Taobao Bad Loan Auctions: What Online Prices Reveal About the NPL Market

China’s nonperforming loan (NPL) market is notoriously opaque. There is no official data that tracks changes in investor demand for bad loans or the price at which banks and asset management companies (AMCs) are able to dispose of them. What little insight we have comes from the local Chinese press, where reports are sporadic, anecdotal, and lag behind real-time developments.

Greater transparency surrounding NPL prices is important in determining the sustainability of China’s deleveraging effort. The higher the price at which banks sell their bad loans, the less they need to raise new capital to cover losses. And a buoyant NPL market allows banks to clean up their balance sheets more quickly, while a sluggish one complicates authorities’ efforts to return the financial system to health.

However, there is one subset of the NPL market where sales and price data is publicly available: Taobao. The Alibaba subsidiary has been auctioning NPLs on its platform since mid-2015. A rarity among online auction exchanges, which delete any online record of their auctions upon completion, Taobao publicly discloses the sale price of every NPL (or NPL bundle) sold on its platform and keeps a digital record. While Tabao auctions represent only a small portion of the overall NPL market, changes in the volume and price of bad loans sold on the platform have the potential to offer useful insights into developments in the broader market.

With the guidance of the Research Computing Center at the University of Chicago, we scraped every NPL auction notice posted on Taobao’s dedicated auction site from mid-2015 through March 2019. We collected data on the seller, auction date, and the success or failure of each auction, and in the case of successful auctions, we also recorded the sales price and the outstanding value of the underlying unpaid loans. We then stripped out all auctions where the seller was neither a bank nor an AMC (a number of auto finance companies, NPL investment funds, and independent auction houses also sell NPLs on Taobao). The result is a database of 5,382 auctions of NPLs. Most of those failed to attract a bidder, but 2,660 resulted in a sale.

Of course, the Taobao data has limitations. Taobao’s unique characteristics—in terms of the composition of buyers and sellers and the factors that influence prices—mean that the online market is not a proxy for, or a microcosm of, the broader NPL market. But the direction of prices and the overall dynamics of the Taobao NPL auction market appear to be shaped by the same macroeconomic trends that determine how the broader market prices bad loans.

This report examines the Taobao data in detail and aims to explain where it fits into China’s NPL disposal ecosystem. It will also show how Taobao’s aggregate pricing has changed over the last couple of years, explain some of the characteristics unique to the Taobao NPL market, and discuss why Taobao is a poor proxy for the broader market. And finally, it will explore how Taobao is still a useful window into the broader NPL market, despite its shortcomings.

Significant But Not Yet Representative

China’s NPL market is divided into a primary and secondary market, and Taobao plays a role in both.[1] Although the primary and secondary markets are priced differently, they nonetheless influence each other. For instance, AMCs are willing to pay banks more if there’s sufficient demand in the secondary market and pay less if secondary market prices are weak.[2]

The primary market—which is by far the larger of the two—is where banks sell their bad loans, mostly to the AMCs. By our estimate, banks raised about 750 billion yuan selling NPLs in 2018, a figure that excludes NPLs packaged into asset-backed securities (ABS) and those swapped for equity. Of that amount, commercial banks raised 10.9 billion yuan in 2018 from selling NPLs on Taobao, bypassing the AMCs.[3]

Taobao plays a far larger role in the secondary market, where the AMCs sell a portion of the NPLs they’ve acquired to third-party investors, usually at a higher price. In the secondary market, the AMCs raised nearly 33 billion yuan in 2018 using Taobao (see Figure 1). The true size of the secondary market is difficult to know since most AMCs don’t disclose the volume of bank loans they sell forward. However, Cinda and Huarong—two of the Big Four AMCs—raised 64.4 billion yuan combined from NPL sales last year (see Figure 2). Hence, while Taobao’s role in the overall NPL market is small, it’s not insignificant.

Figure 1. Primarily Secondary
The proportion of NPL sales on Taobao made by AMCs compared to banks
Sources: Taobao; MacroPolo.

Figure 2. Follow the Money
Partial estimate of the size of NPL flows in 2018 through various disposal channels

How to read this chart: The numbers represent the amount paid by AMCs and other investors for NPLs that passed through the various disposal channels. Figures with an asterisk represent the amount that AMCs earned from their NPL sales. Question marks indicate insufficient data to estimate the volume of NPL flows.
Note: Data for Great Wall and China Orient are based on MacroPolo estimates.
Sources: MacroPolo; AMCs; Peterson Institute for International Economics; Chinese media.

Not a Microcosm

There are several reasons why Taobao isn’t representative of the broader NPL disposal ecosystem: price, institutional concentration, and geographic focus.

First, when it comes to price, Taobao NPLs typically sell for far more than similar transactions in the primary and secondary markets. That’s largely because AMCs and institutional investors tend to buy NPLs in bulk, reducing the cost relative to Taobao where loans are typically sold off in ones and twos. Meanwhile, banks and AMCs have to pay auction fees to Taobao, in addition to paying independent appraisers to value each loan posted on the site. While that may not fully account for the pricing difference, NPLs sold on Taobao nonetheless sell at a significant markup from the rest of the primary and secondary markets.

