The Secret Papers of Madame Olivetti

Annie Vanderbilt

16667.FM SecretPapersOfMadameOli3 3

7/10/08 1:40:43 PM

Prologue
He comes to her in the evenings, after work. They make love and then talk, she in English, he in French, while beyond the open windows and doors of her bedroom the sky turns a lively pink, not unlike the color that flushes her cheeks. What will your mother be making you for dessert tonight? Choco- late pots de crme? she wonders. Or something less sweet. Perhaps a tart of quince and sour lemons? He grins, uncomprehending, a simple man who asks for nothing more than the generosity of her body and the lazy postcoital murmur of her voice speaking to him in a language he does not understand. I am feeding you my life, she tells him. In heaping spoonfuls. He groans, aroused and hungry, as if her words are a feast of exotic dishes she has set before him. Below them, in the village, in a neat stone house that overlooks the sea, his mother is preparing dinner. She suspects that Yves is up the hill with the American widow, but she cares only enough to punish him in small ways: a bitter smile when he compliments her on the tarte au poire she has made for his dessert; a hostile arms length between them when she kisses the air, not his cheeks, to wish him good night. It is seven p.m. when Yves turns his truck up the rutted gold-dust road to Lilys house. It is eight thirty when he rumbles past the news vendors shop, headed for home.The villagers watch but say nothing, for Yves is one of their own. So, too, is Lily, and it is no one elses business. This is nothing new. There have been plenty of sexual goings-on in La Pierre Rouge, the house of the red stone, over the past century. Youll be late, Lily warns, this time in French.
16667.FM SecretPapersOfMadameOli11 11

7/10/08 1:40:44 PM

P r o l og u e
I dont care. She reminds him gently, She is your mother. You live in her house. You must think of her dignity. I think of this. He caresses her breast and the smooth inward slope of her thigh. Sex for them is a healing unguent, liberally applied but short term and, thus, light on the soul. Quickly then., she says. Twenty minutes later, rattling through town in his old Deux Che- vaux truck,Yves waves at the news vendors wife. Frowning, she points at her watch. It is almost nine oclock. For the first time in years he will be late for dinner. The thought flits through Yvess mind: the grumpi- ness of Madame Bibot must sour the flavor of her husbands dinner. He toots his horn to try to cheer her up, but he is thinking of Lily. He is wondering what is the story of her life, unaware that he has heard every word of it. She has fed it to him, in English, in heaping spoonfuls.
16667.FM SecretPapersOfMadameOli12 12
ily Crisp was aware of the whispers that heralded her arrival in the village, the eyes peeping out from behind half losed shutters and starched lace curtains. She trusted that the c whispers carried only kind feelings toward her and her family, although, as she drove through the center of town, she noticed the news vendor and his scowling wife locking up their shop for the evening. Now there was a woman who shunned goodwill, Lily thought. Madame Bibot was probably remarking that Madame Crisp looked unreasonably vibrant and healthy for a widow. Monsieur Bibot, who had a wandering eye, though he tended toward women more buxom and Germanic than Lily, would doubtlessly have said something to enflame his wife, some nonsense about Lilys skin glistening with the sheen of sorrow, which adds luster to a womans complexion. Or how Lily and her husband had made a fine couple and, with their two children, such a happy family. The fairy tale of familial happiness no longer charmed Madame Bibot. In fact, glancing in the rearview mirror, Lily saw that the news vendors wife was glaring at her husband. Monsieur Bibot was gesturing boldly, perhaps attempting to calm the waters by suggesting that grief had certainly aged Lily Crisp. She must be. what? On the near side of fifty? On the far side, Madame would be replying tartly, and Lily smiled because she knew that Madames assessment of Lilys age, and Lilys assessment of Madames response, were both on target.

16667.01 SecretPapersOfMadameOli1 1

7/10/08 1:41:09 PM

A n n i e Va n d e r b i l t
Lily had visited the village most summers over the past twenty years. At first she had come with her husband, Paul, and her son, Pierre. Paul had seldom stayed more than a weekhe was busy with his cows, gainful employment of which the villagers approved. This was something they could understand: livestock, hard work, and commitments. His concern for his herd of cows in a land called Idaho, reportedly filled with snakes and dried bushes, balanced the polite awe in which they had always held him, for he was a wealthy man, had inherited money, unlike his mother, who had been born in the village, not a penny to her name; but with that saucy beauty, who needed a fortune? She had married one. Lily remembered those early visits, spending a month at the house that Paul had inherited, reading to Pierre under a plane tree or walking down the hill to the beach with his hand clasped in hers, building sand castles, splashing in the sea, setting up an umbrella to protect her sons fair skin from the burning rays, then dragging their weary, salt-sticky bodies back up the hill to shower in a dribble of rust-colored water. Pierre had been in his teens when Lily had presented him with a baby sister. The next summer there had been no shortage of raised eyebrows and knowing smilessixteen years and Paul Crisp, still busy between the sheets, and potent!when this beautiful child, with her startling blue eyes and corn-silk hair, had turned up in the village. Was it only two years ago, Lily mused, when the four of them had been sitting on the beach in folding chairs and Justine, aged seven then, had asked her father, Whats the name of the water out there, Daddy? What do you call it? Lilys Lily Pond, he had told her. Lilys Lily Pond, she had chimed in with him, and, crawling onto his lap, had squirmed into position with her back against his chest, straddling his knees. He had wrapped her up in his arms and set his chin on the top of her silky blond head, and the
16667.01 SecretPapersOfMadameOli2 2
The Secret Papers of Madame Olivetti
two of themone dark-haired, one lighthad gazed at the sea, limpid and pale, bleached to the blue of the faded blue sky. Justine had asked, Is that because Mom floats her dreams on it? Absolutely. On blow-up hippos and walruses. And dragons? And elephants. Do the dreams ever fall offlike little girls? she had said, in a hushed voice. Never ever, her father had assured her. Lily had put down the book she was reading, smiled over at them, and asked, Shall we run down to the sea and float with my dreams? Oh, yes, Daddy, can we? Justine had cried, and he had scooped her up and carried her, squealing with delight, across the sand, to where he sailed her back and forth, belly down and arms extended, over the water. Justine had whooped and yelled for her brother, Pierre, to come and see her, she was flying, she was flying like an angel. Squinting through the windshield, Lily returned her thoughts to her driving and swore softly as her rental car, its engine grinding, lurched up the gold-dust road to her house. Midway up the hill, the car stopped with a judder, then charged ahead in a great jump and shuddered. Dont die on me, she threatened, calling down the wrath of the gods on this cursed Peugeot the color of a turnip. It jolted forward, up and over a ledge that scraped the oil pan. Think light, she urged as the vehicle swerved into the driveway and halted abruptly. She cut the engine. Relief washed over her, and, in the silence that followed, she could almost hear the ebb and swirl of the profound exhaustion that engulfed her limbs. Here at last she could sleep. Here at last she could sort through her memories and write them down, let in air and light and tidy the clutter of a scattered life.

