I. The Current Theoretical Framework and Empirical Analysis of the Interest Rate Market

(A) Theoretical Framework

In the context of insufficient aggregate demand, the classic IS-LM framework can be used to explain current interest rates (using the yield on ten-year government bonds as an indicator of market interest rates). The IS curve reflects the output at different interest rates when the goods market is in equilibrium, with the core assumption being that, ceteris paribus, there is a negative correlation between interest rates and investment, meaning lower interest rates lead to stronger investment and higher output. The LM curve reflects the equilibrium in the money market, showing the interest rate levels corresponding to different outputs given a fixed real money supply. Together, they form the classic IS-LM curve.

The current downward trend in interest rates began in 2021 and accelerated in 2023. The year 2021 also marked a period of concentrated exposure of real estate risks in China, with investment declining at the same level of interest rates. Ceteris paribus, the IS curve shifted to the lower left. Concurrently, the central bank increased net money supply to offset the impact of declining investment, causing the LM curve to shift to the lower right. Based on this framework, the downward trend in interest rates is also a logical outcome.

Advertisement

(B) Further Analysis of the Downward Trend in Treasury Yields in 2023

Before 2021, the yield on ten-year government bonds fluctuated with both increases and decreases, oscillating around 3.5%, with 2.7% being a bottom for long-term bond yields. During 2014-2016, there was a prolonged downward trend. This was mainly driven by two factors. On one hand, this period was a time of active advancement in China's interest rate marketization reform, and the bond market was undergoing a process of repricing. On the other hand, this period also coincided with a downcycle in China's real estate market, with economic growth continuously slowing down, significant downward pressure on prices, and PPI experiencing 54 months of negative growth. Monetary policy saw five interest rate cuts and six reserve requirement ratio reductions, leading to a decrease of about 200 basis points in long-term bond yields. Subsequently, with the advancement of supply-side structural reforms and the monetization of shantytown renovation reforms, yields began to rebound. At the beginning of 2020 during the pandemic, yields temporarily fell to around 2.5%, but rebounded quickly after lockdowns were lifted.

After 2021, the fundamentals of the ten-year government bond yield shifted to a long-term downward trend, with two rare rapid declines. The first occurred from March 2021 to September 2021, dropping from 3.26% to 2.83%, a cumulative decrease of 43 basis points. This rapid decline in yields was mainly due to the market's strengthened expectations of central bank interest rate cuts.

The second occurred from November 2023 to February 2024, falling from 2.7% to 2.33%, a cumulative decrease of 37 basis points. This downward trend in long-term bond yields was largely due to a shift in inflation expectations. Inflation expectations are a composite of several factors, including economic growth, capital return rates, risk preferences, and expectations for future macroeconomic policies. In 2023, the real estate industry continued to adjust, and policies to control local debt risks also led to restricted financing for local financing platforms. The market experienced an asset shortage. With the real estate sector dragging down the economy, macroeconomic growth further slowed, and the characteristics of insufficient aggregate demand became more pronounced. With insufficient aggregate demand and declining capital return rates, macroeconomic policies maintained sufficient stability. Under these circumstances, related price indices began to adjust downward, especially with CPI and PPI experiencing negative growth. During the previous downward price cycle, PPI had a continuous negative growth for 54 months. The current situation is more severe than the previous cycle, and with the ongoing negative growth of PPI and the sluggish CPI, inflation expectations have changed, which is also a result of changed growth expectations for the coming period.

II. Model Analysis of the Trend of Ten-Year Government BondsWhen the long-term bond yield drops to around 2.3%, simple qualitative analysis is insufficient to infer the future trend of yield rates. To address this, we construct a Vector Autoregression (VAR) model and set the yield trends under different scenarios.

(1) Model and Variable Selection

We select the Vector Autoregression (VAR) model as our analytical tool. This model can capture the dynamic impacts between variables and is one of the commonly used analytical tools in modern economics. More importantly, the constructed VAR model can provide out-of-sample forecasts in the form of variable autoregression without setting specific scenarios.

Based on the aforementioned analysis, macroeconomic trends, asset returns, and inflation expectations are the main factors affecting the ten-year long-term bond yield. Additionally, according to the classic IS-LM model, liquidity also has an impact on interest rates. On the basis of theoretical analysis and through quantitative testing, we select nominal GDP growth rate, Wind All A (excluding finance, oil, and petrochemicals) net asset return rate, deflator index, and M2-social financing year-on-year growth rate as explanatory variables, representing growth, quality, price, and liquidity indicators, respectively. The dependent variable is the ten-year government bond yield.

In terms of sample selection, 2013 marks the beginning of China's interest rate marketization reform and the start of the continuous improvement of the yield curve. We take 2013 as the starting point and June 2024 as the sample interval for analysis.

