There is a statistical detail in the social integration data. The subtraction of the two periods' stock does not mean the increase in the current period, that is, the incremental stock ≠ increment. The common method to estimate the growth rate of social financing stock is to estimate the new social financing of different sub-items, and then add up to get the total amount of new social financing, and then calculate the year-on-year growth rate of stock. This means that directly using new quantity to estimate the growth rate of stock will introduce estimation error. This article aims to discuss the details of this error.
How big is the error? Historical experience shows that the annual error is about 1%, and the error cannot be ignored.
From 2065438 to 2008, the monthly δ stock and incremental error of social financing were between 3% and 25%. Looking back at history, there is always an error between the increment of social integration and the δ stock, and in most cases the former is greater than the latter. Comparing the results of the two calculation methods, errors can be observed: method 1, current increment/previous inventory *100%; Method 2, (current share-former share)/former share * 100%. The calculation probability of method 1 is higher than that of method 2, and the difference is about 1% (annual frequency). Method 2 calculates the real stock growth rate, so it can be seen that the stock growth rate is directly estimated by the newly added amount, and the annual error is about 1%.
Why are there mistakes? The error mainly comes from five subdivisions.
The breakdown of social financing shows that the difference between δ stock and increment mainly comes from foreign currency loans, entrusted loans, trust loans, corporate bond financing and other five sub-items.
The error of trust loan is mainly the annual 1 month stock adjustment. Taking the trust loans with the caliber of Trust Industry Association as the shadow index, it is found that the δ stock and incremental error of trust loans are almost all contributed every year 1 month, and the main error of 1 month is the pulse adjustment of the stock of trust loans financed by the society in that month. We guess that the adjustment in June 5438+ 10 may be related to the inventory statistics at the beginning of the year. The transparency of entrusted loan data is not high, and the reason for the error is unknown for the time being. But from 20 15, the error is almost negligible.
The error of foreign currency loan is caused by exchange rate fluctuation. The current average exchange rate is used for the increment calculation of foreign currency loans, and the ending exchange rate is used for the calculation of stock. Correcting the influence of exchange rate, that is, the increment and increment stock denominated in US dollars are obtained according to the corresponding exchange rate conversion, and it is found that the increment stock and increment error are basically smoothed out, which verifies that the error of foreign currency loans mainly comes from exchange rate fluctuations.
The reasons for corporate bond financing mistakes are different. Corporate bond financing with social financing caliber includes various bonds (at least 1 1). The annual net financing of 1 1 bonds with the caliber of Wonder has a high degree of fit with the incremental data of corporate bond financing with the caliber of social financing, and so does the δ stock. However, even with the same caliber, the new amount and δ stock of this 1 1 bond are different, because the value date is the standard of Wonder stock statistics, and the issue date is the standard of net financing statistics. Corporate bond financing with social financing caliber includes a variety of bonds, involving different issuance systems, redemption rules and balance statistics standards, not just the difference between the value date and the issue date. Therefore, its daily statistical monitoring involves a wide range, and there are naturally statistical difficulties, and there are inevitably statistical errors in the new amount and δ stock.
The other main reason lies in the specific background of insurance company compensation and financial deleveraging. The remaining income after deducting RMB loans from the total social financing is called "other items", which actually includes insurance company compensation, investment real estate and other financing. We think that in most cases since 20 16, other mistakes are mainly compensation from insurance companies; The error of 20 17 is not only related to the insurance company's compensation, but also related to the specific deleveraging background.
What is the guiding significance? The forecast result of incremental stock growth in 20 19 is more stable than in previous years.
By summing up the δ stock and incremental errors of the five sub-items, it is concluded that the estimated total error probability of the growth rate of the old caliber social welfare stock of 20 19 is only 1.04%, which is the neutral error level since 20 10. The large probability of δ stock and incremental error of the new caliber social financing stock is less than 0.9%, that is, the incremental estimation of the growth rate of social financing stock in 20 19 years is more stable than in previous years.
