Key takeaways
From the perspective of investment, the factors that affect the stock price include valuation and earnings, relative to the frequent fluctuations in valuation, profit forecasts are relatively certain, therefore, profit forecasts have always been the core content of stock investment; the systematic study of earnings is relatively large, according to the object can be divided into all A, industry, individual stocks, according to the content can be divided into profit tracking and profit forecasting; compared with profit forecasting, profit tracking analysis is the past, has occurred profit situation.
This paper only discusses the all-A earnings research, and introduces a top-down all-A earnings research framework for the overall performance of all A, which consists of three parts, profit tracking, profit forecasting, and profit expectation difference. Earnings tracker analyzes earnings data for past reported quarters, and earnings forecasts study earnings data for quarters that have occurred and are not disclosed in earnings reports and forecasts for future earnings data.
For all-A earnings tracking, the market generally adopts a bottom-up research method, through the collection of corporate financial data, the sum of all A earnings data, due to the lag in financial report disclosure, resulting in a significant reduction in the timeliness of all-A earnings tracking, and its significance is also reduced.
For all-A earnings forecasts, we use a top-down approach that leverages macro analysts' consensus expectations for economic indicators to make earnings forecasts. The top-down macro method is based on the premise that macroeconomic indicators are correlated with the overall earnings of enterprises, and there is a sufficient theoretical basis between the method and the premises, and there is a strong statistical correlation.
Under the top-down all-A earnings forecasting research system, the core of earnings forecasting is to establish a logical relationship between macroeconomic indicators and corporate earnings. To this end, we adopt the idea of multiple linear regression, taking macroeconomic indicators as independent variables and all-A earnings data as the dependent variables to construct a top-down all-A profitability model. The revenue growth model is mainly based on the year-on-year PPI and the year-on-year industrial added value as macroeconomic indicators, and the fitting effect is good, reflecting that mainland enterprises are mainly industrial and manufacturing companies, and the overall performance of enterprises is affected by the ex-factory price of products, and the higher the ex-factory price of products, the performance of enterprises will be greatly improved. The gross profit margin model is mainly constructed based on the year-on-year PPI and PPIRM year-on-year and industrial value-added year-on-year as macroeconomic indicators, reflecting the impact of product prices and raw material prices on the gross profit margin of enterprises. The profit model is mainly based on the year-on-year industrial added value and credit pulse as the macroeconomic indicators to construct, after excluding the financial sector, the fitting effect of the all-A profit growth model is greatly improved; the model shows that there is a strong logical relationship between the overall profit of the enterprise and the industrial added value and the credit pulse, and the industrial added value and credit pulse are highly correlated with the completion of fixed asset investment, whenever the overall profit of the enterprise increases significantly, the enterprise often converts part of the undistributed profits and the funds obtained from corporate financing into investment to purchase new equipment, Expenditures in the form of new production capacity to improve the production capacity of enterprises, therefore, there is a certain causal relationship between profit growth and industrial added value and credit impulse.
Based on the consensus expectations of macro analysts, we have constructed a 2024 earnings forecast, and expect revenue growth, gross margin, and profit growth to show a gradual recovery trend as a whole (see the table below). In 2024, the annual revenue growth rate of all A is expected to be 8.17%, the gross profit margin of all A is expected to be 35.94%, the profit growth rate of all A is expected to be 9.81%, the growth rate of all A (non-financial) revenue is expected to be 8.05%, the gross profit margin of all A (non-financial) is expected to be 24.50%, and the profit growth rate of all A (non-financial) is expected to be 3.83%. On the whole, the recovery of corporate profit growth is expected to be better than that of revenue growth, and the profit recovery pace of financial enterprises may be earlier than that of non-financial enterprises.
Risk warning: the risk of data statistical errors, model training risk, model overfitting risk, history does not represent the future, and the risk of inconsistency between model profit forecast and actual profit
1. Introduction
From the perspective of investment, the factors that affect the stock price are valuation and earnings, relative to the frequent fluctuations in valuation, profit forecasts are relatively certain, therefore, profit forecasts have always been the core content of stock investment; the systematic study of earnings is relatively large, according to the object can be divided into all A, industry, individual stocks, according to the content can be divided into profit tracking and profit forecasting; compared with profit forecasting, profit tracking analysis is the past, has occurred profit situation.
