Measure It Differently, and Inflation Is Behind Us

WSJDipublikasikan tanggal 2023-07-12Terakhir diperbarui pada 2023-07-12

Abstrak

If core inflation came in just below 3%, the Federal Reserve would breathe a huge sigh of relief, stocks would head to the races and consumers could relax about the rising cost of living.

The Fed is striving to tame inflation, which is at different levels depending on how it is measured. PHOTO: ERIN SCOTT/BLOOMBERG NEWS

If core inflation came in just below 3%, the Federal Reserve would breathe a huge sigh of relief, stocks would head to the races and consumers could relax about the rising cost of living.

It isn’t merely a dream: Measure U.S. price changes the way Europe does, and inflation was already there in May. Measure them as the U.S. does, and on Wednesday new figures are predicted by economists to show core inflation far higher, at 5% for June.

The U.S. and Europe use different methods to calculate inflation data, but the Bureau of Labor Statistics calculates American price rises the European way too, although the statistic remains obscure.

Right now, measuring U.S. inflation using the two methods shows radically different results. Investors who think they have a handle on the current consensus—that underlying inflation is falling but not fast enough for the Fed—should be troubled by the alternative message coming from the much lower European version of the figures.

U.S. core inflation—which excludes volatile food and energy—measured using the standard consumer-price index was 2.3 percentage points higher than the European-style inflation, known as the harmonized index of consumer prices. It is the biggest gap there has ever been.

The main reason is that Europe’s measure, known as HICP, doesn’t include the imaginary cost of what a homeowner would pay to rent their house, which makes up about a third of the U.S. core CPI. Known as “owners’ equivalent rent” or imputed rent, the measure has long had its critics.

Exclude something that no one actually pays, and which is calculated from guesses by homeowners of the rental value of their house, and core inflation’s looking basically fine, at a fraction under 3%. I’ve concentrated on core inflation, because food and oil prices swing so much that they make it hard to tell if the economy is generating inflation pressures the central bank needs to tackle.

So why is everyone still so concerned about inflation?

The answer is partly about biases and partly about change, but it is mostly about worries that the economy is running too hot to be confident that inflation will come down to the 2% target.

The bias is that CPI is long established and widely used in the U.S. The Bureau of Labor Statistics produces its own HICP inflation data as an experimental measure; many economists and investors don’t even realize it is available.

To make matters even more confusing, CPI gets almost all the focus, even though the Fed sets its inflation target based on the personal-consumption expenditures price index, PCE, from the Bureau of Economic Analysis. PCE comes in lower than CPI, but still puts large weight on imputed rent that isn’t actually paid.

Even if the Fed thought HICP was better—and there is no sign it does—there is no way it could change the measure without drawing political heat. Inflation figures are already subject to deep skepticism from some economists, who point out that improvements to the indexes usually make inflation come in lower than on older methods.

The most important question that the gauges seek to answer is whether the underlying pressures are so strong that the economy needs to be restrained further. Will a strong jobs market mean workers flush with pay increases can consume more, keeping demand up and so allowing companies to raise prices?

If so, the Fed will have to keep raising rates. The lower core HICP inflation suggests the problem of a strong jobs market is less worrisome than on the CPI measure. But hawks are concerned that continued above-inflation pay increases will push inflation up, as companies pass on higher costs. Persistent inflation above 2% might push up expectations of inflation, in turn leading to more pay demands.

Alternatively, will companies facing more expensive borrowing trim spending, hire fewer people and resist pay demands? If so, wages will moderate, demand fall and the Fed relax—at least so long as slowdown doesn’t turn into recession. Doves point to tentative signs that jobs are less plentiful, with initial unemployment claims up and wage increases decelerating for the hourly, less-skilled and lower-paid workers who were most in demand. Doves are also reassured by inflation expectations from consumers and investors not far from the 2% target.

Indicators that have a history of moving in tandem often converge again after a period of moving apart. CPI and PCE might well come down toward HICP, because rent increases have slowed. Unfortunately, even if core inflation does drop more, it will still be too high for comfort on the CPI and PCE gauges that investors and the Fed focus on.

Inflation isn’t the only place where the economic signals are haywire. Lots of the usual indicators of what the economy is doing and where it is going are telling different stories, something I’ll come back to in the next Streetwise. Meanwhile, all of this leaves me concerned—and confused, never a good place to be when trying to figure out what the market will do next.