16667.01 SecretPapersOfMadameOli3 3
Somewhere nearby a dog was barking. She glanced around her. It was growing late, the skys mauve glow having faded into blackness. Yet, the suns warmth lingered. Heat enfolded her body as if she had been wrapped up in moist leaves and shunted into an oven.The air trembled, garnering each whiff of moisture that trailed up from the resting sea. She fanned her skin, cooling the sweat that had rolled down her thighs, her shoulders, and dampened her armpits. Not surprisingly, her temples throbbed. She had drunk a full bottle of wine with dinner; it had gone down like water, in great cooling gulps that, afterward, had made her head spin and her mouth grow sticky. Nighttime was when the sadness assailed her.The pall of loss spread over her then, like a cloth dipped in ashes and smoothed across her skin. She had felt smothered at first, as though if she struggled the sorrow might do greater damage than slow the beating of her heart or interrupt her dreams. But drinking with grief as ones sole companion was not a pastime she intended to pursue; although it blunted the sadness, it brought down an early curtain on the evenings, and there would be thirty-eight evenings until her childrens visit. Self-imposed, she might add. Her choice to come here, to the south of France, to La Pierre Rouge, the old stone house on a hill above the sea that would be her refuge. For one moment, before hefting her luggage from the backseat of the car and carrying it to the house, Lily closed her eyes, thought about Paul, and let the fragrant air of the Cte dAzur drift through her. All night long the dog barked, fretful barks followed by a precarious silence in which she heard the wind, its soft breath hot as a lovers kiss on the plane tree leaves beyond her window. Hours after midnight, when she opened her eyes and the darkness that embraced her was more sensuous even than the weight of sleep, she remembered creeping into the shelter of Pauls arms.
16667.01 SecretPapersOfMadameOli4 4
She had never had far to creep, for he had liked to feel her near him. Their bed was a narrow double, and Pauls leg, thrust out to the side, often pinned her ankles. She would nudge his thigh, and, at her touch, he would sigh as if a forest maiden had entered his dreams and laid a hand where he most wished she would. His body would sometimes respond with a quick, hard thrust into the mattress and she would laugh aloud. She laughed now at the memory, so alive and tactile that she could almost feel the mattress jiggle and the cozy, bearish heat of him. Paul had died in Idaho, quickly, and with their son, Pierre, unhurt beside him. Lily pictured the two men sitting quietly in Pauls truck, tired, driving back to the ranch in the cold November rain and listening to the radio. Men at ease in each others company. Men with dark, curling, unkempt hair and thick black eyebrows that met in the middle over strong French noses.There were five ducks in the cooler behind the seat, three greenheads and a couple of teal, she remembered the officer had said.They had shot five ducks, only five, and were driving home to shower, shave, and meet her for lunch. They were to celebrate her fiftyfi rst birthday. Justine, almost eight then, was perched on the countertop beside her in the kitchen. She was cutting out cookie-dough figures, punching silver balls into the turkeys eyes and the pilgrims bellies. Lily was opening the letter that had been lying on the kitchen table since the mail had been delivered earlier that morning. She had been saving itsavoring the opening, the reading, the learning about Monsieur Duprs wonderful stay in the old stone house on the French Riviera. Another glorious September, she supposed. Another glamorous, unknown, unmentioned companion whose dark perfume would linger in the sheets, and whose black silk panties would lie forgotten under the red couch or in a clump of lavender, leading her to speculate, once again, about their renter. What books had he read this past September,

16667.01 SecretPapersOfMadameOli5 5
and which one would she findhad he left for heron the long wooden table? Which winesred or white? from Burgundy? Bordeaux?had he drunk with his meals, and how many bottles would there behis giftwhen she peeked next summer into the stone-dark cave? Six, no doubt. Punctilious and dependable, he always left six. Just as every July for the past eight years he had sent a check in advance of his stay. Just as every November he had written a thank-you note couched in superlatives. Monsieur Dupr was their only renter. She had just extracted his letter from the envelope when the doorbell rang. Ill go, Mom, Justine had said brightly as she jumped down from the counter, where her gamins legs, in shorts in November, had been swinging. The announcement of death had come just that swiftly: Justines legs swinging and her swinging blond hair as she hurried to the door, opened it, and was asked by the officer, who would later mention the ducks in the cooler, if her mother was in. Just a minute, Justine had replied, her smile widening. Shes right behind me, and Daddyll be home soon. Did you come for Moms birthday party? The policeman had taken off his hat and held it awkwardly in both hands. No, I didnt, hed said. Lily had come up behind her daughter and run her hand over Justines silky hair, so pale it appeared white in an unexpected shaft of sunlight that lanced through the open door. Can I help you? shed said, and then reading the discomfort in the mans eyes, had put her hands softly on her daughters shoulders and urged her around. Why dont you run upstairs to my bedroom and bring me my shawl? she said. Its getting chilly. But Justine had crumpled, with the shake of a muffled sob, against Lilys stomach, and Lily, who had begun to shiver, looked into the policemans eyes and mouthed Pauls name over Justines head. The man nodded.
16667.01 SecretPapersOfMadameOli6 6

7/10/08 1:41:10 PM

My son? she said out loud. Hes fine. She had bitten into her lower lip to stop its trembling and held her daughter close, and her face had turned into a steely mask and her eyes had seemed sheeted in some black throbbing substance that slowly closed out the light. Something between a groan and a wail tore through her body, ripping her heart to pieces without making a sound. She had stared at the officer and Justine had snuffled, Mom, whats wrong? Whens Daddy coming home? And Lily had lifted her daughters delicate chin, tilted it upward, and looked down into her sparkling blue, tearfi lled eyes. Everythings going to be all right, shed said. Everythings going to be all right, Justine. Everythings going to be all right. Lily crushed her knees against her chest and held on tight. This was pointless, all this suffering. Why revisit that moment? This was not why she had come here. This was not the story she intended to write, how her husband of twenty-six years had died after a modestly successful duck hunt, died instantly, at least she had that, and in peace beside their son. Or how Pierre had grabbed the wheel, grabbed it too late, and Paul, whose heart had already stopped beating, had died a second death, his neck broken, the truck smashed on the lava while the ducks in the cooler, dead as Paul, had sustained no injury. As if dying once were enough. Impatient for sleep, for the night to end and the day to begin, Lily closed her eyes. As always he was there, behind her eyelids, a smudge of recollection more heat and shimmer than an actual man. She breathed in his chin, his mouth, his nosele nez Pauls heroic, proud, improbable, quintessentially French proboscis, and she smiled. As she fell into sleep, the dog barked again.