This article only discusses the All-A Earnings Study, and introduces a top-down All-A Earnings Research Framework for the overall performance of All-A, including earnings tracking and earnings forecasting.
Profit forecasting has always been a popular research direction, and many theoretical studies on profit forecasting have been put forward in academic circles, and investment circles have also invested in the results of profit forecasting, so the research methods of profit forecasting are more comprehensive. At present, there are two research methods for the all-A earnings forecast of major securities companies in the market, one is to make qualitative judgments with reference to Merrill Lynch's clock theory, and the other is to use the consensus expectations of macro analysts for quantitative analysis from top to bottom.
Compared with the discussion of earnings forecasting, profit tracking is often despised, which may be due to the serious lag in the mainstream research methods of profit tracking, resulting in profit tracking becoming "profit review";
In the top-down all-A earnings research system, profit tracking and profit forecasting need to be closely integrated, and the above two points are the main sources of investment income generated by profit analysis.
2. Top-down all-A profit system
All-A profitability research, a collection of the entire market, mainly analyzes revenue growth and profit growth, and the steps include tracking what has happened and predicting future trends.
At present, the systematic research on earnings in the market is mainly oriented to industries and individual stocks, and there is no mature all-A earnings research system in the form of industry research framework and in-depth reports on individual stocks; the reason is that macro analysts pay more attention to policies and macro indicators, while industry analysts pay more attention to individual stock earnings, and all-A earnings happen to be in the middle of the two, and get less attention; however, the importance of all-A earnings research is obvious, which can give macroeconomic analysis more micro, A wealth of new perspectives help macro analysts to look at macro indicators dialectically, and secondly, it can effectively help investment institutions to choose macro timing.
The top-down all-A earnings research system consists of three parts, earnings tracking, earnings forecasting, and earnings expectation difference (the framework is shown in Figure 1). Earnings tracker analyzes earnings data for past reported quarters, and earnings forecasts study earnings data for quarters that have occurred and are not disclosed in earnings reports and forecasts for future earnings data.
2.1 Profit Tracking
Profit tracking, standing at the current point in time, uses the disclosed financial data to measure the operating conditions and results of the enterprise in the past quarters, and understands the past operating results of the enterprise by measuring the revenue growth rate and profit growth rate.
For all-A earnings tracking, the market generally adopts a bottom-up research method, through the collection of corporate financial data, the sum of all A earnings data, due to the lag in financial report disclosure, resulting in a significant reduction in the timeliness of all-A earnings tracking, and its significance is also reduced.
2.2 Profit Forecast
When it comes to earnings forecasts, we generally refer to individual stock earnings forecasts, rarely for broad-based indices or the entire A-share market; individual stock earnings forecasts generally adopt a bottom-up analyst approach, with analysts' consensus expectations of corporate performance as the forecast value, while there are two methods for all-A earnings forecasts, one is that most domestic research institutions use Merrill Lynch clock theory to make qualitative judgments on overall earnings, and the other is to use the consensus expectations of macro analysts on economic indicators from top to bottom to make profit forecasts. Considering that qualitative judgments are subjective, we recommend a top-down approach to earnings forecasting.
The top-down macro method is based on the premise that macroeconomic indicators are related to the overall earnings of enterprises, and there is a sufficient theoretical basis for this method and premises. Grinold and Kroner (2002) proposed a model that links stock returns to GDP growth, arguing that corporate earnings growth is composed of GDP growth and corporate excess growth, where GDP growth is determined by labor growth and production capacity[1]. At the micro level, Issah and Antwi (2017) point out that macroeconomic factors have a significant impact on corporate earnings, and that corporate performance is a function of the previous year's ROA and macroeconomic variables, and that macroeconomic variables and the previous year's ROA can affect firm performance [3]. On the whole, whether from a macro perspective or a micro perspective, there is a high correlation between macroeconomic indicators and the overall profitability of enterprises.
From the perspective of data statistics, there is also a strong correlation between macroeconomic indicators and the overall profitability of enterprises, which confirms the core basis of the top-down macro law. According to the results of correlation analysis, the macroeconomic indicators are strongly positively correlated with the overall profitability of enterprises, the current price of GDP, PPI and export value are strongly positively correlated with the growth rate of revenue of all A (see Figure 2), among which the current price of GDP has the highest correlation of 0.80 year-on-year; the added value of industrial enterprises above designated size has a year-on-year There is a strong positive correlation between the current price of GDP year-on-year and the growth rate of all-A profits (see Figure 3), of which the year-on-year correlation of industrial added value is high, reaching 0.56. To sum up, the overall operating results of enterprises are mainly affected by macroeconomic development, and the revenue growth rate and profit growth rate of all A are strongly positively correlated with GDP growth.
