Every time someone shows me a chart with China's credit impulse laid over copper or emerging-market equities, the pitch is the same. The two lines track each other, the credit line moves first, so you now have a two to four quarter heads-up on the reflation trade. I used to nod along to that. Then I went and rebuilt the series myself, tick by tick, and came away a lot more careful about how I use it. The indicator is real and the mechanism behind it is genuinely useful, but a chunk of what people quote is the version that worked in an earlier decade, and it has been quietly degrading for years.
What the credit impulse actually measures
Start with the raw ingredient, which is total social financing, or TSF. This is China's broadest official measure of new credit flowing into the real economy. It captures bank loans, but also the shadow stuff that a narrow loan number misses, so things like corporate bond issuance, entrusted loans, trust loans, bankers' acceptances, and government bond financing. The People's Bank of China publishes it monthly. The headline you usually see in the press is the flow of new financing in a given month, and there is also a stock figure that represents the total outstanding pool.
The credit impulse is not TSF itself. It is the change in the flow of new credit, scaled to the size of the economy. The standard construction is the year-over-year change in the flow of new TSF over a trailing twelve-month window, divided by nominal GDP. So you are not asking whether credit is growing. Credit in China is almost always growing. You are asking whether it is growing faster or slower than it was a year ago. That second derivative is the whole point. A positive and rising impulse means the credit spigot is opening wider than before, which historically front-ran a pickup in real activity and, downstream of that, in the prices of the things China consumes.
The reason it leads rather than coincides is mechanical. Credit gets extended first. Then it funds construction, land purchases, factory expansion, and inventory building. Only after that do you see the physical demand show up in iron ore, copper, oil, and the earnings of exporters into China. That transmission chain takes time, and the historical lag people quote, roughly two to four quarters, is really just the length of that pipeline from loan approval to a truck showing up at a port.
How to build it and where to get the data
You can source the TSF numbers straight from the PBOC's statistical releases, which come out monthly. If you want it cleaned and in a spreadsheet without wrangling Chinese-language releases, most of the big data terminals carry TSF stock and flow, and several central-bank and BIS datasets track China credit-to-GDP, which is a slower-moving cousin of the same idea. Here is roughly how I put a usable version together:
- Pull the monthly new TSF flow and sum it on a trailing twelve-month basis to smooth out the enormous seasonal swings. January in particular is a monster month for Chinese lending and will wreck an unsmoothed series.
- Take the year-over-year change in that trailing-twelve-month flow. That difference is your numerator.
- Divide by trailing nominal GDP. Since China only reports GDP quarterly, interpolate to monthly or just accept a quarterly cadence.
- Plot the result and lag it forward by two to four quarters when you overlay it on the asset you care about, because you are testing whether today's credit predicts tomorrow's price.
One warning that will save you an afternoon. Different people define the impulse slightly differently. Some use the twelve-month change, some annualize a shorter window, some use TSF excluding government bonds, some include it. These variants can point in genuinely different directions in a given quarter, so if you are comparing your chart to someone else's and they disagree, it is almost always a definition mismatch and not a data error. Pick one construction and stick with it so you are at least consistent with yourself.
What it historically led, and how well
The cleanest relationships were always with the things most directly tied to Chinese fixed-asset investment. Industrial metals sit at the top of that list, because China has historically been the marginal buyer of copper, iron ore, aluminum, and steel. When the impulse turned up, base metals tended to follow a couple of quarters later. Emerging-market equities came next, especially the commodity exporters and the Asian supply chain, since so many of them are geared to Chinese demand one way or another. Broader global reflation trades, meaning the reflation-sensitive corners of developed markets and cyclical sectors, were a looser and noisier third tier.
The rule of thumb I actually keep in my head is simple. The impulse is a decent early tell on the direction of Chinese demand, it is a weaker tell on magnitude, and it is close to useless on timing to the month. Treat it as a regime signal, not a trade trigger. If the impulse has clearly turned and stayed turned for two or three months, that is a reason to lean your commodity and EM exposure in that direction over the next few quarters. It is not a reason to buy copper on the print.
The post-property-bust caveats
Here is where I get uncomfortable with the tidy overlay charts. The credit impulse worked as well as it did because, for a long stretch, a huge share of new Chinese credit flowed into property and property-adjacent construction, which is about as commodity-intensive as economic activity gets. Every yuan of that credit pulled in steel and copper and cement almost mechanically.
After the property downturn, that transmission weakened. Beijing spent years trying to steer credit away from property and land speculation and toward manufacturing, green energy, and technology. A yuan lent to a semiconductor line or a battery plant does not drag in iron ore the way a yuan lent to a residential tower did. So you can now get a rising credit impulse that produces much less of the commodity pull you would have expected historically. A lot of recent credit expansion has also gone into refinancing existing debt and local-government balance-sheet repair, which shows up in TSF but funds very little new physical demand. That is credit that keeps the lights on rather than credit that builds something.
The practical upshot is that you cannot just resurrect the old lag and trade it. What I do now is pair the impulse with a second read on where the credit is going. If the impulse is rising and you can see it landing in construction, land, and heavy industry, the old commodity relationship still has some life. If it is rising but the money is clearly going into refinancing and favored high-tech sectors, treat the commodity signal as heavily discounted. The impulse still tells you the tap is open. It no longer tells you, on its own, what comes out of the pipe.
So I still watch it, and I still build it from the raw TSF every quarter. I just stopped treating a rising line as a green light on its own, and started treating it as one input that has to be squared against the composition of the credit and against what iron ore and copper are actually doing. When all three agree, that is worth something. When they disagree, the credit impulse is usually the one that is lying to you.