16667.01 SecretPapersOfMadameOli7 7
eated at her desk, Lily shoved back the old portable Olivetti typewriter she had brought with her from Idaho. A few days had passed since the dog had kept her awake with its barking. Where was it now? Muzzled? Banished to the garage? Locked in a closet? Oras she preferred to believetucked cozily under its owners duvet? Her writing had gone smoothly. The sea was her confidante. It smiled as the dawn raised a blush to its skin, a glass-calm surface that suddenly, between the moment Lily looked down at the page where she had been working and then up, sported a series of wrinkles. How quickly the wind came up in the mornings, as if pulled on a leash by the sun. With more ease than she had expected, a routine was emerging. She wrote for an hour before it grew light, drank a small cup of coffee, then jogged down the hill through the village and along the beach, until she turned back to swim. There were few people about so early in the morning: an old man sweeping the streets; a merchant, yawning, bringing in boxes of peaches and courgettes; lights glowing in the bakers window; birds chitchatting above the sidewalks, before the tourists and the heat drove them into the oak woods and the oleander hedges. This was her favorite time of day; she fetched warm croissants and a baguette as she swung back through town, then trudged up the hill, ate a leisurely breakfast, and continued to write. From time to time she lost her concentration.The sea changed its moods and coloration as often as a French coquette: now the
16667.01 SecretPapersOfMadameOli8 8
sun slapped its cheeks with a brassy, gold-flecked rouge; now scudding clouds lay melancholy bands across its forehead. The ravishing sight of water and sky, so vibrant a blue against orange tiled roofs and dusky green hillsides, offered her solace, pushed her to forget. She was easily distracted, enthralled by the view. Madame! A mans voice bellowing up at Lily startled her out of her reverie. Madame Je suis ici. Im here, she called out. Up here. Im coming right down. She glanced quickly in the mirror, trying to pat her billows of red hair into shapehad she even bothered to comb it that morning? When the writing and memories took hold of her, she forgot to drink and eat, or brush her teeth. The day before she had bundled her hair on top of her head and jabbed a chopstick through it.That had worked nicelybut now she saw that she had dribbled coffee down the front of her shirt. What an image of middle-aged disarray and neglect she presented! Hurrying onto the porch and leaning over the railing, she caught sight of a large, rough-looking man wearing bright blue work pants and a blue jacket. He smiled up at her from the terrace. Madame Crisp? he said. You must be Yves Lebrun. You have come to fix the roof over the kitchen, she replied in French, for she had been forewarned that beyond the standard classroom phrasesgood morning, good-bye, and what is your name?he had mastered no English. Moderating her pace to appear less harried, she descended the staircase. Monsieur Lebrun. Enchante de faire votre connaissance. Delighted to make your acquaintance, she said, extending her hand. Yves shook it. Enchant, Madame. Then, when he had let go of her hand, he asked, You are called Lily Crisp? Yes, that is true. Crisp. This is a name I have not heard before. It is different, isnt it?

16667.01 SecretPapersOfMadameOli9 9
Its British. My husbands great-grandfather came from England. Its an old name. Then you must have much history, he said. Oh, yes, Lily thought, I have too much history. That is precisely the problem. That is why I have come here. But she said to Yves, Dont we all have history.? He gave a small laugh, shrugged, and indicated the roof. It is time to begin my work, he admitted, before your kitchen floats down to the sea when it rains. If it rains. I wonder. the dryness, it is not good for the trees and the hills. And nowhe pointed at a distant hillside thick with its dusky canopy of cork oaksthey will build a golf course on the mountainside. Over there. Can you believe it? How will they keep the golf balls from rolling into the sea? And where will they find enough water for the grass? Shaking his head and muttering to himself, he turned and walked brisklynot in the comfortable shamble Lily had expected from so hefty and loquacious a handyman t o his truck to fetch his tools and a ladder. She called out, Ill be upstairs in my room, writing, if you have any questions. She saw him raise his handhe had heard her; and so she climbed back up the stairs, smiling inwardly as she seated herself at the old chestnut table with cabriole legs that she used as her desk. She reviewed their conversation. What was it Yves had said? Crispit is different, isnt it? Lily chuckled, recalling Esther, her outspoken older sister. For Gods sake, Lil, why did you marry Paul without telling anybody? You barely know him! And why change your name? Lily Fern is a great name. Lily Crisp sounds like the star of a pornographic movie. Or a cookie. Lilys neighbors were in residence. This she knew from the music that wafted up from the pink house with the lime green shutters. She enjoyed the Telemann and the Mozart in the morn-
16667.01 SecretPapersOfMadameOli10 10

7/10/08 1:41:11 PM

ings when she wrote, although not late at night when she had gone to bed and the Belgian couple had just sat down to dinner. The house was young and enchanting, as were its owners. From her hillside perch, she could peer down upon them: the woman listened to Debussy while reading in her garden; the man grew tomatoes and aubergines, and had planted lemon, plum, and fig trees on the sun-drenched slope between Lilys house and his. Lilys house. La Pierre Rouge. Her house. Not Pauls. Not his mothers or his grandfathers, but Lilys. By a twist of fate, which the villagers doubtlessly found amusing, this house that was French down to the cold stone core of its bones now belonged to an American, to a Midwesterner. La Pierre Rouge was set high on a hillside, quite alone, up a steep and narrow, winding dirt road that was deeply rutted. During his lifetime Pauls grandfather must have kept the road open manually with a pickax and shovel, filling the potholes and ruts with gravel and chipping away at the crystalline bedrock that projected into the broken track. Around the final bend, from the forest of cork oaks and pines, the house rose up squarely, ages carred and handsome, built of rough-chiseled blocks of local stone quarried from the hillsides. Shiny, black-painted shutters flanked windows and doors that opened wide to let in the heady fragrance of early-summer blooms and the delicate southern light. Inside, the patina of dark wood set against the brilliant colors of patterned cloth warmed the cold stone-walled interior. There were huge wooden beams overhead, red tiled floors with threadbare Persian carpets, and a crush of comfortably upholstered peasant furniture, some of which, like the house itself, was over a century old. According to family legend, Georges Lafond had brought his young Italian wife in a wooden cart, pulled by a plump and frisky pony, to this rocky plot of ground, where he had built her a house. She had carried with her in the cart a massive, squared-

16667.01 SecretPapersOfMadameOli11 11
o ff chunk of red limestone from the village in the Dolomites where she had been born and had lived until reaching puberty. A bride of two weeks, she was fourteen years old and still sore between her legs. Her husband, smitten with the ripeness of her young buttocks and peach-scented breasts, had vowed to set the block of red stone plumb over the center of the main doorway, but she had said no, and had refused as well its symmetrical placement over or under a window, or midway across the rugged, golden gray wall that faced the sea. No marriage is perfect, nor do I seek perfection, she had told him, remembering the muscular but by no means satisfying tussle in their conjugal bed the first night of their union. The red stone must be mortared willy-nilly into the wall so as not to show pride before the eyes of God. Georges had obeyed her wish, and in the course of a long and satisfactory marriage she had borne him nine children, five of whom had survived into adulthood. The oldest, Grard, a stonemason like his father, when he died at the age of ninety-one, had bequeathed the house of the red stoneLa Pierre Rougeto Justine Lafond, his surviving child and only daughter. Upon Justines death, the house had passed to her son, Paul. And now it was Lilys, and as the dawn grew brighter and Haydens trumpets tooted a fanfare of joyous greetingbonjour to the birds and the cork oaks and the flotilla of clouds draped softly, like jellyfish trailing streamers, and to Lily arrested in her writing and reflecting on the proud-nosed clan who had built this houseshe blessed her deceased mother-in-law, who had never coveted jewels or fancy clothes but had believed in land. The summer after her father died, Justine Lafond had returned to France, having sailed away over forty years earlier. She had ordered new plumbing, tile baths, a washing machine, doublew ide French doors opening onto the terrace, fans for the upstairs ceilings, and a new firm mattress in the master bedroom.
16667.01 SecretPapersOfMadameOli12 12
(The old one, which had folded together in the middle like a fallen souffl, she had given to the woman who beat the Persian carpets and swept the floor.) She had also bought up the entire hillside above, on either side of, and below La Pierre Rouge, down to the narrow strip of rocky earth where the Belgian couple would later plant their orchard. It was this buffer of land that now provided Lily with the solitude and the setting, if not the consummate silence, she required for her writing.