Core assumption risk: Sino-US trade friction exceeds expectations, and domestic demand declines beyond expectations.
catalogue
main body
First, the subtraction of social financing stock is not equal to the current increment.
1. 1 The subtraction of social financing stock does not represent the addition of current social financing.
The central bank also reports social financing stock and incremental data. However, the result obtained by subtracting the ending stock of two periods is not equal to the increment of social financing in the current period. Taking 20 18 as an example, in February 65438, the stock of new caliber social welfare was 200.75 trillion yuan, and in February 165438 10, the stock of new caliber social welfare was 199.29 trillion yuan, resulting in 14535 billion yuan. Compared with the new social financing1589.76 billion yuan reported by the central bank in February, the difference is1362.6 billion yuan. Similarly, in the month from 20 18 to 12, there is also a phenomenon that the δ stock is not equal to the new amount, and the error between the two sets of data is between 3% and 25% [18]. Looking back at the historical data, there has always been an error between the increment of social integration and the δ stock, and in most cases the error is positive, that is, the increment of social integration is greater than the δ stock.
1.2 The error in estimating the growth rate of social financing stock with new social financing cannot be ignored.
A common method to estimate the growth rate of social financing stock is to estimate the new social financing in T period (new social financing T), and then divide it by the previous social financing stock t- 1 to get the predicted growth rate of social financing stock in T period. However, δ stock is not equal to the increment of social financing in the current period, which means that the estimation method of calculating the year-on-year growth rate of social financing stock T with new social financing T will introduce errors. To what extent does this error affect the estimation accuracy? This is a problem that needs to be faced in estimating the growth rate of social financing stock.
In order to observe the error, we compare the calculation results of the two algorithms. Algorithm 1, current increment/previous inventory *100%; Algorithm 2, (current share-previous share)/previous share * 100%. Among them, the growth rate of social financing stock directly reported by the central bank is calculated by method 2, that is, the result of δ stock/stock is the real growth rate of social financing stock.
Figure 3 (only going back to the old caliber social integration) shows that the result obtained by algorithm 1 is basically higher than that obtained by algorithm 2, and the difference between the two results remains around 1% in most years; A few years exceed 1%, such as 2006 (1. 1%), 2007 (1. 1%), 2009 (1.9%) and 20. Individual years even exceed 3%, such as 2004 (3.4%), 2005 (4.8%) and 2008 (3.5%), and these years with great differences are mainly concentrated before 20 10. There are similar errors between the new caliber of 20 18 (adjusted caliber in September) and the caliber adjusted in July. Using method 1 and method 2, the difference (seasonal frequency) is less than 0.8%, as shown in Figure 4.
Second, the error mainly comes from five subdivision projects.
2. The error of1mainly comes from five subdivision terms.
Before explaining the main source of error, let's briefly review the frequency characteristics of social financing data released by the central bank. In terms of social integration of stock, the central bank began to publish stock data (including sub-items) directly only on 20 15, and the data frequency is seasonal. Monthly social and financial stock data (including sub-items) will be released in 20 16. From 2002 to 20 14, the central bank did not directly announce the scale of social financing stock (including sub-items), but announced the annual growth rate of social financing stock (including sub-items). Therefore, theoretically, we can use the social financial stock data from 20 15 and beyond and the year-on-year growth rate of the stock from 2002 to 20 14 to calculate the annual frequency social financial stock data of the corresponding years. As for the increment of social integration, in 20 12, the central bank published the monthly increment data of social integration, and in the same year, it supplemented the monthly increment data of social integration from 2002 to 20 1 1 year. So far, the monthly incremental data since 2002 1 has been published directly.
When analyzing the social integration data released by the central bank, two points are worth noting. First, the year-on-year growth rate announced by the central bank is only accurate to one decimal place (from 20 18, accurate to two decimal places), so the reverse calculation method of year-on-year growth rate will have the following situations-the earlier the year, the more accumulated errors caused by decimal point accuracy. Second, the new social organizations added every month can be divided into "early value of the month" and "early value of the month". The central bank publishes the preliminary statistical value of social financing scale (including sub-items) data in the form of news bulletin every month around 10- 15, which is the "initial value of the month", and the revised data around 16- 19 every month is the final "value of the month".