Under the logical relationship between the profitability of all A enterprises and the macroeconomic indicators, by extracting the historical data, taking the past macroeconomic indicators as the independent variable and the historical profit data of the whole A as the dependent variable, a multiple linear regression model is established to find and simulate the logical relationship between the earnings of the whole A enterprises and the macroeconomic indicators.
Under the top-down approach, the all-A profitability research can play a role in changing the quarterly frequency to the monthly frequency, analyzing and predicting the profitability of the all-A on a monthly basis, which greatly breaks through the ability limit of the current mainstream profit analysis methods in the market. The top-down macro method, with the macroeconomic indicators as the independent variable and the historical profit growth rate as the dependent variable, establishes a multiple linear regression model, which measures the logical relationship between the macroeconomic indicators and the all-A profit. Through monthly earnings data, the "profit bottom" and "profit inflection point" can be grasped earlier, which plays an important role in macro research and investment.
By object, earnings forecasts can be divided into earnings forecasts for the current and undisclosed earnings quarters and earnings forecasts for future quarters.
2.2.1 Earnings Forecast: Quarters for which financial reports have been incurred and undisclosed
For quarters that have occurred but have not disclosed financial reports, our all-A earnings forecasts adopt a top-down research approach, using disclosed macro indicators as independent variables, and estimating earnings data through the logical relationship between macro indicators and earnings data. Compared with the bottom-up approach used in earnings tracking, the top-down macro method greatly alleviates the lag by using macro data, as shown in Figure 4, and the essential difference between the two methods is that the data used is different, with an average lag of 2 months in the disclosure of financial report data, while a slight lag of 15-20 days in the disclosure of macro indicators. The improvement of the timeliness of earnings data can naturally enhance the value of earnings information, which plays a more important and timely role in macro research and investment.
In essence, the top-down macro approach effectively balances the relative relationship between data accuracy and timeliness, and gives full play to the value of profit information. Careful comparison of the two methods, the bottom-up financial method has absolute accuracy and low timeliness, the top-down method greatly improves the timeliness by exchanging part of the data accuracy; for the revenue growth rate of all A, the revenue growth rate measured by the top-down macro method and the actual revenue growth rate of the bottom-up method have a small error and high accuracy, see Figure 6; for the profit growth rate of all A, there is a small error between the profit growth rate measured by the top-down macro method and the actual profit growth rate of the bottom-up method, and the accuracy is average, but there is consistency between the two from the perspective of trend, see Figure 7。
2.2.2 Earnings Forecast: No Quarter
For non-occurring quarters, similar to the previous article, we make full use of the inherent correlation between macroeconomic indicators and corporate earnings, and transform the economic indicators under the consensus expectations of macro analysts into earnings forecasts of all A-companies through the discovered logical relationship.
In fact, the top-down macro method is generally adopted in the profit forecasting of academic circles, and there are many theoretical and empirical studies on this method. Nolen (2012) found that corporate earnings growth has a long-term mean reversion effect, that is, when corporate earnings growth is too high or too low, it will return to the market average in the next 1-3 years[4]. In 2009, economist Micheal Biggs proposed the "credit impulse", which is defined as the new credit demand divided by nominal GDP, which can be used to predict the business cycle of the real economy; when enterprises in the real economy expect economic prosperity, increase leverage to expand production capacity, resulting in an increase in credit impulses, and increase corporate future profits; when enterprises in the real economy expect an economic recession, deleveraging reduces production capacity, resulting in a decline in credit impulses and future earnings of enterprises. Yan Shu et al. (2013) brought together more than 140 macroeconomic variables representing trends in the real economy, price information, financial conditions, and labor markets, and found that macroeconomic information can improve the accuracy of predicting firms' future earnings [5].
