Are ADP and NFP Employment Numbers Correlated?

Ben Lengerich
4 min readJul 15, 2023

--

Two widely used employment indicators, ADP National Employment Report (ADP) and Non-Farm Payrolls (NFP), provide valuable insights but are controversial. Some analysts claim that there’s no correlation between ADP and NFP numbers, but a deeper analysis shows there’s more nuance beneath the surface. Specifically, outliers and context-specific correlations challenge the understanding of ADP and NFP.

Understanding ADP and NFP

To accurately interpret employment data, it’s essential to understand the methodologies and characteristics of ADP National Employment Report (ADP) and Non-Farm Payrolls (NFP) indicators.

A. ADP National Employment Report:

  • Methodology: ADP collects data from a diverse sample of private-sector employers, providing timely industry-specific insights.
  • Advantages: It offers early release, industry-specific data, and focuses on the private sector.
  • Limitations: ADP excludes government jobs, may have bias towards larger companies, and data revisions are possible.

B. Non-Farm Payrolls:

  • Methodology: NFP utilizes household and establishment surveys to provide a comprehensive view of employment trends.
  • Advantages: It covers both private and government employment and has a long historical dataset.
  • Challenges: NFP has a time lag, revisions may occur, and potential sampling errors can affect accuracy.

Understanding these nuances is crucial for interpreting employment data accurately. Next, we will explore the impact of outliers on the correlation between ADP and NFP, and strategies to address this challenge.

Enhancing Correlation Analysis: Outliers and Data Smoothing

Accurate correlation analysis between ADP and NFP requires addressing outliers and applying data smoothing techniques. Let’s examine the data. In all of the following experiments, we are using all NFP/ADP since 2010 with the COVID-19 outliers removed.

A. Impact of Outliers on Correlation Analysis:

  • Figure 1 shows the influence of outliers on the correlation between ADP and NFP.
  • Including outliers (blue) suggests there’s no correlation between ADP and NFP.
  • However, removing outliers (orange, outliers are defined as the min/max 2 percent of samples) reveals a meaningful positive correlation (r=0.43) between ADP and NFP.
The measured correlation between ADP and NFP is heavily influenced by outliers.

B. Enhancing Interpretation through Data Smoothing:

  • Figure 2 demonstrates the impact of data smoothing on the correlation between ADP and NFP.
  • 2-month smoothing increases the correlation to r=0.55 (without outliers).
Since both measures are noisy, the correlation between ADP and NFP is heavily influenced by smoothing.

Addressing outliers and employing data smoothing techniques contribute to more accurate correlation analysis between ADP and NFP.

In the next section, we will explore the context-specific correlation patterns between ADP and NFP, shedding light on their relationship and its implications for employment analysis.

Context-Specific Correlation Patterns:

Correlation analysis between ADP and NFP unveils interesting context-specific patterns that deepen our understanding of their relationship. We can use Contextualized.ML to find context-specific correlations in just a few lines of Python code.

A. Positive ADP Numbers:

  • When examining employment contexts with more positive ADP numbers, a tighter correlation between ADP and NFP emerges.
  • Figure 3 demonstrates this trend, highlighting the strengthened correlation at higher ADP values.
  • The positive ADP numbers indicate robust private-sector job growth, aligning with the overall employment trend reflected in NFP.
Correlation increases for more positive ADP readings.

B. Positive NFP Numbers:

  • Conversely, as NFP numbers become more positive, the correlation between ADP and NFP weakens.
  • Figure 4 showcases this pattern, illustrating a lower correlation at higher NFP values.
  • The divergence between ADP and NFP at positive NFP numbers suggests variations in employment dynamics, potentially influenced by government hiring and other factors.
Correlation decreases for more positive NFP readings.

Understanding these context-specific correlation patterns provides valuable insights into the nuances of the labor market. The tighter correlation at larger ADP numbers suggests ADP’s effectiveness in capturing private-sector employment trends, while the weaker correlation at larger NFP numbers underscores the broader influences and complexities affecting overall employment figures.

Implications

The correlation analysis between ADP and NFP provides valuable insights into employment trends:

A. Context-specific Correlation:

  • ADP exhibits a tighter correlation with NFP at positive ADP numbers, effectively capturing private-sector employment trends.
  • Conversely, as NFP numbers become more positive, the correlation weakens, reflecting variations influenced by government-related factors.
  • This context-specific correlation may be attributed to the different patterns of hiring: government hiring typically aspires to be counter-cyclical (increasing during economic downturns and decreasing during expansions), while private hiring is pro-cyclical (following the overall economic cycle). Thus, when economic times are bad, NFP looks better than ADP and the correlations break down.

B. Outlier Management and Data Smoothing:

  • Outliers in NFP data can distort the overall employment trend, often stemming from government-related events or policies.
  • ADP outliers may result from sector-specific shocks or natural disasters, impacting private-sector employment data.
  • Addressing outliers is crucial for accurate interpretation of employment trends, and it can be done through statistical techniques and visualization tools.
  • Additionally, data smoothing techniques help mitigate short-term fluctuations caused by outliers, revealing underlying employment trends.

By considering the context-specific correlation patterns, effectively managing outliers through statistical techniques, applying data smoothing techniques to reveal underlying trends, and understanding the cyclical nature of hiring, we can enhance the accuracy of employment analysis and make more informed economic decisions.

--

--

Ben Lengerich

Postdoc @MIT | Writing about ML, AI, precision medicine, and quant econ