16667.01 SecretPapersOfMadameOli13 13

c o n v e r s a t i o n

g u i d e
A CONVERSATION WITH ANNIE VANDERBILT
Q. This is your first novel. What inspired it in particular and why did you wait to write your first novel in, let us say, the second half of your life rather than the first? A. Actually, I wrote my first novel twenty-five years ago. In 1983, just after I finished it (or thought I had finished itin retrospect, it was a first draft and needed many rewrites or a quick trip to the shredder), I fell off a cliff. That accident stopped my writing life cold; all my creative energies went toward healing my foot and being able to walk again. Unfortunately (or fortunately), the only copy of the manuscript burned up along with our house in 2005, so I wont be doing any of those revisions. I then wrote a book, as yet unpublished, about bicycling through Japan with my husband, Bill, interweaving our adventures on wheels with the tale of my fall from the cliff, my survival and recovery. A good friend and freelance editor worked with me on this book and it was through her critical eye and heart-stopping honesty (she never hesitated to write Yuck, Trite, You can do better, or This makes me swoon) that I honed my skills. When I was fifty-one (one year younger
16667.01 SecretPapersOfMadameOli281 281

7/10/08 1:42:06 PM

than Lily), I finished the Japan book and immediately began writing The Secret Papers of Madame Olivetti. It took ten years. I was leading a three-part life at the time and trying to achieve some sort of balance. I was writing a novel because Lily Crisp just popped out on the page one day, and then Paul and Victor and finally Yves and all the others, and it gave me such pleasure to tell their story and spend time with them. During those same years my sister and I were caregiving an elderly aunt and my parents and seeing them, one after the other, into death. And finally, I was trying to keep some fun and spice in my marriage by joining my husband, Bill, on some of the backcountry and overseas adventures we have always enjoyed together. So I would be partway through a draft of the book and then there would be a health-care crisis with my mother or a monthlong expedition to some mountain in Bolivia or jungle river in Honduras. By the time I returned I would pick up the thread of plot and characters and find that I had changed, that I had new insights from my adventures or caregiving, so the book changed, too. It was never a smooth, clean reentry; I moved backward before I could move forward again. All this took time, none of which I regret, because when you write a novel in your fifties, you have more perspective on life (you hope) and can let the voluptuous side of living and loving take center stage. Q.What did you hope to achieve in writing The Secret Papers of Madame Olivetti? A. When I began working on this book, I wanted to take the readers senses and immerse them in the landscapes in which

16667.01 SecretPapersOfMadameOli282 282
the story takes place: lakeside Wisconsin; a French village on the Cte dAzur; the rain forest in Chiapas, Mexico; the stark Idaho hills. I saw the scenes, smelled them, and felt them, and then tried to paint those sensations in words. Within this framework I envisaged a plot that was densely layered, moving fluidly between past and present, so that a delicious soup of intrigue, lushness, passion, disaster, humor, and quirkiness would enhance the background flavors of my settings. Longt erm married love, romantic love, sexual love, parental and sibling love, the loss of a loved oneall these interest me and went into the mix. I created Lily Crisp, a strong and yet vulnerable woman who would wind her way through this often messy labyrinth of love and death and emerge, in her own imperfect way, with a clearer understanding of the fragility and resilience of human nature. Q. As you mentioned, the novel has several distinct settings. Are you as well traveled as your knowledge of all these places would suggest? Do you have a personal connection to any of these places? A. Bill and I started going on adventures from the day we were married.We began in the Peace Corps in India in 1968, then backpacked to Everest Base Camp in Nepal in 1972 and managed after that to sandwich wilderness and overseas adventures into our lives for the next thirty-five years. I have a personal connection to all the places about which I write. I grew up in Wisconsin and spent my summers in a rustic two-room pine cabin that belonged to my aunt. Every day I swam across the lake towing the canoe behind me so no motorboats would run me down. And yes, our neighbor
16667.01 SecretPapersOfMadameOli283 283

7/10/08 1:42:07 PM

scattered her husbands ashes in the lake and my sister and I would snorkel the shoreline in a state of titillation and horror. Ive lived in Idaho for the past twenty-five years and know the dry sagebrush hills and open-sky landscape from hiking and biking the trails and back roads and waking up every morning to that clear, high-altitude sunshine. As for France, my husband and I have bicycled across the south of France three times and, in 1980, lived for three months in an old stone house overlooking the sea. The village and the warm waters of the Mediterranean lay below us, through the cork oaks. Finally, Mexico. We spent two winters living and working in Chiapas, in San Cristbal de las Casas. Every month or so we would drive to the rain forest and camp beside the hut of a Lacandn friend. Often we have walked to the cascada, slipped through the sheet of water, and stood with our backs against the cool rock wall where Victor and Lily exchange their first kiss. These are my experiences. But in The Secret Papers of Madame Olivetti most of the specifics and all of the characters and what happens in these locations are imaginary. I had great fodder for the mill and loved dragging aromas, landscapes, birdcallsyou name it, I dragged itout of my memory. Q. The typewriter on which Lily writes her memoirs becomes a sort of character in the novel, and in fact the novel is titled after her.What inspired Madame Olivetti? A. When I wrote my first book (the one that burned up in the house fire), I typed part of it on an Olivetti Lettera 32 Portable that my mother sent to the house in France where I

16667.01 SecretPapersOfMadameOli284 284
was writing at the time. It was laborious work, lots of thunking and clicking of the keys, lots of mistakes, lots of cut and paste. But I became quite attached not only to the slow process of creating on a typewriter but to the machine itself. It had character and tooth. I lost my Olivetti in the fire, but it was alive and well and sleeping in the closet all those years of working on this book. Q.Why is the cat Alonso in the story? A. Alonso mirrors Lily. Both are sensual creatures who enjoy their meals. Both fiercely guard their independence but love deeply. Both relish their solitude but welcome company on a selective basis. Both wander off and misbehave but have gentle, loyal hearts. Plusany woman who can have a close and satisfying relationship with a cat and a typewriter is the kind of woman I would want to know and write about. Q. I found Lily Crisp to be a complex and fascinating womanand one of the sexiest female protagonists Ive come across in some time. She is very much a physical being, completely comfortable in her own skin. There are also times when she seems fearlesswilling to commit herself wholeheartedly to relationships with men she loves, even when she hasnt known them for very long. What about her particularly interested youor is she based on you? A. I certainly dont have Lilys bizarre way of dealing with loss through self-violence of one sort or another, nor do I see my life as a muddle. I would have told Paul the truth and risked losing him rather than hold it all inside. Lily keeps her
16667.01 SecretPapersOfMadameOli285 285
secrets. I wear my heart on my sleeve. But both of us savor good food, believe in passion and romance in life, love to travel, and can laugh at ourselves. I wanted to write about a woman who was comfortable with her sexuality, a naturally erotic but not neurotic being, who could love deeply and was a good mother and wife, sister and daughter, but who had a quirky, off-the-cuff, solitary side to her. I love the complexities and contradictions in people: hence, Lilys combination of fearlessness and vulnerability, her ability to jump into experience, sexual or otherwise, and then take responsibility for her misbehaviors and misjudgments. I was interested in a woman who is earthy and grounded but who makes mistakes and at times is her own worst enemy.Yet with her lusty sense of humor, her recognition of her own inadequacies without whining about them, and her ability to forgive both others and herself, she always manages to get back on keel and sail onward. Q. In her early fifties, two years after her husbands death, Lily is taking some time to reassess her life. In your experience, do most women go through a period of reassessment at some point in their lives? A. I think that most women, as well as most men, go through periods of reassessment in their lives. Many of us require time alone to do this, although most of us dont have the luxury of retiring to France with our typewriters (and a sexy, empathetic handyman) for five weeks. Especially in our fifties, when the future is still uphill but the downhill stretch is alarmingly visiblenot over but on the horizonmany