To summarize the above theories, the growth rate of corporate earnings has the characteristics of mean reversion, which is a time series function that fluctuates up and down around a certain center for a long time. Overseas research institutions often use the Cobb-Douglas production function to determine the long-term center of GDP growth, which is determined by population growth, capital investment, From the micro level of enterprises, enterprises finance the purchase of equipment to expand production or develop new business, investment slowly turns into production capacity to drive profitability, and the total financing behavior of enterprises is closely related to macro social financing; it takes a certain amount of time for credit impulses to be transmitted to profits, according to our calculations, it takes about 1.75 years for credit impulses to be transmitted to corporate profits, see Figure 8.
2.3 Poor earnings forecasts
After the establishment of the top-down all-A earnings forecasting system, the comparison between the forecasts is part of the top-down all-A earnings research system. Earnings forecasts for the quarters that have occurred and have not been disclosed represent "reality", and earnings forecasts for quarters that have not occurred represent "expectations", and as time continues to advance, the non-occurring quarters gradually become quarters of occurring and undisclosed financial reports, and the "reality" is constantly updated, and the "expectations" or expectations of the past are different. In the all-A earnings research system, the above three points are the main sources of investment income generated by profit analysis.
3. Research on all-A profitability model
Under the top-down all-A earnings forecasting research system, the core of earnings forecasting is to establish a logical relationship between macroeconomic indicators and corporate earnings. To this end, we adopt the idea of multiple linear regression, taking macroeconomic indicators as independent variables and all-A earnings data as the dependent variables to construct a top-down all-A profitability model.
Taking 2013Q2-2023Q3 as the data interval, the macro indicators of independent variables include nominal GDP growth, CPI year-on-year, PPI year-on-year, industrial added value year-on-year, fixed asset investment growth rate, export amount year-on-year, M2 year-on-year, credit impulse, PPIRM year-on-year, and the dependent variables include a variety of dependent variables, which can be divided into all-A and all-A (non-financial) according to the research object, and revenue growth, gross profit margin, and profit growth according to the research content, as shown in the following table.
The construction of the all-A profitability model adopts the standard process of data modeling, feature engineering, modeling parameter tuning, and model evaluation.
1. Feature engineering: Data processing includes handling outliers (culling or replacing anomalous data), handling missing values (populating, deleting, or estimating missing data), and data normalization (converting data into a format with zero means and unit variance so that features of different magnitudes can be fairly compared). Feature extraction extracts meaningful information from raw data and builds new features through aggregation, summary statistics, and the use of domain knowledge. Feature selection aims to identify which features are most useful for prediction, and methods include correlation analysis, principal component analysis (PCA).
2. Modeling parameter tuning: Select the appropriate model according to the nature of the problem, use technology to find the best model parameter settings, find the best parameter combination to improve the performance of the model through multiple rounds of testing, and achieve hyperparameter tuning, and cross-verify to ensure that the model performs stably on different data subsets, so as to avoid overfitting.
3. Model evaluation: select appropriate performance evaluation indicators according to the model type, conduct in-depth analysis of the prediction results of the model, and pay attention to the interpretability of the model decision-making process.
3.1 All-A revenue growth rate
The industrial added value and PPI were used to simulate the operating income of all A (both of which used the standardized quarter-frequency year-on-year value for regression analysis). The regression results are shown in Fig. 10 and Fig. 11, and the model has a good fitting effect, with a goodness-of-fit of 0.81. The following is the standardized model function of the revenue growth rate of all A, which represents the logical relationship between the revenue growth rate under standardization and the year-on-year growth rate of PPI and the year-on-year growth of industrial added value, both of which are positively correlated with the growth rate of revenue.
3.2 Gross profit margin
To predict the gross profit margin, the first problem that needs to be faced is that there is a seasonal disturbance in the gross profit margin, the interference is mainly concentrated in the first quarter, due to the suspension of production during the Spring Festival, the gross profit margin of the enterprise in the first quarter fell sharply compared with the other three quarters, but the macroeconomic indicators are eliminated in the year-on-year value of the next season, in order to exclude the impact of the first quarter, we only select the gross profit margin data in the second, third and fourth quarters for simulation. PPI, industrial added value, and PPIRM were used to simulate the gross profit margin of all A (the data interval excludes the first quarter, and the data of the second, third and fourth quarters are retained, the dependent variable is the standardized gross profit margin, and the independent variables are all standardized quarter-frequency year-on-year values), and the regression results are shown in Figure 12 and Figure 13, and the model fitting effect is good, and the goodness of fit is 0.68. The following is the standardized model function of the all-A gross profit margin, the ex-factory price of the product represented by PPI is directly proportional to the gross profit margin of the enterprise, and the production cost represented by PPIRM is inversely proportional to the gross profit margin of the enterprise.
