China Real Estate Developers — What were Their Common Financial Tricks

PHOTO: QILAI SHEN/BLOOMBERG NEWS

Background

This project is to explore China real estate developers’ financial data and uncover what were the most common tricks they leveraged to cover up their high debt loads before the Three Red Line Policy was enforced by the central government in the Aug of 2020. Due to the rapid increase in borrowing by real estate developers to compete for land resources, the real estate industry has become an extremely high-indebted industry with a scary leverage ratio over the past decade. When the economy was running at the prime time with an easy money policy and high GDP growth rate, real estate developers enjoyed economic prosperity and could obtain external financing at low-interest costs, and conduct refinancing to extend their payback period. However, particularly during the period from 2019 to 2020, China's economy was hit by the trade war, and then further crippled by the COVID-19 pandemic. Even though the China economy had been through a lot of ups and downs during that period, the real estate sector was still playing a significant role in promoting China's economy, one of many reasons is the real estate sector has many close ties to other sectors such as banking, commodity (e.g., Steel & Concrete), manufacturing (e.g., home appliance) and transportation. The development of the real estate sector can cause a ripple to other industries. As you can see from Figure 1., the real estate sector still accounts for 25–30% of China's GDP. That is to say that if you want to understand the China economy, the real estate sector is definitely a good starting point to work on.

Figure 1. Real Estate Sector Contribution to China GDP

Therefore, in order to contain the spread of financial risk due to high leverage and high debt problem that surfaced under the current low GDP growth rate environment, Beijing decided to enforce the Three Red Lines policy to contain the financial risk. Here is the summary of the policy framework:

1. 70% ceiling on liabilities to assets, excluding advance proceeds from projects sold on contract
2. 100% cap on net debt to equity,
3. Cash to short-term borrowing ratio must be at least one.

Developers will be categorized based on how many limits they breach and their debt growth will be capped accordingly. If a firm passes all three criteria, it can increase its debt by a maximum of 15% in the next year. However, developers will be restricted to the access of financing if they breach any one of these three limits.

In the near term, a developer with a weak balance sheet and sizable exposure to second-tier cities may need to cut home prices to boost sales and shore up a cash position. This was self-evident in Evergrande’s campaign last year to offer discounts of as much as 30% — its deepest cuts ever. It may also spur waves of equity sales and spinoffs of non-core businesses such as property management services; Evergrande is already disposing of assets and spinning off affiliates.

Project Goal

What I am going to do is to use the past financial data of 116 property developers in China to extract insights such as the most common tricks they leveraged to cover up their high debt loads to shareholders and also give you some heads-up about what is a better way to analyze their financial data besides calculating those financial ratios. Having a good command of financial ratio analysis definitely can help an analyst go further and deliver actionable and better insights. However, in reality, their accounting practices seem much more complicated and convoluted. That is the reason why a deep dive into a given industry is definitely necessary for being a good financial data analyst.

Data Description

The data contains three-year financial data (2017, 2018, and 2019) of 116 listed real estate developers and comes from the Wind Financial Platform. All accounting items are in the unit of RMB 100 million.

Sec_Code: the security code of a listed real estate developer.
Sec_Name: the company name
Cash_From_Customers: how much cash proceeds have been received from customers. This is a key indicator to evaluate a developer’s sales & marketing capability to drive sales. To give you a heads-up. In China, some customers buy a new property that is still in construction and will be completed within 1 or 2 years. Typically speaking, another important alternative indicator called contract sales would also be used to analyze cash received from customers, but this dataset does not have it.
NetCF_from_Operating: cash flow from net operating activity
NetCF_fromInvesting: cash flow received from investing activity Proceeds_FromBorrowing: how much cash received from borrowers Proceeds_FromBondIssues: how much cash was received from issuing bonds Dividend&InterestPaid: how much dividend a real estate developer has distributed in a fiscal year
NetCF_from_Financing: how much cash received from financing activities
Provision_InventoryDepreciation: provision charged on the unsold property. This is an interesting indicator. Typically if a newly built property cannot be sold, it will be counted as an inventory unit on the accounting book and depreciation rules will be applied. A good real estate developer should have a high turnover to sell newly built properties as soon as possible to avoid such non-cash charges.
Operating_Revenue: how much revenues generated through core business activity
Bills_Payable: any bills that are payable to banks or other financial institutions. Usually, bills payable are short-term debt
A/C_Payable: short term debt to suppliers
Total_CI: total comprehensive income
CI_For_Non_Controlling_Ints: comprehensive income for non-controlling interests
NonControling_Interests: non-controlling interests from minority shareholders in total equity
Total_Equity: total equity on the shareholders’ equity statement

