20171018-法国巴黎银行-COMMODITY_QUANT_STRATEGY_Introducing_the_factor_model_for_Nickel_10页_299kb

Summary of the Nickel Factor Model Analysis

Core Content

This document presents an analysis of the nickel price using a factor model approach, with a focus on identifying the most robust drivers of nickel prices over different time periods since 2007. The analysis employs the Lasso regression method, which is used for both variable selection and regularization to improve model interpretability and prediction accuracy. The findings are contrasted with the MarFA™ model, which is short nickel, and suggest that the recent price appreciation of nickel is a catch-up to fair value, not an overreaction.

Main Points and Key Variables

1. Lasso Regression Methodology

• Lasso is preferred over stepwise selection and ridge regression due to its ability to perform variable selection and regularization.
• It reduces collinearity and allows for a more interpretable and accurate model.
• The model is built using a large set of macro and financial monthly series (over 100 variables).

2. Variables Used in Different Periods

2007–2017

• RMB (trade-weighted)
• China monetary conditions (real interest rates, total loan growth, REER weighted)
• Global Nickel S/R Prods Production
• China Nickel imports (lagged 2 periods)
• China Housing Starts (lagged 2 quarters)

2010–2017

• USD Trade Weighted Index
• China Real Estate Source of Funds: Domestic Loan YoY
• China Commercialized Buildings Newly Built (Cum) YoY (lagged 2 periods)
• China Property - Land Areas Purchased YTD YoY (lagged 3 periods)

2013–2017

• China Loan Local & Foreign Currency YoY (lagged 4 periods)
• USD Trade Weighted Index
• China Freight Traffic YoY (lagged 12 periods)
• US 5y5y Forward-looking Real Interest Rates (lagged 12 periods)
• China Commercialized Buildings Newly Built (Cum) YoY (lagged 2 periods)

3. Model Results and Fair Value Estimates

• Consolidated Model (2007–2017):

  • Nickel is not substantially above fair value.
  • Recent appreciation is a catch-up to fair value.
  • Model central value: USD 11,286/MT
  • Error bands: USD 9,740–12,854/MT
  • Difference with market: 4.0%

• Model 1 (2007–2017):

  • Central value: USD 10,476/MT
  • Error bands: USD 8,610–12,343/MT
  • Difference with market: 12.1%

• Model 2 (2010–2017):

  • Central value: USD 11,582/MT
  • Error bands: USD 10,175–12,989/MT
  • Difference with market: 1.4%

• Model 3 (2013–2017):

  • Central value: USD 11,527/MT
  • Error bands: USD 10,050–13,064/MT
  • Difference with market: 1.9%

• Conditional Projections (Consolidated Model):

  • Dec-17: USD 11,301/MT, Difference: 3.9%
  • Jan-18: USD 11,267/MT, Difference: 4.2%
  • Feb-18: USD 11,251/MT, Difference: 4.3%
  • Mar-18: USD 11,158/MT, Difference: 5.2%
  • Apr-18: USD 11,172/MT, Difference: 5.1%

4. Key Findings

• The financial transmission channel (US dollar, CNY, and China credit expansion) has been a critical driver of nickel prices.
• The Chinese real estate sector plays a key role in influencing nickel prices, with a lagged positive effect.
• The model suggests that nickel is currently slightly overvalued if the last 10 years of data are considered.
• However, when using shorter periods, nickel appears to be at fair value.
• The recent price appreciation is predicted by the model, with nickel being ~12% cheap at the end of September 2017.

Commodity Quant Strategy Overview

• The strategy includes short-term trading and medium-term fair-value estimation.
• The short-term model focuses on financial market drivers to identify trading signals.
• The medium-term model considers fundamental factors such as supply and demand dynamics, inventories, and market positioning.
• The models are applied to various commodities including WTI crude oil, gasoline, copper, aluminium, iron ore, nickel, and gold.

Contacts

Legal and Disclaimer Information

• This document is non-independent research and is a marketing communication.
• It is not investment research and not subject to any prohibition on dealing ahead.
• The information is not guaranteed for accuracy or completeness.
• No liability is accepted for any use of the document or its content.
• No assurance is given that a transaction will be entered into on the terms indicated.
• The information may be subject to change and is provided for informational purposes only.
• Performance data may be based on back-testing and is not indicative of future results.
• Conflicts of interest may exist due to BNP Paribas' role in market-making, advisory, and investment banking activities.
• The document is confidential and may not be distributed without prior written consent.

United States Disclosures

• Options are complex instruments and may involve high risk.
• ETFs are subject to tracking error, currency, and geopolitical risks.
• Restricted securities may be discussed and are only available to Qualified Institutional Buyers (QIBs) or non-US persons under Regulation S.
• No disclosure of confidential information is required, and no liability is accepted for any use of the document.

United Kingdom Disclosures

• The document is communicated by BNP Paribas London Branch.
• It is subject to limited regulation by the Financial Conduct Authority (FCA) and Prudential Regulation Authority (PRA).
• The branch is supervised by the European Central Bank (ECB) and ACPR.
• The document is not a prospectus and is not intended for public offering.

France Disclosures

• The report is produced and/or distributed in France.
• It is not a prospectus under EU Directive 2003/71/EC.
• The document is produced by a BNP Paribas group company.
• It is intended for specific recipients and not for public distribution.