The reason why we spend some energy to sort out the social integration caliber and data frequency published by the central bank is to try to clarify which part of the error comes from the difference between δ stock and new amount, and which part is only introduced because of the conversion of data caliber. For example, before 20 15, there was a lack of social financing stock data directly released by the central bank. Although the annual social financing stock can be calculated according to the year-on-year growth rate of the social financing stock given by the central bank, the social financing stock published by the central bank is only accurate to one decimal place year-on-year, and every backward calculation will introduce an error, and with the extension of the backward calculation cycle, the more errors accumulated in previous years. Therefore, when we observe the incremental stock and incremental difference of annual frequency, we frame the data range at 20 10 and above. In addition, it should be emphasized that there is no directly published monthly frequency stock data before 20 15, and it is also impossible to calculate the growth rate of monthly frequency social financing stock. Therefore, when we discuss the monthly frequency δ stock and incremental error, we frame the data range at 20 16 and above.
Since 20 16, trust loans, entrusted loans, foreign currency loans and corporate bond financing are the main sources of δ stock and incremental errors of social financing, and the monthly errors are concentrated in three sub-items of foreign currency loans and corporate bond financing. The contribution rate of subdivision to the total social integration error depends on two factors, one is the error of subdivision itself, and the other is the proportion of subdivision relative to the total social integration. Although the error term of stock financing itself is large, its proportion in the total social financing is relatively low, so the final contribution of stock financing to the total social financing error is not high. Although RMB loans account for a high proportion of the total social financing, their own δ stock and incremental errors are small, so the δ stock and incremental errors of RMB projects do not contribute much to the total social financing errors. Since the regulation of bills, the financing error of off-balance-sheet bills has gradually converged, although its relative proportion is not low, but since 20 16, the contribution of bill items to the total error of social financing has declined rapidly. The following focuses on the analysis of trust loans, entrusted loans, foreign currency loans, corporate bond financing, and other five sub-items.
2.2 Trust loan error or main reason 1 monthly stock statistics
Comparing the trust loan data of trust industry association and social finance, the former stock is 97%-98% of the latter stock balance; In most cases, the annual increase of the former is basically 94%- 10 1% of the latter. It can be seen that the trust loans counted by the trust industry association (hereinafter referred to as the association's caliber) can be used as an excellent shadow indicator to observe the social financing caliber trust loans.
First of all, we find that the δ stock and incremental error of social financing caliber trust loans basically reach the annual peak in 65438+ 10 month every year, and the error of 65438+ 10 month can basically explain most of the annual errors. In other words, analyze the δ stock and incremental error of social financing caliber trust loans, and mainly explain the 1 month error.
Secondly, comparing the caliber of social finance with that of association, it is found that the trend of stock data is basically the same except for 1 month every year; There is no similar seasonal phenomenon in the increase. As a shadow indicator, the association's caliber points to the error between the annual 1 month δ stock and the increment of social financing trust loans, which is mainly caused by the change of social financing trust loan stock.
At present, it is impossible to effectively judge the final reason for the "pulse adjustment" of the annual 1 social financing caliber trust loan stock. We guess it may be related to the statistical arrangement of trust loan data at the beginning of each year. In any case, at least we can know the δ stock and incremental error of social financing trust loans, and it is important to observe 65438+ 10 month.
Due to the lack of transparency of entrusted loan data, it is not yet possible to know the reasons for δ stock and incremental error of entrusted loan projects. Since 20 15, the error of entrusted loan has greatly converged, and its contribution to the total error of social financing is almost zero. We will not make a detailed analysis of the entrusted loan error for the time being.