3.3 All-A profit growth rate
The regression results are shown in Fig. 14 and Fig. 15, and the model fitting effect is average, and the goodness of fit is only 0.40. The following is the standardized model function of the profit growth rate of all A, and the industrial added value and credit pulse are positively correlated with the profit of all A, and the profit growth rate is more sensitive to the change of industrial added value than the credit pulse.
3.4 All-A (non-financial) revenue growth
The regression results are shown in Fig. 16 and Fig. 17, and the model fits well with goodness-of-fitting, with a goodness-of-fit of 0.81. The following is a standardized model function of all-A (non-financial) revenue growth, and the fitting effect of the revenue growth model is slightly improved after excluding banks, securities companies, and insurance companies.
3.5 All-A (non-financial) gross margin
Compared with the gross profit margin of all A, the gross profit margin of all A (non-financial) excludes the financial sector without gross profit margin, and the overall central level has declined, which is in line with the gross profit margin of actual industrial enterprises. PPI, industrial added value, and PPIRM were used to simulate the all-A (non-financial) gross profit margin (the data interval excludes the first quarter, only the data of the second, third and fourth quarters are retained, the dependent variable is the standardized gross profit margin, and the independent variables are all standardized quarter-frequency year-on-year values), and the regression results are shown in Figure 18 and Figure 19, and the model fitting effect is good, and the goodness of fit is 0.74. The following is a standardized model function for all-A (non-financial) gross margin:
3.6 All-A (non-financial) profit growth
The regression results are shown in Fig. 20 and Fig. 21, and the fitting effect of the model is greatly improved compared with that of all A, with a goodness-of-fit of 0.57. The following is a standardized model function for all-A (non-financial) profit growth:
4. Analyze and predict
Summarizing the above six models (see the table below), the revenue growth model is mainly constructed with PPI year-on-year and industrial added value as macroeconomic indicators, and the fitting effect is good, reflecting that mainland enterprises are mainly industrial and manufacturing companies, and the overall performance of enterprises is affected by the ex-factory price of products, and the higher the ex-factory price of products, the higher the ex-factory price of products, the performance of enterprises will be greatly improved. The gross profit margin model is mainly constructed based on the year-on-year PPI and PPIRM year-on-year and industrial value-added year-on-year as macroeconomic indicators, reflecting the impact of product prices and raw material prices on the gross profit margin of enterprises. The profit model is mainly based on the year-on-year industrial added value and credit pulse as the macroeconomic indicators to construct, after excluding the financial sector, the fitting effect of the all-A profit growth model is greatly improved; the model shows that there is a strong logical relationship between the overall profit of the enterprise and the industrial added value and the credit pulse, and the industrial added value and credit pulse are highly correlated with the completion of fixed asset investment, whenever the overall profit of the enterprise increases significantly, the enterprise often converts part of the undistributed profits and the funds obtained from corporate financing into investment to purchase new equipment, Expenditures in the form of new production capacity to improve the production capacity of enterprises, therefore, there is a certain causal relationship between profit growth and industrial added value and credit impulse.
Considering that the macroeconomic indicators used in the model are mainly based on industrial data, the model can naturally better measure the profitability of industrial enterprises; in the modeling process, we have tried to add sensitive macro indicators of financial enterprises, but such macro indicators are prone to multicollinearity problems with industrial added value and PPI, which have a counterproductive effect on the fitting of logical relationships.
Based on the consensus expectations of macro analysts, we have constructed a 2024 earnings forecast, and expect revenue growth, gross margin, and profit growth to show a gradual recovery trend as a whole (see the table below). In 2024, the annual revenue growth rate of all A is expected to be 8.17%, the gross profit margin of all A is expected to be 35.94%, the profit growth rate of all A is expected to be 9.81%, the growth rate of all A (non-financial) revenue is expected to be 8.05%, the gross profit margin of all A (non-financial) is expected to be 24.50%, and the profit growth rate of all A (non-financial) is expected to be 3.83%. On the whole, the recovery of corporate profit growth is expected to be better than that of revenue growth, and the profit recovery pace of financial enterprises may be earlier than that of non-financial enterprises.
