2026 年 03 月 01 日 证券研究报告 | 策略周报 外部烽烟再起，稳健为主 策略周报 分析师：郝一凡 分析师登记编码：S0890524080002 电话：021-20321080 邮箱：haoyifan@cnhbstock.com 分析师登记编码：S0890524100002 电话：021-20321091 邮箱：liufang@cnhbstock.com 021-20515355 相关研究报告 1、《波动明显上升，适度回归稳健 —策 略周报》2026-02-01 2、《理性降温，景气度仍是避风港 —策 略周报》2026-01-18 3、《跨年波动或有上升，不改高景气主 线 —策略周报》2026-01-04 4、《继续耐心布局高景气 —策略周报》 2025-12-21 5、《由守转攻，布局高景气方向等风起 —策略周报》2025-12-07 投资要点 分析师：刘芳 【债市方面】止盈意愿有望减弱，区间震荡为主。春节后在两会前，债市存 在一定止盈压力，观望情绪上升。但外部冲突风险上升，风险偏好回落或利于 债市，债市调整风险有限。后续关注政策目标设定，若经济增长、赤字率目标 等设定并未明显强于预期，则债市风 ...

今 日 早 评 重点品种： 【短评-天然橡胶】泰国原料继续上涨。杯胶上涨0.3泰铢/ 公斤至56.8泰铢/公斤，胶水上涨0.7泰铢/公斤至68泰铢/公 斤，水杯价差扩至11.2泰铢/公斤；越南1月天然橡胶与混合胶 合计出口同比下降24%。1月共出口合计15.9万吨，其中出口到 中国12.6万吨，同比下降24%；截至2026年2月23日，中国天然 橡胶社会库存136.6万吨，环比增加7万吨，增幅5.4%。中国深 色胶社会总库存为92.6万吨，增7.1%。中国浅色胶社会总库存 为44万吨，环比增1.9%。评：东南亚产区处于季节性低产期， 原料供应持续收缩，海外原料价格处于上涨轨道，加工厂成本 高企。我国保税区库存累积。下游需求，下游逐步复工复产， 需求端提振预期仍在。综上，天然橡胶震荡偏强。 【短评-白银】美国上周初请失业金人数为21.2万人，预期 21.5万人，前值从20.6万人修正为20.8万人。评：就业数据略 有改善，市场对美联储降息预期有所减弱，并预期在鲍威尔卸 任之前，都不会有进一步降息行为，打压白银做多情绪。白银 进一步上涨动力不足，白银跟随黄金被动波动，中期暂看高位 震荡。 投资咨询中心 2026年 ...

Core Insights - The bond investment strategies of small and medium-sized banks are shifting from capital gains to a focus on stable coupon income due to narrowing yields and limited risk management tools [1][2][3] - Banks are adopting a strategy of shortening duration, controlling positions, and focusing on short-term bonds, which reflects a broader trend affecting the entire bond market's funding structure [1][2] Group 1: Investment Strategy Adjustments - Many banks are strictly limiting long-duration trades and primarily investing in short-duration bonds while selectively using leverage [2] - Some banks have maintained their duration but reduced trading volumes, opting for other assets like deposits and interbank certificates during periods of inactivity [2] - The overall strategy has shifted towards a balanced approach due to internal profit pressures and market volatility, with state-owned banks enhancing trading attributes while rural commercial banks significantly reduce bond market allocations [2][3] Group 2: Market Conditions and Challenges - The bond market is characterized by limited returns, with the ten-year government bond yield expected to fluctuate within a narrow range of 30 basis points in 2025, leading to reduced coupon income [3] - Most small and medium-sized banks lack risk hedging tools, making them vulnerable to market fluctuations, as they do not have the qualifications for derivative trading [3][4] - The reliance on adjusting cash positions rather than utilizing sophisticated risk management tools has led to a cycle of forced selling during downturns and missed opportunities during upswings [4] Group 3: Asset Allocation Trends - There is a clear stratification in bond asset allocation among different types of banks, with state-owned banks being the primary holders of government bonds, reflecting their focus on asset safety and liquidity [6] - Joint-stock banks and city commercial banks have similar holding structures, primarily investing in government and local bonds, while rural commercial banks maintain a higher proportion of financial and credit bonds in pursuit of better yields [6] - The pressure to achieve profit targets often leads to a shift of funds into financial markets when credit lending falls short, but the subdued bond market limits the potential returns on these investments [6] Group 4: Future Outlook - The ten-year government bond yield has decreased from approximately 2.82% at the beginning of 2023 to 1.68% by the end of 2024, resulting in significant floating profits for banks [7] - As net interest margins continue to narrow and bond risk-reward ratios decline, banks are increasingly pressured to realize floating profits to secure returns [7] - If a rate cut window opens in 2026, the pressure to realize floating profits may ease as the bond market enters a phase of declining interest rates [7]