2.3 Foreign currency loans are due to exchange rate fluctuations.
Foreign currency loans are foreign currency loans provided by financial institutions to non-financial enterprises, individuals, institutions and groups in the form of loans, discounted bills, advances, negotiable bills and forfaiting [2]. It is worth noting that although it is in the form of "foreign currency", foreign currency loans are denominated in RMB in social and financial statistics, so exchange rate fluctuations will directly affect the scale of foreign currency loans denominated in RMB. When calculating the increment of social financing, the current average exchange rate is adopted for foreign currency loans; When calculating the social financing stock, foreign currency loans use the ending exchange rate [3], so exchange rate fluctuation is the main reason that affects the incremental stock and incremental difference. Correcting the influence of exchange rate, that is, converting the stock and new social financing into US dollars respectively according to the above exchange rate conversion principle, we find that the gap between δ stock and incremental data has been largely smoothed out (as shown in figure 15). The data of individual months are still slightly different, which is mainly related to the large fluctuation of exchange rate at the end of the period.
2.4 Corporate bond financing is due to statistical methods.
Corporate bond financing refers to all kinds of bonds issued by non-financial enterprises, including corporate bonds, ultra-short-term financing bonds, short-term financing bonds, medium-term bills, collective bills for small and medium-sized enterprises, non-public directional financing instruments, asset-backed bills, corporate bonds, convertible bonds, separable convertible bonds and private debt for small and medium-sized enterprises (including but not limited to the above eleven types of bonds). We compare the annual net financing amount and δ stock of 1 1 bonds [4] with Wonder's caliber, and find that the net financing amount of 1 1 bonds is basically the same as the new financing amount of corporate bonds with social financing caliber, at least the change trend is extremely consistent, as shown in Figure 16; Delta shares have a similar performance, as shown in figure 17. However, even with the same caliber, there are differences in the newly-increased net financing and δ stock of eleven types of securities. The main reason for the differences is that the stock statistics take the value date as the standard, and the net financing statistics take the issue date as the standard.
Corporate bond financing with social financing caliber includes a variety of bonds, involving different issuance systems, redemption rules and balance statistical standards, not limited to the differences in statistical standards between the issuance date and the value date. In other words, the daily statistical monitoring of corporate bond financing involves a wide range, which naturally has certain statistical difficulties, which is also the main reason for the difference between the new amount of corporate bond financing and the δ stock. Considering the complexity of corporate bond financing statistics based on social financing, and the fact that the new amount and δ stock gap of corporate bond financing based on social financing do not show obvious laws, we will not pursue the details of the reasons for the differences.
2.5 Other items are mainly due to the specific background of insurance companies' compensation and deleveraging.
Old-fashioned social financing statistics include RMB loans, foreign currency loans, entrusted loans, trust loans, undiscounted bank acceptance bills, corporate bond financing, domestic stock financing of non-financial enterprises, insurance company compensation, investment real estate and other financing * * * 10. The central bank usually publishes the new and stock data of the first seven categories, but does not publish the data of insurance company salary, investment real estate and other financing. The surplus after deducting the first seven items from the total social financing data is called "other items", which actually includes insurance company compensation, investment real estate and other financing.
Insurance company compensation refers to all kinds of funds provided by insurance companies to fulfill their compensation obligations within the validity period of insurance contracts, including property insurance compensation, health insurance compensation and accidental injury insurance compensation. This indicator will break even at the end of the year, and there is no concept of stock and balance. That is, other stocks = investment real estate stocks+other financing stocks, other new increments = insurance company compensation increment+investment real estate increment+other financing increments, other increments-δ other stocks = (investment real estate increment-δ investment real estate stocks)+(other financing increments-δ other financing stocks)+insurance company compensation increment.
