というNBER論文が上がっている（H/T タイラー・コーエン（前回エントリのリンク先に同じ）；ungated版へのリンクがある著者の一人のページ）。原題は「Dollar Asset Holdings and Hedging Around the Globe」で、著者はWenxin Du（コロンビア大）、Amy W. Huber（ペンシルベニア大）。
以下はその要旨。
We analyze a large number of industry- and company-level filings of global institutional investors to provide the first comprehensive estimate of foreign investors' U.S. dollar (USD) security holdings and currency hedging practices. We document four stylized facts. First, driven by increasing portfolio allocations, foreign investors expanded their USD security holdings six-fold over the past two decades. Second, following the 2007-09 financial crisis, foreign mutual funds, insurers, and pensions raised their USD hedge ratio by an average of 15 percentage points, despite higher hedging costs implied by large and persistent deviations from covered interest-rate parity. The total FX hedging demand from these sector reached $2 trillion in 2019. Third, there is considerable heterogeneity in hedging practice across countries and sectors. Fourth, the global banking sector provides limited dollar hedging on net, underscoring the important role non-banks play in fulfilling the hedging demand of foreign institutional investors. We employ a mean-variance framework to benchmark investors' demand for USD assets and currency hedging practice, emphasizing the influence of expected returns on optimal portfolio construction and the apparent divergence between model predictions and observed hedging behaviors. We show a strong correlation between hedging demand and the cross-section of CIP deviations.
（拙訳）
我々は、多数の産業および企業レベルの世界の機関投資家の報告を分析し、海外投資家の米ドル建て証券の保有ならびに通貨ヘッジの実務について初めて包括的な推計を提示した。我々は4つの定型化された事実を明らかにした。第一に、ポートフォリオ配分の増加に伴い、海外投資家は過去20年間に米ドル建て証券保有を6倍に拡大した*1。第二に、2007-09年の金融危機後、海外のミューチュアルファンド、保険会社、および年金は、カバー付き金利平価からの大きな乖離が持続したことでヘッジコストが上昇したにもかかわらず、米ドルヘッジ比率を平均して15%ポイント引き上げた。これらの部門の為替の総ヘッジ需要は、2019年に2兆ドルに達した*2。第三に、国と部門によってヘッジの実務には顕著な違いが見られる。第四に、世界の銀行部門が提供するドルヘッジはネットベースでは限定的であり、海外の機関投資家のヘッジ需要を満たす上でノンバンクが果たす重要な役割を浮き彫りにしている。我々は、平均分散の枠組みを用いて、投資家の米ドル資産への需要と通貨ヘッジの実務のベンチマークを提示し、最適ポートフォリオ構築の期待リターンの影響と、モデル予測と観測されるヘッジ行動との間に見られる乖離を強調する*3。我々は、ヘッジ需要と、カバー付き金利平価からの乖離のクロスセクションとの間に強い相関があることを示す*4。
*1:本文では「This is not simply a reflection of larger foreign wealth. Rather, foreign investors are tilting their portfolios toward USD securities. Compared to pre-GFC, mutual funds, insurance, and pensions increased the share of USD securities in their overall portfolios by 7.7 percentage points post-GFC, and increased the share of USD securities in their non-domestic investments by 6.6 percentage points.」と記している。
*2:本文では「The hedge ratios for insurance companies, pension funds, and mutual funds were 44%, 35%, and 21%, respectively, as of 2020. The total hedging demand from these three sectors alone amounted to almost $2 trillion per annum. These investors’ hedge ratios were on average 14.7 percentage points higher post-GFC than pre-GFC. This new hedging regime developed despite elevated and fluctuating deviations from CIP, which increased the cost of hedging. In fact, the amount of foreign investors’ hedging activities tends to be higher when deviations from CIP are also wider. We calculate that the cost of hedging due to short-term CIP deviation averaged $2.7 billion per annum between 2017 and 2020 for the insurance and pension industries.」と記している。
*3:本文では「The investor’s problem can be thought of allocating a portfolio over the local-currency risk-free asset, a risky USD asset, and the short-term FX forward to hedge USD currency risk. We show that under general conditions, the optimal hedge ratio increases with respect to the volatility of FX risk, the covariance between FX risk, and the level of U.S. asset return. Furthermore, the optimal hedge ratio is affected by deviations from uncovered interest rate parity (UIP) and CIP. In particular, the optimal hedge ratio decreases in the expected return on the FX trade of going long USD and short the local currency, or the UIP deviation, and decreases in the additional cost of hedging USD back to the local currency beyond the interest rate differential, or the CIP deviation. Finally, the model also implies that USD holding and the optimal hedge ratio both increase in the excess return on the U.S. asset over the U.S. risk-free rate, all else being equal. After deriving the drivers of optimal hedging in the canonical mean-variance model, we compare predictions of these drivers with the data. In the time series, the rise in the hedge ratio is consistent with a rising expected return on the U.S. asset and rising USD allocation. However, in the cross section, using a panel regression with industry fixed effects, we see that the regression coefficients of hedge ratios on the expected FX returns, the covariance between FX and U.S. asset returns, and the cross-currency basis all bear the opposite sign from mean-variance predictions. 」と記している。
*4:本文では「This model of constrained intermediary delivers the prediction that the hedging cost rises in the hedging demand. Moreover, because the intermediary’s balance sheet is segmented across currencies in the spirit of Siriwardane, Sundaram, and Wallen (2022), shocks to local hedging demand explain the cross-sectional variations in the CIP basis. Consistent with this prediction, we show for each of our sample currencies that, its aggregate hedging demand, normalized by GDP, is strongly and negatively correlated with its average 3-month CIP basis. The cross-sectional R-squared of the relationship is equal to 0.77.」と記している。