Group 1 - The article discusses the prediction of medium to long-term government bond yields using a mixed neural network model that incorporates macroeconomic factors, monetary policy, and market sentiment [5] - The optimal model predicts that the ten-year government bond yield will decrease by approximately 3 basis points (BP) from the end of January 2026 to the end of February 2026, and by about 6 BP from the end of 2025 to the end of 2026 [5] Group 2 - In the banking sector, the total market wealth management scale is expected to grow by around 3.5 trillion yuan, reaching approximately 33-34 trillion yuan by 2025, driven by factors such as deposit disintermediation and the release of floating profits [5] Group 3 - In the real estate market, the new housing transaction area in 30 core cities is projected to decline by 19% year-on-year in 2025, while the average transaction price is expected to increase by 0.7% year-on-year [6] - The average price of second-hand houses in key cities for 2025 is reported as follows: Beijing at 28,194 yuan per square meter (up 2.6% year-on-year), Shanghai at 36,962 yuan per square meter (down 4.4%), Guangzhou at 25,832 yuan per square meter (down 7.1%), and Shenzhen at 56,703 yuan per square meter (down 1.6%) [6]

Group 1: Market Overview - The bond market is predicted to see a decline in the ten-year government bond yield by approximately 3 basis points from January to February 2026, and by about 6 basis points from the end of 2025 to the end of 2026 [2] - The internet media sector is expected to experience significant growth in the animated drama market, driven by AI video models, with a recommendation to focus on industry capacity release and specific companies like Reading Group [3] Group 2: Real Estate Sector - As of January 11, 2026, the new housing market in 20 cities recorded a total transaction of 13,000 units, down 50.4% year-on-year, with notable declines in major cities such as Beijing (30%), Shanghai (41%), and Shenzhen (78%) [4] - The second-hand housing market in 10 cities saw a total transaction of 24,000 units, a decrease of 26.2% year-on-year, with significant drops in Beijing (38%), Shanghai (13%), and Shenzhen (37%) [4] Group 3: Company Research - The report on North China Huachuang indicates that state capital cooperation is driving industry empowerment, with projected net profits of 7.628 billion CNY for 2025, 10.030 billion CNY for 2026, and a new forecast of 12.812 billion CNY for 2027 [5] - The report on Shannon Chip Creation highlights the benefits of rising storage prices driven by AI, with a net profit forecast of 605 million CNY for 2025, an increase to 1.043 billion CNY for 2026 (up 48%), and a new forecast of 1.251 billion CNY for 2027 [6]

Core Viewpoint - The introduction of the "Interbank Certificate of Deposit (NCD) Buy/Sell Power Comparison Index" by Xinhua Finance enhances the monitoring of short-term liquidity in the bond market, providing a comprehensive tool for understanding market sentiment and institutional trading dynamics [1][4]. Group 1: Index Overview - The NCD Buy/Sell Power Comparison Index focuses on highly liquid and actively traded money market instruments, filling a gap in short-end trading sentiment monitoring [1]. - The index is based on real-time trading data, tracking the buy/sell behavior of different types of institutions during specific time periods, resulting in a quantifiable sentiment indicator [1]. - The index value ranges from -100 to 100, where absolute values indicate the strength of buy/sell power comparison, with positive values indicating stronger buying power and negative values indicating stronger selling power [1]. Group 2: Institutional Classification and Trends - The index supports five types of institutions: banks, securities firms, broad-based funds, insurance companies, and other types [2]. - Recent data shows significant institutional differentiation in the NCD market, with banks showing a strong net buying position, while insurance and fund institutions are reducing their holdings, indicating a potential shift in market sentiment [4]. Group 3: Functionality and User Access - The "Bond Market Sentiment Radar" aggregates various real-time information, including the NCD Buy/Sell Power Comparison Index and market trends, allowing investors to gain insights into market depth and efficiency [4]. - Users can access the "Bond Market Sentiment Radar" through the Xinhua Finance professional terminal or integrate the index's real-time data into their internal systems for tailored business applications [6].

