We next set out the model and analyze a specul<strong>at</strong>or’s optimiz<strong>at</strong>ion problem. We then specializeto a symmetric sett<strong>in</strong>g to obta<strong>in</strong> analytical results. Section 4 offers numerical characteriz<strong>at</strong>ions,show<strong>in</strong>g how asymmetries affect outcomes. A conclusion follows. All proofs are <strong>in</strong> the appendix.IThe modelIn our multi-asset model of specul<strong>at</strong>ive trade <strong>in</strong> a stock market, specul<strong>at</strong>ors are str<strong>at</strong>egic and <strong>in</strong>ternalizehow their trades affect prices, and hence the <strong>in</strong>ferences and trades of other specul<strong>at</strong>ors.N risk-neutral <strong>in</strong>formed specul<strong>at</strong>ors and exogenous noise traders trade claims to M assets. Pricesare set by risk-neutral, competitive, un<strong>in</strong>formed market makers.We denote the vector of asset values by v ′ = (v 1 ,...,v M ). We consider a very general formul<strong>at</strong>ion<strong>in</strong> which asset values are l<strong>in</strong>ear functions of K underly<strong>in</strong>g fundamentals, e ′ = (e 1 ,...,e K ).These K fundamentals are jo<strong>in</strong>tly normally distributed with means th<strong>at</strong> we normalize to zero andan arbitrary variance-covariance m<strong>at</strong>rix, Σ e . The value of asset j is given byv j = v j1 e 1 + v j2 e 2 + ... + v jK e K .We can write the vector of asset values as v = V e, where V is the M × K m<strong>at</strong>rix,⎛⎞v 11 v 12 ... v 1Kv 21 v 22 ... v 2K..⎜ . . .. . ⎟⎝⎠v M1 v M2 ... v MKWe allow for the possibility th<strong>at</strong> each specul<strong>at</strong>or has access to many sources of priv<strong>at</strong>e <strong>in</strong>form<strong>at</strong>ionabout asset values. Specifically, we let specul<strong>at</strong>or i see a vector of signals about the valuefundamentals,s i = A i e, where A i is an L i × K m<strong>at</strong>rix,⎛ ⎞ ⎛s i 1 A i 11 A i 12 ... A i 1Ks i 2A i 21 A i 22 ... A i 2K=.⎜ . ⎟ ⎜ . . .. .⎝ ⎠ ⎝s i L iA i L i 1 A i L i 2 ... A i L i K⎞⎛⎟⎜⎠⎝⎞e 1e 2.. ⎟⎠e KExample: One specul<strong>at</strong>or sees an asset’s value. There are M assets, N = M specul<strong>at</strong>ors,and K = M fundamentals. The value of asset j is v j = e j , i.e., V is the M × M identity m<strong>at</strong>rix.5
Only specul<strong>at</strong>or i sees v i = e i , i.e., s i i = e i and s i j = 0, j ≠ i. The Ai m<strong>at</strong>rix captur<strong>in</strong>g specul<strong>at</strong>ori’s <strong>in</strong>form<strong>at</strong>ion is an M × M m<strong>at</strong>rix, which has only zero entries save for A i ii , which is one.Example: Each specul<strong>at</strong>or receives one signal for each asset. There are M assets, N specul<strong>at</strong>ors,and K = MN fundamentals. The value of asset j is v j = e (j−1)N+1 +...+e (j−1)N+i +...+e jN .Thus, V is an M × MN m<strong>at</strong>rix, where the j th row of V captures the value of asset j, so it hasv j,(j−1)N+i = 1, for i = 1,... ,N, and zeros elsewhere. Specul<strong>at</strong>or i sees s i j = e (j−1)N+i for assetj = 1,... M. Then A i is an M × MN m<strong>at</strong>rix, where j th row of the A i m<strong>at</strong>rix captures specul<strong>at</strong>ori’s <strong>in</strong>form<strong>at</strong>ion about asset j: it has A i j,(j−1)N+i= 1, and zeros elsewhere.In addition to his signals s i , specul<strong>at</strong>or i knows the prices <strong>at</strong> which orders for each asset willbe executed. Th<strong>at</strong> is, our economy is a noisy r<strong>at</strong>ional expect<strong>at</strong>ions economy <strong>in</strong> which agents arestr<strong>at</strong>egic and <strong>in</strong>ternalize how their trades affect prices and the <strong>in</strong>form<strong>at</strong>ion content of prices. Itis <strong>in</strong> this sense th<strong>at</strong> our model extends and melds the models of Adm<strong>at</strong>i (1985) and Caballe andKrishnan (1993), allow<strong>in</strong>g us to capture key aspects of a dynamic market <strong>in</strong> a simpler st<strong>at</strong>ic sett<strong>in</strong>g.Let x i j be specul<strong>at</strong>or i’s order for asset j, let xi′ = (x i 1 ,xi 2 ,... ,xi M) be the vector of his orders,and let X ′ = (X 1 ,X 2 ,... ,X M ) ′ = ( ∑ Ni=1 xi 1 ,∑ Ni=1 xi 2 ,...,∑ Ni=1 xi M) be the vector of net order flowsacross assets from specul<strong>at</strong>ors. In addition to trade from specul<strong>at</strong>ors, there is exogenous liquiditytrade of asset j of u j . We let u ′ = (u 1 ,u 2 ,... ,u M ) be the vector of liquidity trades. Liquiditytrade is distributed <strong>in</strong>dependently of the value fundamentals, and hence of specul<strong>at</strong>or signals. Weallow for a general correl<strong>at</strong>ion structure <strong>in</strong> liquidity trade across assets, assum<strong>in</strong>g only th<strong>at</strong> liquiditytrade is jo<strong>in</strong>tly normally distributed with zero mean and variance-covariance m<strong>at</strong>rix Σ u .Markets are cleared by competitive market makers who use the trad<strong>in</strong>g activity throughout theentire stock market when sett<strong>in</strong>g prices. In particular, a market maker observes the net order flowfor his asset, and the prices of other assets. Equivalently, with l<strong>in</strong>ear pric<strong>in</strong>g, market makers seethe vector of net order flows for all assets, (X + u) ′ = (X 1 + u 1 ,X 2 + u 2 ,... X m + u m ). Marketmakers set the price of each asset equal to its expected value given the vector of net order flows,i.e., p j = E[v j |(X + u) ′ ]: competition drives market-maker expected profits from fill<strong>in</strong>g each orderdown to zero. At the period’s end, assets are liquid<strong>at</strong>ed and trad<strong>in</strong>g profits are realized.We focus on equilibria <strong>in</strong> which specul<strong>at</strong>ors adopt trad<strong>in</strong>g str<strong>at</strong>egies th<strong>at</strong> are l<strong>in</strong>ear functions ofboth their priv<strong>at</strong>e signals and net order flows, and market makers set prices th<strong>at</strong> are l<strong>in</strong>ear functions6