if companies can create a facade of "surface transactions" through related-party transactions, dealer inventory loading, bill mortgages, and used car listings, then this seemingly reliable variable precisely becomes a narrative tool "most easily used to tell a story."
This is a profound statistical counterintuition phenomenon—
We usually assume that the higher the "exposure level" of a variable, the higher the cost of falsifying it. But in certain systems, precisely because the exposure level is high, manipulating these variables can maximize the "narrative effect," making them more worthy of being falsified.
The consequences of this counterintuition are disastrous:
- Investors misjudge trends: Using sales data as the primary basis for decisions leads to increased investment driven by inflated data.
- Companies reinforce their own illusions: Mistaking internal circulation sales for market demand, thereby expanding production capacity and misjudging real demand.
- Industry regulation lags: Because the distortion of key variables is not easily detected in time, it often only begins to be traced after a "collapse."
Don't Let Variable Narratives Replace Variable Facts
Kahneman's viewpoint essentially questions this: humans are too easily hijacked by narrative logic, rather than letting the data speak for itself. The same applies in the new energy vehicle industry. Do not use narrative variables like "license plate quotas," "penetration rates," or "replacement cycles" to cover up real data gaps.
Capital markets can accept volatility and uncertainty, but they struggle to accept "fake variables." An untrustworthy sales data point is like a manipulated independent variable in a predictive model; it makes the coefficients of all related variables suspicious and ultimately renders the entire systemic judgment invalid.