Data Preprocessing

Step 1. Translation
You can see from Figure 2, as dataset columns were in Chinese, the first step was to translate the columns and security names into English.

Figure 2. Dataset Overview

I prepared a list of English account line items and security names that would be used to replace columns in the dataset. This process is kind of manual and tedious. At this point, I can’t see there is any better alternative to automate the process. We may also use BeautifulSoup to scrape the data from Yahoo Finance website and retrieve English security names from Yahoo, which was exactly how I retrieved the English name list. However, as the Yahoo Finance HTML structure would change from time to time, it would require more effort to update and maintain Python scripts.

Figure 3. English Accounting Items

Figure 4. English Security Name

Data Exploration & Missing Value Handling

In this project, we are not going to use all columns but just keep our focus on accounting variables that matter the most to our project goal. Candidate variables include ‘A/C_Payable’, ‘Bills_Payable’,
‘CI_For_Non_Controlling_Ints’, ‘Cash_From_Customers’, ‘NetCF_fromInvesting’, ‘NetCF_from_Financing’,
‘NetCF_from_Operating’, ‘NonControling_Interests’,
‘Operating_Revenue’, ‘Provision_InventoryDepreciation’,
‘Total_CI’, ‘Total_Equity’.

1. Data Distribution
   There are a few key takeaways we can get from the distribution charts. For example, in 2019, the distribution of these accounting variables is either higher right-skewed (e.g., total equity) or left-skewed (cash flow from investing). This makes sense as we are aggregating these 116 developers’ accounting information to be an industry level or market level. In the real estate sector, the “Winner Takes All” model does exist. If a developer has a strong financial position, it can easily raise funds from external parties and aggressively bid for valuable land resources from the local government. Also with the strong financial position, real estate developers can easily build up close ties with local government to negotiate. All these factors reinforce the “Winner Takes All” model in China's real estate sector.

Figure 5. Distribution of Accounting Variables

2. Fill in missing values
There are some key variables from which we can see some `Nan` value in the dataframe. Fortunately, we don’t have a lot of missing values in these columns. Also as these columns are fundamental financial data, they are to some extent correlated to each other.

With a basic understanding of the accounting data, one of the best ways is to fill in missing values with the normalized ratio method. For example, if A/C payable is highly correlated with Operating Revenue (In fact, it is because A/C payables are positively related to sales. If more sales activities occurred, A/C payables will also increase but not necessarily at the same pace but should show the same trend). If we want to fill in the missing values in 2019, we can compute the ratio of A/C Payable to Operating Revenues in 2018 and then apply the same ratio in 2019 by multiplying it with its Operating Revenues of 2019 to estimate the A/C payables of 2019.

One of the reasons I believe SimpleLiinearRegression does not work to help us to predict the missing values here is first of all our dataset is relatively small, we only have 3-year data for 116 developers (in total, only 348 data points). Secondly, If we apply Linear Regression, it is inevitable we have to build up our linear regression either across all developers’ data in the same year or the same developer with three-year data. Either way has high multi-collinearity, which violates the underlying assumption of linear regression. For example, companies in the same industry definitely will be impacted by the same policy. Also, they have competing relationships (e.g., one company prospers, the other might decline) in the market as well.

We can also increase the robustness of this ratio, by taking the average of the other two years’ data as well. For example, we can take the average of both 2017 and 2018 and apply the ratio method to fill in missing values in 2019.