We construct the model with a monthly frequency. For variables with higher original frequencies (such as government bond yields), we take the monthly average of daily data; for variables with lower original frequencies (such as ROE of listed companies), we use interpolation to fill in quarterly data. Considering that time series models require each variable to pass the stationarity test, we perform first-order differencing on the adjusted monthly variables.

To better verify the model's effectiveness, we use data from January 2013 to March 2023 as the training set, retaining the most recent 12 monthly data as the test set; from April 2023 to March 2024, we use rolling modeling and forecasting to verify the effectiveness of the fitting results.

(2) Data Description

In theory, there is a positive correlation between economic growth, net asset return rate, price deflator index, and government bond yields. The higher the economic growth, the more overheated the economy, requiring a higher equilibrium interest rate level; the higher the net asset return rate, the higher the growth quality, which can also match higher yields; the price deflator index is the most direct reflection of inflation indicators, and the higher the inflation level, the correspondingly higher the yield. The M2-social financing represents the remaining fund supply, and the higher this number, the more abundant the remaining fund supply, which means the lower the yield. In other words, theoretically, this indicator should have a negative correlation with yields.

From the actual data, consistent with the aforementioned theoretical analysis, the ten-year government bond yield shows a clear positive correlation with growth, returns, and prices, while M2-social financing shows a clear negative correlation with yields. Further testing of the relevant indicators reveals that there is cointegration between the dependent and explanatory variables, which can be analyzed using the VAR model.(III) Model Forecasting and Out-of-Sample Extrapolation Results

We used the most recent 12 months of data as the test set. Based on the training set, we employed the VAR model to model historical data on a quarterly basis and forecasted the Treasury bond maturity yield variable for one quarter ahead. Considering the stationarity of the data, the yield was differenced and used as the explained variable. Under the first-order differencing condition, the trend of the in-sample indicators was basically consistent with the actual indicators, and out-of-sample forecasts had the same sign as the actual indicators in 10 out of 12 test set samples. After restoring the first-order differenced data to the long-term interest rate variable, the model's out-of-sample forecast values maintained a relatively consistent trend with the 10-year Treasury bond maturity yield.

The characteristic of the VAR model is that it can use the model to roll forward predictions based on existing data without specifying scenarios. Based on the model's forecasting results, we obtained data from July to the end of the year. The results show that after July, the 10-year Treasury bond yield will rebound from the current level and return to a range around 2.35%. This indicates that, barring significant changes such as risk events similar to the Yongmei default that lead to short-term major adjustments in central bank monetary policy, or the introduction of large-scale economic stimulus plans that drive yields to rise quickly, the 10-year long-term debt will rise slightly and remain stable at the current level. It is worth noting that the nominal GDP growth rate will experience a significant decline in the coming period. Given the endogeneity of the model data, we believe that if there is no sustained effort from fiscal policies such as medium and long-term Treasury bonds in the third quarter, it is possible that the nominal growth rate will fall below 4% in the fourth quarter.

(IV) Yield Trend Under Different Assumption Scenarios

We fitted the VAR(3) model parameters based on disclosed data, assuming that the quality indicator (A-share ROE indicator) meets the consensus expectations of analysts and reaches approximately 8.55 by the end of 2024; assuming that the liquidity indicator (M2 - social financing year-on-year growth rate) remains unchanged; and conducting scenario analysis on the forecast of the 10-year Treasury bond maturity yield around the changes in growth and price indicators.

In the second half of 2024, we provided a test range of 4.5-6.0% for the quarter-on-quarter GDP growth rate and a test range of -0.5-0% for the year-on-year change in the GDP deflator, with the following results:

- When the GDP grows by 4.5% year-on-year and the GDP deflator changes by -0.5% year-on-year, the 10-year Treasury bond maturity yield is 2.235%;

- When the GDP grows by 6.0% year-on-year and the GDP deflator changes by 0.0% year-on-year, the 10-year Treasury bond maturity yield is 2.233%;

- When the GDP grows by 4.5% year-on-year and the GDP deflator changes by 0.0% year-on-year, the 10-year Treasury bond maturity yield is 2.229%;

- When the GDP grows by 6.0% year-on-year and the GDP deflator changes by -0.5% year-on-year, the 10-year Treasury bond maturity yield is 2.227%.Overall, the impact of nominal growth rate and changes in the deflator on yields is not significant. With nominal growth rates of 6% and 4.5%, and deflator rates of 0% and -0.5%, there are a total of four combinations, with yields fluctuating between 2.22% and 2.23%. Different scenarios suggest that over the next two quarters, the yield on the 10-year government bond will rebound from the current level.