16667.01 SecretPapersOfMadameOli286 286
of my friends and I have felt the need to step back and gain perspective on our lives and priorities. Sometimes the only way we can chart where were headed, or might choose to head, is to take a look at where weve been, as Lily does in typing out her memories and recounting them to Yves. Its so easy to zip along in life and suddenly, since time moves exponentially faster with age, youre not fifty but seventy and you havent stopped once to ask, Who the hell am I, and is this what I want to be doing? Do I still want to drag out the same old whips and beat myself up because I made some mistake or misspoke? Do I still want to keep my pleaser tag lit up in neon and hanging around my neck? Do I still want to say no, no, instead of ripping off my clothes and diving naked into the water? Q. Im curious about your reading habits. Do you find the time to read? Have you ever been part of a book club or reading group? Is there a list of books that you feel have particularly influenced your life or helped shape you as a writer? A. I have always read voraciously. For me, reading is one of lifes great pleasures, along with fine dark chocolate devoured on a daily basis and walking on a beach or in the mountains or even down the street and back. I have been in the same book club, called Qui Legit, for the past fifteen years.We read books that have stood the test of time, including classics as well as recent novels. We allow gossip with soup before the discussion begins but then try to keep to the book. Since we have no leader, we each try to read critically and guide the discussion into interesting channels. Over the years we have
16667.01 SecretPapersOfMadameOli287 287
bonded in a very special way through discussing books. We now have a shared memory of 180 books we have read and talked about. The books that have resonated for me are The Lover by Marguerite Duras, because of the lush, atmospheric landscape laced with the eroticism of the love affair; The Leopard by G. di Lampedusa, because of the painterly sensual quality of his writing; The English Patient by Michael Ondaatje, because of the remarkable sense of place, both in the ruined villa and the desert, and the intensity and poignancy of the love stories; and Kalimantaan by C.S. Godshalk, because of the humid, dripping, heated sensuous jungle atmosphere and the strange turnings of love. Sound familiar?
16667.01 SecretPapersOfMadameOli288 288

1 This approach is similar to the one adopted in Regalia and Rios-Rull (2001) to investigate the causal relationship between the narrowing of the gender gap and the increase in single female households.
of the decrease in the gender wage gap between the 1970s and the 1990s. This result is consistent with the estimates obtained by ONeill and Polachek (1993). An alternative explanation for the increase in womens average work hours that is related to changes in the wage structure is the decline in the gender wage gap (Jones et al., 2003). I also use my model to evaluate the consequences of a 13% decline in the gender wage gap, holding returns to experience constant. While I nd that this change can account for 18% of the actual increase in lifetime hours of work for married women, it cannot account for the large increase observed for married women with young children. This result has an intuitive explanation. Higher relative wages by gender should cause all women to work more, irrespective of their age. On the contrary, the change in the returns to labor market experience differentially affects women of childbearing age because it makes labor market interruptions early in the life cycle relatively more costly. This paper contributes to a recent literature that studies the causes of the increased labor market participation of US women over the past century. One set of explanations stresses the importance of technological progress in the consumer durables sector and in medicine. According to Greenwood et al. (2005), the diffusion of new household technologies allowed women to decrease the time spent in home production and released time for market work. Goldin and Katz (2002) point out that the availability and diffusion of more reliable contraceptive methods, such as the pill, allowed women to plan their fertility and the timing of births, making it easier for them to plan a career. Other developments are the shift from manufacturing to services, which employ more women in jobs such as clerical work and sales (Goldin, 1990), changes in the wage structure that favored women (Pencavel, 1998, Jones et al., 2003), changes in the cost of children relative to lifetime earnings (Attanasio et al., 2004), and female-biased technological progress, which increased the relative attractiveness of working in the marketplace (Galor and Weil, 1996). Finally, the rise in the female labor force participation can also be explained in terms of the change in societys attitudes towards the role that women should play in the home and in the market place. On this topic, Fernndez et al. (2004) examine the role of preferences formation within the family. Since the focus of the analysis is the change in life-cycle behavior of married women with children, I abstract from modeling fertility and marriage decisions.2 The study of the effects of a changing wage structure on these decisions is the focus of a recent paper by Caucutt et al. (2002). They use a dynamic general equilibrium model of family formation and investment in children to study the determinants of womens timing of births and labor supply decisions. They nd that the rate of return to labor market experience is an important factor in the decision of women to delay child birth through its impact on wage growth. Their results are complementary to and reinforce the results reported in this paper. The idea that the link between labor market participation and wages is important for understanding gender differences in earnings growth dates to Weiss and Gronau (1981), who identify the basis of this link in the process of human capital accumulation that occurs on-the-job. Subsequent papers estimate structural models of married womens labor force participation decisions in the presence of learning-by-doing human capital accumulation. Most notably, Eckstein and Wolpin (1989) construct and estimate one of the rst dynamic models of married womens labor force participation decision in the presence of endogenous wage determination. Altug and Miller (1998) use a different methodology to estimate a model that incorporates both labor force partic2 For a survey of life-cycle models of fertility, see Hotz et al. (1997). More recently Greenwood et al. (2003) built an overlapping generation model that incorporates both marriage and fertility decisions.

ipation and labor supply decisions in the presence of learning-by-doing and habit persistence in the agents preferences for leisure.3 This paper uses insights from this literature as the building blocks of its theoretical framework. The paper is organized as follows. Section 2 documents the dynamics in hours of work by married women. I present the life-cycle model in Section 3. Section 4 describes the estimation of the learning-by-doing human capital production function that is used in the quantitative exercise. Section 5 discusses the calibration strategy. The results of the exercise and the sensitivity analysis are presented in Section 6. Finally, I provide some concluding remarks in Section 7. 2. Changing patterns in womens market work One of the striking changes in the labor market over the last three decades has been the dramatic increase in hours worked for women. In particular, average hours worked per adult female increased by 43% between 1970 and 1990. At the same time, average hours worked per adult male remained essentially constant (they increased by 0.43%) between the two periods. Over this period, several characteristics of the female population changed substantially, such as educational attainment, fertility, marital status and out-of-wedlock childbearing. These facts have been extensively documented in the empirical literature.4 The purpose of this section is to document that a key component of the increase in womens hours worked between 1970 and 1990 is the drastic change in the behavior of married women with young children. The extent of this change is reected in the change in cross-sectional age proles of hours worked for married women between 1970 and 1990. Figure 1 uses Census data from Mc Grattan and Rogerson (1998) to plot the age proles of weekly hours worked per person across decades. As Fig. 1 displays, in the 1970s, the age proles of hours worked per person are double-peaked for married women. Women work more hours at the beginning of their adult life, then temporarily withdraw from the labor market while they have children, and eventually engage again in market activities as the children grow older. This is the typical shape of womens proles, as rst pointed out in the work on womens earning and labor force participation by Mincer (1974), and Mincer and Polachek (1974). As shown in Fig. 1, the shape of womens proles is quite different in the 1990s. In particular, the average number of hours worked per woman increase for every age group, and this increase is largest for 25 to 44 year old women. As a consequence, the 1990 prole has a single-peaked shape, which typically characterizes men.5 The comparison between hours proles for different demographic groups further clarify the extent of the change. To this effect, Fig. 2 compares Census data in 1970, panel A, and 1990, panel B, for married men and women, married mothers, and single women. Married women and married mothers of young children show a behavioral prole different from that of men and single women in the 1970s. In the 1990s, the gap between average number of hours worked by