Second, loans auctioned on Taobao are dominated by a handful of financial institutions. Notably, Cinda is the single biggest seller of NPLs on Taobao, despite buying fewer NPLs in 2018 than Great Wall and China Orient.[4] In 2018, Cinda raised 13.4 billion yuan from NPL sales on Taobao (31% of the total), Zheshang Asset Management—one of Zhejiang province’s AMCs—raised 6.7 billion yuan (15%), Great Wall raised 6.3 billion yuan (14%), and Ping An Bank raised 4.6 billion yuan (11%) (see Figure 3).

Figure 3. Institutionally Concentrated…
Proportion of NPL sales made by most active financial institutions on Taobao
Sources: Taobao; MacroPolo.

Finally, the lion’s share of NPLs auctioned on Taobao are concentrated in a handful of provinces, with Zhejiang, Guangdong, Inner Mongolia, Yunnan, and Sichuan accounting for 69% of the NPL sales value in 2018 (see Figure 4).

Figure 4. …And Geographically Focused
Proportion of NPLs auctioned on Taobao, by province
Source: Taobao; MacroPolo.

So for the Taobao database to be a truer microcosm of the broader NPL market, the loans would need to be more evenly distributed across the provinces, less concentrated among a small handful of sellers, and originate more evenly across all major AMCs and commercial banks.

One Direction

That said, while the loans auctioned on Taobao sell at a premium relative to more established markets, prices changes on that platform is in line with both the broader direction of the primary and secondary markets. Unfortunately, measuring how Taobao’s NPL auction prices change relative to the rest of the market is difficult. There is no official indicator, and neither the banks nor AMCs consistently disclose data on what they earn from NPL sales. The only sporadic price indicators usually come from Chinese media reports.

Based on our data, the Taobao auction “price premium” is inconsistent, ranging between 13 and 25 percentage points (see Figure 5). However, for the few quarters for which we have data, the Taobao NPL prices move in the same direction as the primary market. Sources actively involved in the secondary market also say that directionally, the Taobao price curve aligns with their observations of price changes.

Figure 5. A Different Key—and A Different Range
Quarterly weighted discount on Taobao NPLs vs primary market prices disclosed by AMC executives
Notes: 1) Distressed debt investors measure the prices they pay as a percentage of the outstanding unpaid principal of the underlying loan. In other words, if a bank holds a 100 yuan bad loan and sells it at auction for 40 yuan, that would appear on the chart as 40%; 2) The horizontal lines are the average price AMCs paid for NPLs, as disclosed in local media by former top AMC executives. For 4Q2017, one of the executives gave a range; 3) While the Taobao data stretches back to mid-2015, prior to 2Q2017 the volume of auctions was too small to be a representative sample.
Sources: Taobao; MacroPolo.

What this seems to indicate is that while the Taobao NPL market has its own unique properties, the auction prices are nonetheless shaped by the same macroeconomic forces that determine the direction of prices in the broader market. Therefore, it’s worthwhile to provide a brief overview of the two main events that have shaped the NPL market over the past year and how they show up in the Taobao data.

1. The Collapse of the NPL Bubble

Toward the end of 2017, NPL prices were widely acknowledged as having entered bubble territory. That was driven by domestic investors leveraging their NPL acquisitions by borrowing from shadow banking institutions. However, the central bank and banking regulator pricked the bubble by cracking down on shadow banking. The result was that local investors no longer had the resources necessary to acquire bad loans in the same volume as they had the previous year. Prices started declining in 1Q2018 and have continued to do so ever since. Both the peak of the bubble and the subsequent price decline appear in the Taobao data.

2. Recognition of 90-day NPLs

In June 2018, the China Banking and Insurance Regulatory Commission imposed new rules requiring banks to recognize all loans overdue by more than 90 days as nonperforming. That brought China in line with international practice, but put banks under additional pressure to dispose of their NPLs. While NPL prices continued to decline in 2H2018, it’s not immediately obvious that the rule change was the proximate cause. However, the effect of the new rule was obvious in the Taobao auction data. In June 2018, the number of Taobao auctions spiked and has remain elevated ever since (see Figure 6). It also coincided with a spike in failed auctions, signaling that NPL prices had come under pressure as the volume of bad loans for sale increased (see Figure 7).

Figure 6. A Big Spike
Taobao NPL auctions spiked in June 2018 in response to regulatory changes

Figure 7. More Failures
More NPLs for auction in 2H2018 failed as prices came under pressure
Sources: Taobao; MacroPolo.

Conclusion

Taobao is a unique, discrete market for NPLs, and the data we’ve collected accounts for only a small part of China’s overall NPL market. However, it appears that Taobao NPL prices are sensitive to the same forces that shape the direction of the broader NPL market, making Taobao auction data a useful window.

The value of the dataset will rise if the volume of sales continues to rise, whether as a result of more aggressive NPL disposals across the board or as banks and AMCs more fully embrace Taobao. As the only publicly available source on NPL pricing in China, Taobao is a potentially invaluable resource. But to realize that potential the market may require more time to mature.

Acknowledgments

I’d like to thank Jeff Tharsen, a computational scientist at the University of Chicago’s Research Computing Center, for shepherding us through the early stages of this project. Without his insight, guidance, and technical expertise, this project would not have been possible. With a background in both classical Chinese and computer science, we could not have hoped for a better guide and more patient teacher.

I’d also like to thank my colleague Joy Dantong Ma, for both taking the time away from her usual work to support this data project and for demonstrating far superior Python abilities than she had led us to believe.

Endnotes