III. Reasons for Central Bank Intervention in the Government Bond Market and Possible Effectiveness

(1) Reasons for Central Bank Intervention

After the bond market experienced a continuous downward trend since 2023, starting from April 2024, the central bank has been continuously reminding the bond market of risks. The first quarter's regular meeting had stated, "In the process of economic recovery, attention should also be paid to changes in long-term yields." Subsequently, the central bank issued warnings about bond market interest rates on multiple occasions. In the five trading days following April 24, the 10-year government bond yield rose from 2.22% on April 23 to 2.34%, an increase of 12 basis points; the 30-year government bond yield rose from 2.44% on April 23 to 2.55%, an increase of 11 basis points. Considering the longer duration of the 30-year bond, its actual price adjustment will also be larger. However, after a month, long-term bond rates continued to decline. The central bank issued another warning, with the central bank governor Pan Gongsheng clearly stating: "At present, especially attention should be paid to the term mismatch and interest rate risk of a large number of non-bank entities holding medium and long-term bonds, maintaining a normally upward sloping yield curve, and maintaining the market's positive incentive effect on investment." On July 1, the central bank announced in the open market operations notice, "It is decided to carry out government bond borrowing operations for some primary dealers in open market operations in the near future."

The central bank's emphasis on long-term government bond yields is not for the purpose of stabilizing the exchange rate as the market believes, but for the stability of the banking system.

In terms of exchange rates, taking the offshore market, which better reflects market demand, as an example. In this economic cycle, the renminbi exchange rate against the US dollar has broken through 7.1 four times. At these four points, the level of government bond yields was not consistent, indicating that there is no direct correlation between government bond yields and the renminbi exchange rate. At the same time, at the first three points where it broke through 7.1, the central bank did not intervene in the exchange rate by intervening in government bond yields. More importantly, when the central bank began to verbally intervene in yields, the exchange rate did not show a strong trend of depreciation.

The so-called stability of the banking system is mainly due to two factors. First, the rapid increase in the bond positions held by commercial banks has formed a certain stampede effect to a certain extent, which may lead to over-adjustment of yields in the short term. Once the yields return to the long-term equilibrium level, it is very likely to bring losses to banks, and some city commercial banks with larger positions may even face certain risks. Since 2023, the scale of government bonds held by commercial banks has soared from 15.8 trillion to 20.5 trillion, an increase of 4.8 trillion, and the proportion of positions held has also exceeded 70% of all government bonds. Second, looking at a long period of time, there is a strong correlation between government bond yields and bank net interest spreads. The current net interest spread of banks has dropped to 1.54%, which is already lower than the previously recognized break-even point of 1.7% for banks. If government bond yields continue to decline, it may lead to further contraction of the net interest spread, leading to risks for some small and medium-sized city commercial banks with weak risk resistance.

(2) Effectiveness of Central Bank Intervention

From the international experience, central banks can better control the yield of short-term interest rates by adjusting policy interest rates and liquidity control, but for long-term interest rates, the effectiveness of central bank intervention is uncertain for a period of time. For example, in March 2004, the Federal Reserve began to raise interest rates, and the interest rate increased by 400 basis points in two years, but the yield on the 10-year government bond did not fluctuate much, and even decreased slightly after the Federal Reserve stopped raising interest rates. Similarly, during the international financial crisis, the Federal Reserve implemented three rounds of QE and one round of QT to lower long-term bond yields. The effectiveness of QE is obvious. Gagnon (2016) summarized the results of 24 quantitative studies from 2008 to 2015, concluding that a 10% GDP scale of QE can reduce the yield on the 10-year government bond by 68 basis points, with a median of 54 basis points. However, further research can show that the yield reduction brought about by QE operations requires continuous purchases. Once stopped, the long-term bond yield is very likely to stabilize and rebound to a certain extent. For the operation of OT, the effect is not obvious.According to the news, the central bank has borrowed hundreds of billions of government bonds from primary dealers, accounting for about one-thousandth of GDP. Based on the experience of the United States' QE, the impact of such a scale of operation on long-term government bond yields is minimal, with only a short-term psychological effect. If the central bank wants to continue intervening in the government bond market, there are two ways to do so. One is to continue borrowing bonds from the market and then selling them, and the other is to sell the 1.5 trillion special government bonds it holds. Unless there is a particularly extreme situation, this method will basically not be considered. For the first method, after excluding savings-type government bonds and government bonds already used for repurchase transactions, there are about 20 trillion government bonds available for purchase in the market, and 8 trillion government bonds available for sale in the hands of primary dealers. Moreover, considering the need for banks to manage their positions, the term of the central bank's credit borrowing will not be too long. At this time, a reverse operation could hedge the effects of previous operations.