3 In both models, fertility choices are taken as given. Francesconi (2002) extend the Eckstein and Wolpin framework to incorporate fertility decisions. See Blundell and MaCurdy (1999) for an extensive review of the literature on dynamic models of labor supply. 4 See Blau (1998) for a thorough account of the trends in labor participation and wages of US women between 1970 and 1995. 5 Note that the data on average hours worked per person include changes in both the extensive and the intensive margin. The age proles of labor force participation are almost at (slightly decreasing) in the 1970s, but they are concave in the 1990s. The age proles for average hours worked per worker (the intensive margin) are similar to the ones per person: more women are working in the 1990s, and they are working more hours conditional on their labor force participation.
Fig. 1. Weekly average hours worked per married woman.
married women and by married men (or single women) decreases. The largest increase in work hours is experienced by 2544 year old mothers with children of preschool age, with the average number of hours worked per person more than doubling. At the same time, for the same age group, average hours worked by single women and married men remained constant or decreased slightly for every age group. As a consequence, married womens and married mothers age proles have the same shape as mens and single womens proles in 1990.6 Between 1970 and 1990 the demographic distribution of young women changed substantially. In particular, the fraction of married women with children of preschool age declined, whereas the fraction of single women in the population doubled.7 Simple calculations show that the change in the demographic composition of the female population between 25 and 44 years of age can account for about 24% of the total increase in womens hours of market work. The increase in hours of market work by married women with children explains 70% of the residual. At the same time, the increase in hours of work by single women can account for only about 3% of the change.8 This fact reinforces the observation that the increase in average hours of work by young women is due mainly to a change in the working behavior of married women with children. Although the facts presented in this section are based on cross-sectional data, the same pattern is observed for life-cycle histories. Goldin (1989) uses historical data to show that womens labor participation at different stages in the life cycle started to change with the 1880s to 1910s

12 Here I consider a unitary framework where the two, symmetric, members of the household maximize joint utility under the assumption of income pooling. The main results of the analysis would not change if we were to consider an efcient non-unitary model of household decision making. 13 Hotz and Miller (1988) make similar assumptions for the childcare production function. Also in Weiss and Gronau (1981) as children grow older womens productivity at home decreases.
Household decision problem Given prices g and R and the initial stock of human capital a household solves the following decision problem:
t1 U (ct , nmt , nf t ) + V (xt )

subject to:

(ct + st ) (1 + R)t1

g{m,f } t=1

gt ngt (1 + R)t1
gt+1 = G(gt , ngt ) ngt +

g {f, m}

+ hgt = 1 g {f, m}
a0 = aN +1 = 0 where gt = g gt is the hourly wage of a worker of gender g at time t, g is the genderspecic efciency wage, R is the rental rate of physical capital, and the adult life-span N is equal to 4 (10-year) periods. I assume that married couples begin their life with no asset and they consume everything they have during the last period of their life (no bequest). 3.1. Functional forms This section describes the functional forms used in the simulation exercises. Preferences As it is standard in the literature on labor supply, I assume that preferences are separable in consumption and leisure and also across time. In particular, I assume period utility functions of the form: (nf t + hf t )f (nmt + hmt )m U (ct , nmt , nf t , hmt , hf t ) = ln ct A + m f V (xt ) = b ln xt where A > 0 and g > 1, g {f, m}. To be noticed, in this model g = 1/(g 1) cannot have the same interpretation as in a model with exogenous wage proles since here the intertemporal elasticity of substitution changes over the life cycle. As mentioned above xt represents children welfare. The parameter b represents the weight that parents give to the enjoyment of children welfare. Childcare production The literature provides little guidance in modeling the childcare production function. There is no evidence on the shape of this function, although the degree of substitutability between maternal time and market produced goods and services should play an important role.14 I assume a constant elasticity of substitution production function, since it provides a parsimonious representation of childcare production. I consider the following functional form: xt = t hf t + (1 t )st

15 We can easily consider a more general childcare production function that also include fatherly time. Although we would need to make assumptions on the degree of substitutability between motherly and fatherly time. In this case the function would generalize to:

xt = t h t + h mt f

+ (1 t )st
where 1/(1 ) is the elasticity of substitution between motherly and fatherly time. 16 Hotz and Miller (1988) show that the childcare intensity in motherly time declines as the children ages, a = t1 a. t The assumption that = 0 cannot be rejected. is signicantly different from 1. 17 These studies show that the increase in return to experience and the increase in actual experience for women explain a large portion of the decreasing gender wage gap over this period. 18 The importance of past labor supply decisions and wages as determinants of current wage levels has been documented in the literature. Using PSID data, Altug and Miller (1998) nd a signicant learning-by-doing effect for married women. Imai and Keane (2004) document a similar nding for men based on NLSY data.
human capital technology parameters, and hence, of the returns to labor market experience, for the two decades. I consider the following specication for the human capital production technology: gt+1 = (1 g )gt + g ngtg gt
where gt is the human capital stock of an individual of gender g at time t and ngt is hours worked by an individual of gender g at time t, g {f, m}. Human capital depreciates at a constant rate. According to this specication, the stock of human capital at time t + 1 is given by the previous periods human capital stock minus depreciation, plus an increasing function both of the individuals stock of human capital and of the total number of hours worked in the previous period. This functional form implies that current human capital is a sufcient statistic for an individuals past work history.19 The marginal rate of return to extra hours of work is given by the rst derivative of this function, and is proportional to. The parameter can also be generalized to include a pure age-effect. This assumption would capture the fact that incentives to accumulate human capital may change as individuals age. The parameters of the production function are estimated separately by gender for the 1970s and the 1990s. Note that I assume the human capital technology to be linear in previous period human capital stock, gt. Under this assumption, the rate of growth of human capital does not depend on the level of human capital stock in the previous period.20 As discussed by Weiss and Gronau (1981), this assumption makes it possible to separate the effect of labor supply on the level and the rate of growth of human capital accumulation. An analogous human capital production function is estimated by Imai and Keane (2004) in the context of a structural dynamic model that allows for both human capital accumulation and savings.21 Their estimates are based on a sample of primeage men during the 1980s and the early 1990s. Unfortunately, given the focus of my analysis, their estimates cannot be used in my calibration exercise. 4.1. Estimation methodology Since the human capital stock is not observed, Eq. () cannot be estimated directly. However, we do observe wages. As is typical in the literature on human capital, wages can be dened as the product of the individual human capital stock times the rental rate of human capital in the economy (i.e. the efciency wage). That is, the wage rate for the ith individual of gender g is given by: igt = igt gt , where the rental rates, gt , differ by gender. In what follows the subscript g is dropped for expositional purposes. However, the parameters of the production function are allowed to vary by gender and across decades in the estimation. The empirical counterpart of Eq. () is obtained by multiplying both terms of the equation by t+1 , dividing by t , substituting it = it t , taking logs and rearranging. We obtain: t+1 + it+1 () ln it+1 ln it = ln (1 ) + nit + ln t