Report Industry Investment Rating No relevant content provided. Core Viewpoints - The current RMB exchange - rate appreciation is more likely the third scenario where domestic entities' risk preference changes, so it does not benefit Hong Kong stocks, and the sustainability of the RMB's strength is questionable. It is hoped that future RMB appreciation will be the first scenario, corresponding to the burst of the US tech bubble [17]. - In the de - leveraging cycle, the bias of the stock - bond ratio in favor of equities is limited, and the value style has a higher probability of relative dominance [9][58]. Summary by Directory 1. National Balance Sheet Analysis - **Liability Side**: In November 2025, the liability growth rate of the real sector was 8.6% (previous value 8.7%), expected to decline to around 8.3% in December. The money market continued to loosen marginally last week. The central bank's stance indicates that the goal of stabilizing the macro - leverage ratio remains unchanged, waiting for the quantitative fiscal target from the Two Sessions in 2026 [2][16]. - **Fiscal Policy**: Last week, the net increase in government bonds was 1707 billion yuan (higher than the planned 1148 billion yuan), and the planned net increase next week is 174 billion yuan. The government liability growth rate is expected to decline to around 12.4% in December from 13.1% in November [2][18]. - **Monetary Policy**: Last week, the average weekly trading volume and price of funds increased, and the term spread widened. The one - year Treasury yield dropped to 1.29% at the weekend. The market may have over - anticipated a 2026 interest - rate cut. The term spread between the ten - year and one - year Treasuries widened to 55 basis points. The money - market loosening may be approaching its limit [3][18]. - **Asset Side**: In November, physical quantity data showed signs of stabilizing at a low level. The full - year nominal economic growth target for 2025 is 4.9%, and it is necessary to confirm whether 5% will be the central target for China's nominal economic growth in the next 1 - 2 years [3][19]. 2. Stock - Bond Cost - Effectiveness and Stock - Bond Style - **Economic Cycle**: Since 2011, China has entered a downward cycle of potential economic growth, which seems to have ended in Q4 2024. The Chinese government put forward three policy goals in 2016. Currently, the convergence on the liability side is not over but has limited space. If the valuation of the US technology sector is re - evaluated, global funds may flow to China [6][20]. - **Last Week's Market Performance**: The money market continued to loosen, risk preference rose, and the growth style dominated. The stock - bond ratio favored stocks. The ten - year Treasury yield rose 1 basis point to 1.84%, the one - year yield dropped 7 basis points to 1.29%, and the 30 - year yield remained stable at 2.22%. The broad - based rotation strategy underperformed the CSI 300 by - 0.32 pct last week and - 5.41 pct since July 2024 [7][21]. - **Investment Recommendations**: Bonds at the long end are considered to have investment value. For equities, a neutral stance is taken, with a focus on style. Before seeing the government's bond - issuance plan, the value style is expected to dominate. This week, the Shanghai 50 Index (60% position), the CSI 1000 Index (20% position), and the 30 - year Treasury ETF (20% position) are recommended [8][22]. 3. Industry Recommendations - **Industry Performance Review**: This week, A - shares rose with increased trading volume. Among Shenwan primary industries, non - ferrous metals, national defense and military industry, power equipment, electronics, and building materials had the largest increases, while beauty care, social services, banks, coal, and food and beverage had the largest declines [28]. - **Industry Crowding and Trading Volume**: As of December 26, the top five crowded industries were electronics, power equipment, machinery, national defense and military industry, and non - ferrous metals. The trading volume of the whole A - share market rebounded. Transportation, non - ferrous metals, and other industries had the highest trading - volume growth rates, while banks, coal, and other industries had the largest declines [31][32]. - **Industry Valuation and Earnings**: This week, the PE(TTM) of non - ferrous metals, national defense and military industry, and other industries increased the most, while that of social services, beauty care, and other industries decreased the most. Industries with high 2024 full - year earnings forecasts and relatively low current valuations include banks, securities, and others [36][37]. - **Industry Prosperity**: Externally, there was a marginal decline. The global manufacturing PMI decreased in November, and export growth rates of some countries changed. Domestically, the second - hand housing price dropped, and quantity indicators fluctuated. The highway truck traffic volume declined, and the industrial capacity utilization rate showed a fluctuating trend [41]. - **Public - Fund Market Review**: In the fourth week of December, most active public - fund equity funds outperformed the CSI 300. As of December 26, the net asset value of active public - fund equity funds slightly increased compared to Q4 2024 [55]. - **Industry Recommendations**: In the de - leveraging cycle, the stock - bond ratio favors equities to a limited extent, and the value style is more likely to dominate. The recommended A + H dividend portfolio includes 13 stocks, and the A - share portfolio includes 20 stocks, mainly in industries such as banks, telecommunications, and oil and petrochemicals [9][58].

