The significant downward revision of US non-farm payrolls triggered discussions.
The adjustment amplitude of US non-farm payrolls data is large, mainly due to statistical factors.
Models that rely on traditional experience rules cannot adapt to major economic fluctuations and are prone to distortion. Before and after the outbreak, the US economy experienced greater fluctuations than usual, and the employment data obtained from the model needs to be significantly revised. The data revision includes: including missing or erroneous data in existing samples, and correcting sample biases.
The reason for the intense fluctuations in the US economy before and after the epidemic is twofold: firstly, due to the high elasticity of both the decline and recovery of the services sector; secondly, because of the reshoring of manufacturing and the AI-driven industrial revolution after the epidemic, there have been changes in the underlying economic operation logic.
The future interest rate reduction trading faces two complexities. 1. Focusing on data or misjudging the actual employment situation, because the model cannot accurately capture the employment reality of excessive fluctuations. According to the frequently adjusted employment data, the market will repeatedly gamble on interest rate cuts. 2. Market and Federal Reserve have different perspectives on data. There may be repeated game in interest rate operations and market interest rate expectations in the future.
Abstract:
The significant adjustment of US non-farm payrolls data is primarily due to statistical factors.
To accurately estimate sample and collected data, US non-farm employment data needs to be periodically revised.
Data revisions include: including missing or erroneous data in existing samples and correcting sample biases.
Historically, during periods of greater economic fluctuations, the number of employees and enterprises will also fluctuate significantly, and the magnitude of data revisions will be wider.
The reason behind this is that 'linear models' cannot accurately estimate non-linear changes in the economy. The underlying logic is that during periods of greater economic fluctuations, neither respondents of existing samples nor the samples themselves have followed historical trends. Therefore, compared with the 'relatively stable' model, the model accuracy is not enough compared to the drastic changes in reality.
After the outbreak, US employment data was significantly revised, reflecting a restructuring of the US economy.
In just four years before and after the outbreak, the US experienced a deep recession, strong recovery, and industrial revolution in between. This change can be said to be short-lived and bring great fluctuations.
In retrospect, it is impossible to estimate the depth of the recession in 2020 using pre-outbreak data, and recession data in 2020 will mislead data for the period of strong economic recovery in 2021. Therefore, the adjustment of US non-farm payrolls data from 2020 to 2021 is huge.
The adjustment of non-farm payroll data is concentrated in the service sector. The service industry is the sector that suffered the most severe decline in 2020. It is also one of the sectors with the strongest recovery in 2021 and beyond. A large number of service enterprises were newly established in the United States before and after 2021, and the growth rate far exceeded the historical average.
These changes cannot be captured by linear old models, causing employment data to be repeatedly significantly adjusted.
Under the new structural adjustment, future US interest rate trading will not be smooth.
The suppression of high interest rates on demand is not explicit, especially for the highly flexible service industry. The suppression of high interest rates on service demand may accelerate at some point. Statistical models may not accurately describe the speed at which US employment declines during economic recession. This also leads to the model overestimating service employment data using the rules of the past two years, and subsequently announcing a significantly deviated correction value according to the characteristics of US employment data revision.
In the future, we must closely monitor the employment data provided by the model, or misjudge the actual employment situation. The market will reassess and frequently gamble on interest rate expectations based on this indicator characteristic.
The market and the Federal Reserve have different perspectives on data, which will cause certain frictional gambles between the substantial interest rate reduction operations and the market interest rate expectations in the future. Therefore, this year's interest rate trading journey will not be smooth.
Article author: Zhou Junzhi, Sun Yingjie, Source: China Securities Co., Ltd. Original title: 'China Securities Co., Ltd.: Why did non-farm payrolls data in the United States make a significant adjustment?'
Editor/ping