Zit t Zit+1 t+1

E[it+1 | it = 1, it+1 = 1] =
2 (Zit t , Zit+1 t+1 , ) + (Zit+1 t+1 )

Zit+1 t+1 Zit t

2 (Zit t , Zit+1 t+1 , )
where and denote, respectively, the density and cumulative distribution functions of a standard normal distribution and 2 denotes the bivariate normal distribution. I will refer to the two
22 This amounts to assuming that the stochastic disturbances {v }T ij j =1 follow an AR(1) process. Alternatively, one could use semi-parametric procedures to correct for the bias (see Vella, 1998 for a survey.) This exercise, although interesting, is outside the scope of this paper.
correction terms as to 1 (Zit ; t , ) and 2 (Zit+1 ; t+1 , ), respectively. They are the bivariate equivalent of the inverse Mills ratio in a standard Heckman two-step selection correction model. I then follow the two-step procedure to adjust the estimates of Eq. () for the sample selection bias. In the rst step I estimate the two correction terms based on a probit specication.23 The resulting terms, 1 (Zit ; t , ) and 2 (Zit+1 ; t+1 , ), are added to Eq. () to obtain: ln it+1 ln it = ln (1 ) + nit + ln
t+1 + 1 (Zit ; t , ) t ()
+ 2 (Zit+1 ; t+1 , ) + it+1
where it+1 are i.i.d. standardized Normal stochastic disturbances. In the second step equation () is estimated by least squares. I estimate () by gender and by decade, the 1970s and the 1990s, using PSID data. The 1970s samples, one for men and one for women, include individuals from the 1970 to 1977 waves of the PSID. The 1990s samples include individuals from the 1990 to 1997 waves. The four samples include men and women aged 20 to 60. Summary statistics are reported in the Appendix. Real hourly wages, it , are obtained by dividing annual real earnings by annual hours worked, nit.24 The year-to-year variation in the rates of return to human capital, t , is captured by introducing time dummies in (). 4.2. Estimation results and the change in returns to experience In this section I discuss the estimation results and document the increase in returns to labor market experience between the 1970s and the 1990s. I estimate Eq. () under the assumption that the depreciation rate and the exponent are the same across genders and decades. Thus the variation in returns to experience, across genders and over time, will be captured by the estimated values for the parameter. The estimates must be constrained in order to generate numbers that are consistent with interior solutions for both mens and womens working hours in the calibration exercise.25 I also 2 specify () as a quadratic function of age: (ait ) = 0 + 1 ait + 2 ait.26 Both assumptions are needed to obtain realistic age-earning proles when I use the estimates of the human capital production function to calibrate the model. Table 1 presents the results when I set = 0.4 and = 0.2, and estimate the remaining parameters for the sample of full-time workers (which is dened as individuals who worked at least 1500 hours during the year). Imposing = 0.2 (a 20% depreciation rate for human capital) is consistent with the estimates reported in Imai and Keane (2004) for prime age men.

27 Given the earnings equation ln w = + X + X 2 + Z + , where X is year of experience and Z is it it it it it 2 it 3 it a vector of additional controls, the returns to labor market experience are given by: ln wit /Xit = 1 + 22 Xit. 28 The following transformation can be used to compare the estimated coefcients. Based on the Mincer specication for wit+1 the wage regression we can obtain: ln( w )/Pit+1 = 1 + 22 (Xit + Pit+1 ) where Xit represents accumulated it labor market experience at time t and Pit+1 is an indicator function that is equal to 1 if the individual worked full-
My estimates for mens returns to experience and for their growth over time are also consistent with those documented by Blau and Kahn (1997). According to their estimates, mens returns to full-time experience increased by 9% whereas womens returns more than doubled. Hence, as one would expect, the correction for non-random selection into the sample has a large impact on the estimates of the returns for female workers, whereas it has no effect on the estimated values for men. The results described in this section are robust to the choice of the values of the xed parameter and as well as to the choice of an alternative specication for the human capital production function. In all these cases I nd that women experienced a relatively larger increase in returns to labor market experience than men, with estimated returns ranging from 16 to 27%.29 5. Calibration The basic calibration strategy is to parametrize the model using data from the 1970s, and then ask how the models prediction change when we allow for changes in the returns to experience to the values estimated for the 1990s. The baseline economy, the 1970s, is calibrated to the PSID data in the following way. First, the parameters of the learning-by-doing human capital production function are set to their 1970s estimated values for men and women. The remaining parameters are chosen to match the 1970s work hours and hourly wage age proles for married women and married men. Given the parameters obtained for the baseline economy, I change the parameters of the human capital production function to their 1990s estimated values and study the models prediction in terms of age proles for hourly wages and hours worked. I also consider a second experiment where, given the baseline economy parameters, the female/male gender differential is changed to its 1990s values. Finally, I compute and compare the contribution of the relative change in returns to experience, and of the decline in the male/female wage gap, to the observed 1970s1990s change in the life-cycle pattern of work hours for married women and men. The model is parametrized to the annual hours worked and hourly wage age proles for married men and married women as in the PSID data for the 1970s. This is the subset of married individuals from the PSID samples used for the estimation of the human capital production function.30 The actual 1970s and 1990s age proles for average annual hours worked per person, and hourly wages are reported in Table A.2 in the appendix. Initial condition The initial levels of human capital, m0 and f 0 , are both set equal to one. Thus I assume that men and women in the 1970s do not differ in terms of their initial level of human capital. The female/male differential is captured by the market premium to female and male human capital, g. I also assume that households start their adult life with no assets.