Core Viewpoint - The bond market in 2025 is characterized as a "non-bull, non-bear volatile market," with interest rates ending a four-year downward trend and the ten-year government bond fluctuating within a narrow range of 1.6% to 1.9% [1] Summary by Themes Market Characteristics - The market is showing new characteristics such as desensitization to macroeconomic factors, increased volatility, and a seesaw effect between stocks and bonds [1] Investor Experience - This year has been particularly impactful for investors in a "low interest rate" environment, with significant reductions in coupon returns and challenges in capital gains, as well as rapid shifts in market trends [1] Key Themes in the Bond Market - Important themes for the bond market this year include: 1. Respecting trends while also trusting common sense 2. The difficulty of executing wave trading strategies 3. The presence of structural opportunities even in a volatile market 4. Profound changes in the bond market ecosystem 5. A shift in asset allocation direction, indicated by the trend of looking at stocks while investing in bonds [1]

Group 1 - In 2025, major asset performance showed significant differentiation, with gold prices rising over 60%, A-shares experiencing a structural bull market, and the bond market entering a phase of volatility [1][6] - The driving forces behind these trends include a domestic financial cycle downturn, a restructuring of the global economic order, and a deepening industrial revolution led by AI [6][7] - The bond market faced challenges due to the "asset shortage" logic weakening and economic rebalancing policies entering a phase of turbulence [6][7] Group 2 - For 2026, the market is expected to be in a complex transitional phase, with a strategic focus on defensive positioning while tactically seeking structural opportunities in commodities and equities [2][10] - The stock market is anticipated to see a profit-driven rally supported by improved economic fundamentals, ample liquidity, and reasonable valuations, particularly in sectors like AI and semiconductors [3][15] - The bond market is projected to experience a wide fluctuation in yields, with the ten-year government bond yield expected to range between 1.6% and 2.1% [3][47] Group 3 - The commodity market is expected to continue structural differentiation, with basic metals like copper and aluminum benefiting from global fiscal expansion and industrial upgrade demands [3][63] - Gold is anticipated to undergo a phase of high-level consolidation, supported by weakened dollar credit and central bank purchases, while facing short-term volatility [3][74] - The investment logic for commodities is shifting from traditional economic cycles to a focus on strategic scarcity and essential resources, particularly in the context of global security and industrial needs [3][64][66]

Core Viewpoint - The article discusses the application of deep learning models, particularly LSTM, in predicting government bond yields, highlighting its advantages in handling complex financial time series data [4][5]. Group 1: Financial Time Series Prediction - Financial time series prediction has evolved through three main stages: traditional econometric models, traditional machine learning models, and deep learning models [4]. - Deep learning models, especially LSTM, are currently among the mainstream methods for financial time series prediction due to their ability to adapt to non-stationary, non-linear, high-noise, and long-memory characteristics [4]. Group 2: LSTM Model for Bond Yield Prediction - A three-layer stacked LSTM model with Dropout regularization was developed to predict the 10-year government bond yield, exploring the application and effectiveness of deep learning in fixed income quantitative analysis [5]. - The model utilized data from early 2021 to December 12, 2025, with approximately 130,000 adjustable parameters, achieving an average absolute error of 1.43 basis points in predictions [5]. - The model predicts a slight decline in the 10-year government bond yield, with a forecasted value of 1.8330% for December 19, 2025, down from 1.8396% on December 12, 2025 [5]. Group 3: Future Optimization Directions - Future optimizations include adjusting the model design regarding time windows, data processing, network architecture, and training strategies [6]. - Expanding input variables to include macroeconomic, market, and sentiment data will enhance the model's predictive accuracy and economic logic [6]. - Combining LSTM with traditional econometric models or other machine learning models to create hybrid models like ARIMAX-LSTM and CNN-LSTM-ATT can improve prediction precision [7].