Goals of Analyzing Real Estate Developers’ Financial Well-being (2017–2019)

Analyzing this specific sector requires a good understanding of the industry insights and how to tweak accounting information to turn it into a signal that reflects the real accounting performance. But in this article, I am not going to touch on complicated accounting manipulation techniques but just to understand the entire industry’s situation and what we can infer/learn from the real-estate sector during this period. In the following session, we are going to address three important questions:

1. How was the entire real estate market changing over the three-year period?
2. What kind of common “financial engineering” were developers leveraging to polish up their real financial position?
3. Can we create a model that can help us to pick up the best and worst company, and long the best and short the worst to see how this pseudo portfolio performs in one year period (from 2020–2021)?

Q1. Market Concentration Analysis — How was the entire real estate market changing over the three-year period.

For the first level of market concentration analysis, we want to identify how many top players would capture at least 80% of total sales/cash flow from customers (aka. Pareto Analysis). As in this dataset, it does not provide sale or revenue data and thus we will use cash flow from customers as an alternative variable.

Cash flow received from customers is slightly different than revenues/sales in terms of the accounting recognition period. Typically, some housing developers in China heavily rely on prepayments from customers, in which case customers pay in advance to buy a property that will be completed within the agreed period (approximately in two to three years). This prepayment will be treated as liabilities on their financial statements and will be recognized as sales/revenues when housing developers deliver units to customers. There are some reasons supporting this presale practice. First and foremost the housing market is booming in the market stage called “Seller Market”. Developers can leverage pre-sale arrangements to test market demands and formulate good pricing strategies. The second reason is presale can act as an effective and zero-cost hedging tool compared with common derivative contracts. Property developers can lock in the price of a property to avert policy risk (pricing control policy enforced by local government or central government). The presale practice, on the other hand, can also assist real estate developers to raise funds with zero interest costs. However, does more cash flow received from customers indicate a good signal from a business perspective? It really depends but in some cases investors need to be more cautious because property developers might be cash-strapped and therefore would discount property prices or provide other perks such as free parking lots or lower management fees in order to get cash from customers (customers are not stupid. If there are no benefits, why would they freeze a large sum of cash for around 2–3 years before the completion of property). Also, if the speed of cash received from customers does not align with the number of units deliver to customers, there might be a risk of under-delivery. Heavy discounts would appear when real estate developers’ financial well-being is under the water. For example, before the burst of China Evergrande Group’s debt problem, the company has applied deep discounts on its apartment sales in order to collect cash to clear debts ASAP.

As you can see, the market landscape did change to some extent. You see that the blue curve (2017) from Figure 6 is more left outward than the orange and green curves. Also, the blue curve is reaching 80% of total cash received from customers much earlier than the other curves. We can imply that over the three-year period, the market is becoming more consolidated by those top market players, given that the real estate market was off the high and entering a downward trend at the time. Even the market was becoming more concentrated, the total cash received from customers still edged up year over year. In detail, total cash received from customers logged a positive growth rate, at 20% from 2017 to 2018. However, in 2019 the growth rate was shrinking to only 9.7%. What was really happening at the time from 2018 to 2019? Also, in 2017 the top 20 market plays siphoned over 80% cash from customers. When it comes to 2019, the number was down to the top 18. The declining trend of the number indicates that at the time the market was going through some struggles and becoming more centralized on the top players, which have a strong market position and good financial well-being.

Figure 6. Market Concentration of Cashflow from Customers (2017–2019)

Also, you can clearly observe that over the three-year period (Figure 7), firms at higher bins (Bin 5 with high equity and stronger financial position) shows a much lower standard deviation of market ranking in terms of cash received from customers, which indicates that their market position is relatively solid compared with those at the lower bins. This further indicates that the market during the three-year period was going through consolidation, in which market shares were taken away by strong market players.

Figure 7. Standard Deviation of Market Rank By Total Equity Bins

Q2. What kind of common “financial engineering” developers leveraged to polish up their real financial position.