IV. The Performance of Spreads in the Context of Asset Scarcity

Spreads come in two forms: term spreads and credit spreads. Under asset scarcity, the manifestations of term spreads and credit spreads are not consistent.

(1) Short-end yields are more influenced by monetary policy

As mentioned earlier, there are many factors affecting long-end yields, while short-end yields are mainly influenced by monetary policy and its impact on liquidity. For example, during the pandemic, under the background of continuous interest rate cuts, liquidity was quickly released, and the decline in one-year government bond yields was faster than that of ten-year yields, with the spread once reaching 130 basis points. Similarly, after the second half of 2023, as the MLF rate was lowered, the spread also began to widen, rising from less than 50 basis points to 70 basis points. Over the past 10 years, the average spread between one-year yields and ten-year yields has been 67 basis points, so the current spread has not significantly deviated from the equilibrium level.

On July 22, the central bank lowered the Standing Lending Facility (SLF) rate and the open market (OMO) 7-day reverse repurchase operation rate by 10 basis points, and also lowered the loan market报价 interest rate (LPR) by the same margin. On July 25, the People's Bank of China conducted an MLF operation in the open market for the second time within the month, with an operation volume of 20 billion and a rate cut of 20 basis points. This operation not only reduced loan rates but, more importantly, narrowed the interest rate corridor by lowering the upper limit of the interest rate corridor, which is conducive to reducing the financing costs of banks. Therefore, for a foreseeable period, as the interest rate center moves down, short-end yields may be further lowered.

(2) The performance of spreads with different ratings under asset scarcity is not consistent

Asset scarcity has led to the current decline in government bond yields, and under the background of asset scarcity, the yields of high-quality assets have also declined, leading to changes in yields with different ratings. On the one hand, the yields with relatively high ratings have also declined, leading to a significant narrowing of spreads; on the other hand, for lower-rated industrial bonds, spreads have shown a trend of first widening and then narrowing.Bonds with high credit ratings are most typically exemplified by government development bonds. In the past, the spread between ten-year government development bonds and treasury bonds of the same maturity was between 50-100 basis points (BP). However, against the backdrop of an asset scarcity, especially with the sluggish growth in the stock scale of government development bonds, the lack of supply has led to a faster decline in the yield of these high-grade bonds. Currently, the spread has been reduced to within 10 BP, which is almost negligible. In other words, under the circumstances of asset scarcity, the market has essentially disregarded the benefits brought about by the tax-deductible nature of treasury bonds. There has been a craze for treasury bonds, which are also backed by national credit. For relatively higher-risk assets, credit spreads have expanded at one point due to increased risk appetite. However, in the second half of 2023, as the asset scarcity intensified, some funds passively increased their allocation to these so-called high-risk assets, leading to a reduction in spreads.

If the yield on ten-year treasury bonds continues to fluctuate after the third quarter, it is not ruled out that more funds will start seeking higher-risk and relatively higher-yielding assets for allocation in the future.

V. Main Conclusions

Firstly, according to the classic IS-LM model analysis, the economic downturn pressure caused by the reduction in investment due to real estate adjustments is the main factor leading to the current interest rate decline.

Secondly, the further decline in real estate leads to a more prominent contradiction of insufficient total demand, and the change in inflation expectations is the main factor for the decline in the yield on ten-year treasury bonds in 2023.

Thirdly, according to the constructed VAR model, in the next two quarters, the yield on ten-year treasury bonds will fluctuate between 2.2% and 2.3%. Another result calculated by the model is that without external intervention, the nominal GDP growth rate in the fourth quarter could potentially drop below 4.5%.

Fourthly, the central bank's attention to the yield on ten-year treasury bonds is not due to exchange rate impacts, but rather due to concerns that a rapid decline in yields could affect the net interest margin of banks, thereby leading to systemic risks in the banking system.

Fifthly, based on the experience of the United States' quantitative easing, the scale of the central bank's purchase and sale of treasury bonds in the open market must be sufficiently large and sustainable to have an impact on yields. Currently, the People's Bank of China's borrowing and subsequent selling of bonds only have short-term shocks and psychological impacts on the market.Sixthly, compared to the yields of long-term government bonds, the yields of short-term government bonds are more significantly influenced by central bank policies. It is anticipated that for a period in the future, as the central pivot of the interest rate corridor shifts downward and central bank monetary policy becomes relatively more accommodative, short-term bond yields may experience a more substantial decline.

Seventhly, the scarcity of assets has led to a rapid convergence of yields on government bonds and high-grade bonds, while the yield spread between government bonds and high-risk bonds first widens and then converges. This indicates that the intensification of asset scarcity has caused funds to passively increase their allocation to some high-risk bonds. It is expected that the relevant credit spreads may further converge.