Bacaan Terkait

Mantan Direktur AI Nvidia Mengguncang Transformer, AI Fisika dengan Konteks 5 Triliun, Memprediksi Seluruh Alam Semesta

Terobosan AI Fisika: Model 5 Triliun Konteks Mampu Simulasi Alam Semesta Tim dari Accelerated Understanding, yang didirikan oleh mantan Direktur AI Nvidia Anima Anandkumar dan suaminya Benedikt Jenik, memperkenalkan model AI revolusioner. Alih-alih menggunakan arsitektur Transformer standar, model mereka memanfaatkan **Neural Operators** untuk memahami dan mensimulasikan fenomena fisika secara langsung dalam **4 dimensi (ruang 3D + waktu)**. Model ini memiliki kapasitas konteks yang belum pernah terjadi sebelumnya: **1 triliun token selama pelatihan dan lebih dari 5 triliun selama inferensi**—setara dengan 5 juta kali kapasitas model bahasa terkemuka saat ini. Ini memungkinkannya melakukan **simulasi "one-shot"** untuk sistem fisik kompleks, menghasilkan lintasan gerak lengkap sekaligus tanpa subsampling atau pembagian data. Model yang sama dapat menangani masalah fisika yang berbeda-beda, dari aliran udara hingga dinamika plasma. Dibalik terobosan ini, tim menolak tawaran menggiurkan dari Jeff Bezos (35% ekuitas, gaji $2 juta, dan janji pendanaan $20 miliar) untuk bergabung dengan Project Prometheus. Mereka memilih independensi untuk mewujudkan visi "AI Fisika". Diduga kuat mereka mendapat dukungan komputasi dari Nvidia dan CEO Jensen Huang, yang sebelumnya sangat terkesan dengan pekerjaan awal Anandkumar di bidang prediksi cuaca berbasis AI. Ini menandai pergeseran paradigma dari AI yang hanya menghasilkan teks/gambar menuju AI yang memahami kode dasar alam semesta—hukum fisika—untuk simulasi dan optimisasi dunia nyata yang akurat.

marsbit12m yang lalu

Mantan Direktur AI Nvidia Mengguncang Transformer, AI Fisika dengan Konteks 5 Triliun, Memprediksi Seluruh Alam Semesta

marsbit12m yang lalu

Analisis Cepat Laporan Keuangan Nvidia: Pendapatan Triwulanan Siap Tembus $100 Miliar, Tahun Depan Masih Bisa Tumbuh 70%

Laporan Keuangan Nvidia Q2 FY2027: Pendapatan Kuartal Mencapai $962 Miliar, Panduan 70% Pertumbuhan untuk Tahun Depan Nvidia melaporkan kinerja kuartal kedua yang luar biasa, dengan pendapatan mencapai $962,21 miliar (naik 106% YoY), melampaui ekspektasi. Laba non-GAAP adalah $539,54 miliar. Perusahaan memberikan panduan untuk Q3 sebesar $1.080 miliar, yang akan menandai era pendapatan kuartalan triliunan dolar. Yang paling mengejutkan, CFO Colette Kress memproyeksikan pertumbuhan pendapatan sekitar 70% untuk tahun fiskal 2028, jauh lebih tinggi dari perkiraan pasar 45%. Proyeksi ini bahkan diberikan dengan asumsi keterbatasan pasokan dan tidak termasuk pendapatan pusat data dari China. Struktur pendapatan didominasi oleh bisnis Pusat Data, yang menyumbang lebih dari 90% total pendapatan dengan $890 miliar (naik 117% YoY). Pertumbuhan masih didorong terutama oleh investasi dari hyperscaler cloud seperti AWS, Microsoft, Google, dan Meta. Siklus produk baru telah dimulai dengan Vera Rubin, platform komputasi rak-level untuk AI, yang telah memasuki produksi massal penuh dan mulai dikirim. Setiap 1 GW daya komputasi yang diterapkan diperkirakan menghasilkan peluang pendapatan $400 miliar. Rubin diharapkan berkontribusi sekitar 20% terhadap pendapatan Pusat Data di Q3. Manajemen menekankan bahwa permintaan AI tidak melambat, tetapi justru meluas dari pelatihan model ke inferensi, AI perusahaan, dan robotika. Namun, pertumbuhan saat ini dibatasi oleh pasokan, termasuk HBM, kemasan lanjutan, dan daya pusat data. CEO Jensen Huang menyatakan bahwa panduan pertumbuhan akan "jauh lebih tinggi" tanpa kendala ini. Untuk mengatasi hambatan infrastruktur, Nvidia bermitra dengan firma modal besar untuk memobilisasi lebih dari $500 miliar dalam pembiayaan pihak ketiga dan mengamankan sumber daya seperti lahan dan listrik. Dengan panduan pertumbuhan 70% untuk tahun depan, Nvidia menyampaikan pesan bahwa siklus AI masih jauh dari berakhir. Ambang batas pendapatan kuartalan triliunan dolar mungkin hanya titik awal untuk tahap pertumbuhan berikutnya.

Odaily星球日报15m yang lalu

Analisis Cepat Laporan Keuangan Nvidia: Pendapatan Triwulanan Siap Tembus $100 Miliar, Tahun Depan Masih Bisa Tumbuh 70%

Odaily星球日报15m yang lalu

Trading

Spot
活动图片