Tricks from A/C payables & Bill payables

• A/C payables & Bill payables are short-term financings that real estate developers would leverage. In China real estate market, developers usually have strong negotiation power to put the payment on hold to suppliers until properties have been completed. Therefore when the liquidity becomes tight, developers tend to defer payment to supplies. Therefore, payables are an important source of short-term financing for developers to relieve the short-term liquidity problems. By increasing payables, developers were transferring some of the liquidity pressure to suppliers.
• In Figure 8. total payables logged a positive growth rate during the three-year period. It increased by 22% from 2017 to 2018, and then the rate was accelerating to 31% from 2018 to 2019. However, merely looking at total payables can’t extract any useful insights. We also need to analyze cash received from customers and operating revenues.

Figure 8. Total Payables

• In Figure 9, the growth rate of operating revenues was constant throughout the three-year period. However, we need to note that operating revenues actually reflect how much cash received from customers were finally recognized as revenues on financial statements. Therefore, operating revenues might be a lagging data point because it recognized sales activities that occurred before the fiscal year. For example, if a real estate developer received a downpayment from customers to buy a property that will be completed with an agreed period, cash received would not be recognized as revenues until properties are finally handed over to customers. In this case, we can also look at cash received from customers to evaluate sale activities.

Figure 9. Operating Revenues

• When it comes to cash received from customers, we can see that in 2019, the growth rate of cash from customers did not align with the growth rate of total payables for the entire market. That raises a red flag because if the trend persists in the future, real estate developers might not be able to use cash to cover total payables, even though developers can negotiate better credit terms but in the end developers have to clear payables.

Figure 10. Cash Received From Customers

• Also, if we look at the line plots for operating revenue against net cash received from operating activity, we can see the gap between the two data is widening. Operating revenues of the industry were increasing while net cash flow received from customers was leveled off. Though this signal does not necessarily mean operating revenues were manipulated, it’s a good signal showing operating revenues alone cannot tell the whole picture of a firm’s financial well-being.

Figure 11. Operating Revenues vs. Net Cash Flow From Operating

Off-balance sheet item — Joint Venture

• The other common technique that developers leverages as mentioned is to form a joint venture with institutional or private investors, and then developers sell their shares of the joint venture to investors and guaranteed that they would buy back at a higher price later. Therefore, how can we look for the signal through accounting data? Actually, it could be complicated as typically this type of joint venture does not require full disclosure as developers usually hold less than 50% of the stake. However, we could use a relatively indirect approach by analyzing the minority interest on their financial statements.
• What we are going to do is to analyze the ratio of Non-Controlling Interest over Total equity vs. the ratio of comprehensive income attributed to NCI over total comprehensive income. Logically speaking if non-controlling interests are normal and belong to equity investment (100%), these two ratios should be very closed and aligned. However, if this non-controlling interest entity involves debt-form investment, it would complicate the situation and we can see the two ratios would deviate from each other.
• In order to better analyze data, we classify data into five quintiles. The higher quantile represents stronger total equity and so on. As you can see from Figure 12, bin 4 & 5 (high total equity), the ratio difference is relatively stable and much smaller than bin 1, 2, and 3. In addition, you can see that bin 5 was trending down from 2017 to 2019, while other bins were trending up. That makes sense as high total equity means a stronger balance sheet, which enables developers to have broader access to external financings such as bond issues and favorable credit terms in paying total payables, while the weaker players have more restricted access to external financing and also higher cost of borrowing. Therefore they have to do some “financial engineering” to source money through uncommon ways to either support their competition or operating activities.

Figure 12. Non-controlling Interest % in Total Equity vs. Non-controlling interest % in Comprehensive Income by Five Bins

• As expected (Figure 13), you can see that developers with stronger balance sheets and equity show that they can have greater access to external financing. Over the three-year period, companies at the higher bins had raised many orders of magnitude of money through bond issues or borrowing compared with those companies at the lower bins (lower equity and smaller balance sheet size)

Figure 13. Total Proceeds Raised through Debt Issues or Borrowing

• This bar chart (Figure 14) has further substantiated my argument that a company with higher total equity has greater power to negotiate for better terms of total payables with suppliers or financial institutions. Therefore, you can see that bin 5 (high total equity) shows much larger total payables than the lower bins over the three-year period.