In 2024, the revenue growth rate of all A is expected to be 8.17%, which will deviate from the central level in 2023, forming an "expected inflection point" in demand, showing a gradual recovery trend throughout the year, or accelerating the recovery rate in the second quarter. The revenue growth rate in the first quarter is expected to be 4.98%, and the revenue growth rate in the second quarter will increase by 2.42pct to 7.40%, and the third and fourth quarters will still gradually recover but the pace will slow down, see Figure 24.
In 2024, the gross profit margin of all A is expected to be 35.94%, which is better than the average level in 2023, showing a gradual recovery trend throughout the year, and in the second quarter of 2024, the gross profit margin of all A is expected to be 35.81%, with a slight decline in the third quarter, followed by a continuous recovery in the fourth quarter, see Figure 25.
In 2024, the profit growth rate of all A is expected to be 9.81%, which is higher than the central level in 2023, showing a gradual recovery trend throughout the year, or the recovery rate will accelerate significantly in the fourth quarter, forming an upward cycle, and the profit growth rate in the third quarter is expected to be 9.92%, and the profit growth rate in the second quarter will increase by 3.78pct to 13.70%, see Figure 26.
In 2024, the growth rate of all-A (non-financial) revenue is expected to be 8.05%, which will deviate from the central level in 2023, and show a gradual recovery trend throughout the year, or the recovery rate will accelerate in the second quarter. The revenue growth rate in the first quarter is expected to be 3.97%, and the revenue growth rate in the second quarter will increase by 3.07pct to 7.04%, and the third and fourth quarters will still gradually recover, but the pace will slow down, see Figure 27.
In 2024, the gross profit margin of all A (non-financial) is expected to be 24.50%, better than the average level of 2023, showing a steady recovery trend throughout the year, and in the second quarter of 2024, the gross profit margin of all A (non-financial) is expected to be 24.33%, with a slight decline in the third quarter, followed by a slight recovery to 24.83% in the fourth quarter, see Figure 28.
In 2024, the profit growth rate of all A (non-financial) is expected to be 3.83%, basically out of the negative growth range in 2023, showing a gradual recovery trend throughout the year, or the recovery rate will accelerate significantly in the fourth quarter, forming an upward cycle, and the profit growth rate of all A (non-financial) in the first quarter of 2024 will still be negative, and after the third quarter, the profit growth rate in the fourth quarter will increase significantly by 20.83pct to 20.42%, see Figure 29.
On the whole, the top-down all-A profit forecasting research can play a more important role in macro research and investment analysis by making full use of rich and multi-level macro indicators to estimate and predict the all-A profit through the logical relationship between macroeconomic indicators and the profitability of all-A enterprises. In 2024, the profit growth rate of all A is expected to be 8.17%, the profit growth rate is expected to be 9.81%, and the annual corporate earnings are expected to gradually improve.
5. References
[1] Grinold R and Kroner K.The Equity Risk Premium: Analyzing the Long-Run Prospects for the Stock Market[J]. Investment Insights,2002,5(3): 7-33.
[2] Abaidoo R and Ofosuhene Kwenin D. Corporate profit growth, macroeconomic expectations and fiscal policy volatility[J]. International Journal of Economics and Finance,2013,5(8):25-38.
[3] Mohammed Issah and Samuel Antwi. Role of macroeconomic variables on firms’performance: Evidence from the UK[J]. Cogent Economics & Finance, 2017，5(1):1-18.
[4] Foushee S N, Koller T, Mehta A. Why bad multiples happen to good companies[J]. Mckinsey Quarterly, 2012, 3: 23-25.
[5] Yan Shu, David C. Broadstock, Bing Xu. The heterogeneous impact of macroeconomic information on firms' earnings forecasts[J]The British Accounting Review,2013,45( 4):311-325.
6. Risk Warning
1. Risk of data statistical errors: The data handling process may lead to data errors
2. Model training risk: Model training risk includes the risk of wrong model selection and parameter error
3. Risk of model overfitting: When the model puts too many indicators, it is easy to cause the model to be distorted and does not have the ability to predict
4. History does not represent the future: The model is based on certain financial logic assumptions, and if the macro environment changes drastically in the future, it may affect the effect of the model
5. Risk of inconsistency between the model profit forecast and the actual profit: There is a risk that the profit data predicted by the model is inconsistent with the actual profit data in the future