There is no difference between the increment of investment real estate and δ investment real estate stock, and there is no difference between other financing increments and δ other financing stocks, so the increment of other items-δ other stocks is equal to the compensation increment of insurance companies. According to the data of the China Insurance Regulatory Commission, for most of the time since 20 16, the increment of other items-δ and the stock of other items are almost equal to the insurance company's compensation, as shown in Figure 2 1. In other words, many times, other differences mainly come from insurance company compensation. In a few time periods, such as 20 17, the increment-δ of other items and the stock of other items are quite different from the insurance company's compensation, and show significant seasonal fluctuations. Considering that investment real estate mainly refers to real estate held by financial institutions to earn rent or capital appreciation, or both, it is not prone to seasonal fluctuations. Therefore, we tend to think that the high probability of data difference of 20 16 does not come from investment real estate projects. Other financing means that the real economy obtains funds from small loan companies and loan companies, mainly including loans from small loan companies and loan companies, changes in financial system structure under the background of 20 17 financial interbank deleveraging, and small loan companies. Or fluctuate with it. In other words, we tend to think that the difference between other increments of 20 17 and δ shares is not only related to the insurance company's compensation, but also related to the specific financial deleveraging background.
Third, the significance of technical details to the prediction of the growth rate of social financing stock.
Foreign currency loans are increasingly becoming an important financing channel for foreign trade enterprises. Because it is denominated in foreign currency, the growth of foreign currency loans will be affected by the superposition of multiple factors such as the world economic situation, domestic and foreign spreads, and foreign exchange management policies. It is difficult to accurately capture the error disturbance caused by the incremental estimation of foreign currency loans. Similarly, corporate bond financing involves different bond types and statistical standards, while other projects include three sub-projects, and entrusted and trust loans involve non-standard projects, so the information is not completely transparent, so it is difficult to accurately capture the errors of the five sub-projects. We resort to the historical comparison method to roughly estimate the δ stock and incremental error of five terms in 20 19 years.
Looking back, from 20 10, the errors of other projects are basically stable in the range of 0.3%~0.5%. We tend to think that the error probability of other projects of 20 19 is 0.3%~0.5%, which belongs to neutral level. In corporate bond financing, except for 20 18 (0.4%), the error since 20 10 rarely exceeds the range of -0.2%~0.2%. We tend to think that the error probability of 20 19 corporate bonds is -0.2%~0.4%, which is a high level in recent years. In terms of foreign currency loans, the error has been basically controlled at -0.2%~0.45% since 2009. With the unilateral depreciation trend of RMB since 200 15, the error of foreign currency loans has also turned from positive to negative. Taking 20 17 as the reference sample under the assumption of RMB exchange rate appreciation and 20 15 as the reference sample under the assumption of RMB exchange rate depreciation, it is estimated that the error of 20 19 foreign currency loans is within the range of-0.16% to 0.1%,from 20. In terms of trust and entrusted loans, the incremental stock difference of trust loans in June 5438+ 10 was 58.2 billion yuan, exceeding the level of the same period in the past two years, and so was entrusted loans. Considering that the error of trust loans is mainly contributed by June 5438+ 10, it means that the probability of introducing errors into trust loans and entrusted loans in 20 19 years is higher than that in 20 17 years and 20 18 years. According to the error number of 65438+ 10 in recent two years, linear extrapolation shows that the subdivision of entrusted and trust loans in 20 19 may introduce an error of 0.02%~0.04%, which is a historical high. When the five sub-errors add up, the probability of the final error is only 1.04%, which belongs to the neutral small level since 20 10.
It is worth noting that the central bank did not publish the details of write-off and ABS in June-February of 20 19. According to the experience in the past two years, there are no mistakes in the stock and increment, ABS and write-off of local special bonds. Then the incremental stock and incremental error of the new caliber mainly come from five sub-items: foreign currency loans, trust loans, entrusted loans, corporate bond financing and other items, and the contribution of these five risks to the incremental stock and incremental error of the new caliber head office is less than 0.9%.
Risk warning: Sino-US trade friction exceeded expectations, and domestic demand fell more than expected.
[1] The error calculation formula is: (increment-δ stock)/δ stock.
[2] Sheng Songcheng et al., Theory and Practice of Social Financing Scale (Third Edition), China Financial Publishing House.
[3] Sheng Songcheng et al., Theory and Practice of Social Financing Scale (Third Edition), China Financial Publishing House.
[4] Check 8 items of corporate bonds, corporate bonds, medium-term notes, short-term financing, PPN, ABN, convertible bonds and separable convertible bonds.
(Article Source: Guo Lei Macro Teahouse)