Figure 14. Total Payables by Bins (Total Equity)

Q3. Can we create a model that can help us to pick up the best 5 and worst 5 companies, and long top 5 and short worst 5 to see how this pseudo portfolio performs in the next 6–9 month returns (from 2020/3/31–2020/12/31)?

Before we pick stocks from these 116 listed firms, we need to define our criteria for stock selection. Here are some of the criteria that I addressed in previous sessions. You know that I was doing this part just for fun, if you think some factors look unreasonable, it is totally acceptable to have differing views as the best strategy does not exist in the stock market :).

Factor 1: market ranking based on cash received from customers and operating revenues should be relatively close to each other. If these two ranks are relatively stable and aligned, it signals that the earning quality of such developers is credible.

Factor 2: constant dividend & interest payout to shareholders & bondholders, the higher the payout the better company would be. This ratio represents the rate of return that bond and equity investors can earn either through bond or equity investments.

Factor 3: ratio of non-controlling interest to total equity and the ratio of total comprehensive income attributed to minority shareholders should be relatively close or aligned to each other. The difference between the two ratios should be constant otherwise it indicates that the company is likely to use joint venture financing.

Factor 4: ratio of cash received from customers to total payables should be higher, the higher the ratio the better the company can use the cash received to cover its short-term liabilities.

In order to avoid looking-ahead bias, we decided to trade stocks after the first quarter of 2020 because most of the firms’ 2019 annual statements should be available for the market. Also, our holding period will start from 2020/3/31 to 2020/12/31.

As you can see from our portfolio performance (Figure 15), the stock-picking methodology is really risky and volatile. The portfolio performance increases by almost 20% within three months and falls to square one at the year-end of 2020. At the same time, the trading idea ignores a number of important factors. Firstly, we cannot conduct short selling in the China stock market. Second, I did not include transaction costs, dividend or capital gain taxes, and dividend distribution which will impact portfolio performance. Lastly, we did not consider the possibility of executing transactions during a trading day but just at the adjusted closing price. At the same time, we know that during this period, China's economy was adversely impacted by the outbreak of COVID-19. Considering the extremely downbeat economic outlook, all industry sectors will be negatively impacted. In summary, these four factors are definitely not robust enough over time and cannot serve as good profit-generating factors at all.

Figure 15. Trading Strategy Performance

Takeaways

With the analysis of China real estate markets, we can get a few takeaways:

1. As the economy was trending down (China was inflicted by the trade war and then COVID-19, the winner will be much stronger with greater access to external financing to bid land resources to maintain market positions.
2. Payables are a common tool that developers can negotiate in order to improve their financial well-being in the short term.
3. We can infer joint venture financing through the ratios between comprehensive income attributed to non-controlling interests and non-controlling interest to total equity. These two ratios’ trends should be aligned if no joint venture financing is involved. Companies with strong balance sheets typically would not opt for such financing sources as this type of joint venture financing has an extremely higher cost of borrowing, but weaker players tend to do so.
4. Even though, we know which signals are good or bad, it is still very hard to leverage those signals to develop a profitable trading model even in paper trading. If a trading idea cannot pass paper trading, just don’t even think about the real trading environment.
5. Obviously, our trading ideas do not include any rebalancing techniques. Given the market situation, our trade model can be improved through rebalancing. For example, we can rebalance our portfolio based on the latest financial statement, such as selling stocks that fall out of Quantile Bin 1 and buying stocks back if they get out of Quantile bin 5 and so on.

Acknowledgment

If you want to check out how I execute python scripts to conduct analysis, please feel free to check out my GitHub repository. In the end, I want to give very special thanks to my wife, who relentlessly supports our family while pursuing her Ph.D. She is a real hero in my heart. Also, a big thank you to Logan and Allen at Rhodium Group. Their research titled “Time to Pay the Piper” is really inspiring to me and gives me a lot of inspiration in writing this article. Please leave me some comments if you have any. Thank you so much for reading.