Table 2 Model parameters Preferences m f A b 3.02 2.0.35 0.75 0.56, 0.23 Rental rates r f m a0 f 0 m0 5% 9.08 13.1 1

Initial conditions

Childcare production
34 In the micro literature the estimates of the elasticity of substitution for men range between 0.1 and 0.45 (see, for example, Altonji, 1986 and Pencavel, 1986). For women there is a wide array of estimates for the intertemporal elasticity of substitution ranging from negative values to large positive values (see Mroz, 1987, and Killingsworth and Heckman, 1986).
Panel A: Annual hours worked
Panel B: Hourly wages Fig. 3. The baseline economy and the data.
rate of time preferences is assumed to be equal to real interest rate. Moreover, the model predicts a childcare consumption share of total household income around 6% in the rst (10 year) period of childrens life and of 7.5 percent in the second period. On average the model predicts a childcare expenditure share of 6.77%. Women spend 32% of period 2 working hours taking care of their children. In the second period the fraction of maternal time devoted to childcare decreases to around 3.7% of total hours worked. 6.1.1. Experiment 1: Change in returns to experience In the following experiment the parameters of the human capital production function are changed to their 1990s estimated values for both men and women while keeping the remaining parameters constant to the baseline economy. I compare the predicted change in age proles for married men and women to the actual change observed in the data. The 1990s actual proles for work hours and hourly wages are reported in Table A.2 in the appendix. In this experiment the estimates of the human capital production function are set to the 1990s values reported in Table 2. They correspond to a female/male differential in the rates of return to experience equal to 0.91. That is, mens returns at age 20 increase by 6% with respect to the 1970s estimates presented in Table 2, whereas womens returns increase by 25%. This implies an 18% increase in the female/male returns to labor market experience (from 0.77 to 0.91). Note that since the estimates for 1 and 2 are very small this experiment basically amounts to changing a single parameter, 0 , for both men and women. The results are described in Fig. 4. Panel A displays the age proles for annual hours worked for the benchmark economy, the 1970s, as well as the predicted proles under this experiment, the 1990s. In all the graphs the dashed line represents 1990s men, the solid line 1970s men, the dotted line 1990s women, and the crossed solid line 1970s women. As panel A shows, given the parameters of the baseline economy, the increase in rates of return to experience alone generates the change from the single-peaked to the double-peaked work hour age prole for married women that this paper aims to explain. At the same time, consistent with the data, the model predicts only modest changes in the life-cycle prole of hours worked for men. This is mainly due to the fact that men are already working forty percent of their time in the baseline economy. For what concerns wages (see panel B), the change in proles generated by the model goes in the same direction as the actual change, although the 1990s proles do not perfectly match what is observed in the data. Both mens and womens proles are steeper in the 1990s than in the 1970s. Moreover, married womens age prole for hourly wages increases with age as compared to the decreasing prole of the baseline economy. These results show that changes in returns to experience alone can generate substantial changes in womens labor force participation and wages over the life cycle. Although the 1990s model has the same prediction as the baseline model for consumption proles, household consume more now since they have more income. Moreover, the model predicts a childcare consumption share of total household income around 7.3% in the rst (10-year) period of childrens life, and of 7.4% in the second period. The model generates a total childcare expenditure share of 7.38% in the 1990s. Overall the total childcare expenditure share is less then one percentage point (of total household income) higher in the 1990s than in the 1970s. The childcare expenditure share is higher in the rst period of childrens life in the 1990s as compared to the 1970s, whereas it is lower in the second period. The fraction of time women spend in childcare drops dramatically. Women spend only 6% of their period 2 working hours

Panel B: Hourly wages Fig. 4. Experiment 1change in returns to experience.
providing childcare (from a 32% in the 1970s). In the second period of childrens life the fraction of maternal time spent in childcare activities drops to 0.003% of total hours worked. 6.1.2. Experiment 2: Change in the pure gender wage gap In the second experiment, I change the rental rates of female and male human capital to their 1990s values, holding the returns to experience and the remaining parameters constant. In particular, I set f = 10.5 and m = 13.3, thus the pure gender wage differential, that is the ratio f /m , is now equal to 0.79. This is equivalent to an increase in the female/male ratio by 13%. Figure 5 presents the results for this experiment. As described in the previous section, panel A displays the predicted change of work hours age proles and panel B presents the predicted change for hourly wages. The change in the pure wage differential generates an increase in average hours worked per person for every age group but it cannot explain the change in the shape of the age prole. Moreover, although womens hourly wages increase, the wage age prole is still declining in age. In this case, married mens behavior does not change much between the 1970s and the 1990s. For this group, both the work hours and the earnings age prole are almost unaltered. Household consumption and savings proles are almost identical to the 1970s proles. The childcare expenditure share slightly increases. The model predicts a childcare consumption share of total household income around 6.4% in the rst (10-year) period of childrens life, and of 7.5% in the second period. The model generates a total childcare expenditure share of 6.96% in the 1990s as compared to 6.77% in the 1970s. The fraction of time women spend in childcare drops. Women spend only 19% of their period 2 working hours providing childcare (as compared to 32% in the 1970s). In the second period of childrens life the fraction of maternal time spent in childcare activities drops to 0.7% of total hours worked. Although in the simulation the gender-specic efciency wage are simply set in order to reproduce the pure female/male wage differential for young workers in the 1990s, the model generates an average female/male wage ratio over all age groups that closely matches the one observed in the data. To conclude, I also consider the joint effect of an increase in both the gender wage gap and the returns to experience to their 1990s value. In this case, the model generates larger average work hours than those observed in the data for every age group. This suggests that the two explanations explored in this paper interacts to generate a further increase in average hours worked by women over the life cycle. The intuitive explanation is that in this case not only womens learning-bydoing prole is steeper than in the 1970s, but also they are initially better off in terms of their human capital. Men slightly decrease their average hours worked. 6.2. Discussion The change in relative returns to experience has a strong impact on married womens hours worked age proles. On the contrary, at least under this specication, the decline in the gender wage gap seems to play a much smaller role in explaining the increase in womens hours of work. In order to assess quantitatively the contribution of these two competing explanations, I compare the percentage change in average hours worked per person observed in the data to the one generated by the model. Table 3 summarizes the results. For 20 to 59 year old women actual annual hours worked per person increased by 75.2% between the 1970s and the 1990s. The model generates a 72.4% increase in hours worked in the rst experiment and a 13.5% increase in the second experiment. Hence, the change in returns to

Panel B: Hourly wages Fig. 5. Experiment 2change in gender wage gap.
Table 3 Change in hours worked per married women: Data vs. Model (percentage values) 2029 Data Experiment 1: returns Experiment 2: gender gap 3.20.102 102.6 12.75.2 72.4 13.5
labor market experience can account for 96% of the increase in lifetime average hours worked per married woman. The change in the gender wage gap can account for only 18% of the increase. Moreover, in the rst experiment the model can account for 40% of the increase in average hours worked per person observed in the data for the age group 2029, and for approximately 95% of the increase observed for the age group 3039. For the age groups 4049 and 5059 the generated increase in average hours worked over-predicts the one observed in the data. On the other hand, the change in the gender wage gap accounts for only a small increase in hours worked in each age group and for 18% of the increase in average hours worked by 3039 year old married women the group whose change in work behavior determines the change in the shape of work hours age proles. The increase in relative returns to experience can generate a large part of the increase in hours worked for women in childbearing age (although it over-predicts the increase for older women). Most importantly, it can explain the change in the shape of life-cycle proles of hours worked for married women between 1970 and 1990from double-peaked to single-peaked. On the contrary, the increase in the gender wage gap can explain only a modest fraction of the increase in hours worked for every age group, and it cannot generate the changing shape of married womens age proles. This evidence suggests that although there are surely other factors that contributed to the change in labor force participation of married women with young children, the relative change in returns to experience seems to be a very important determinant of this change. The model also allows us to measure the contribution of the relative change in returns to experience to the decline in the gender wage gap observed in the data. According to the model, the increase in returns to experience alone accounts for about 42% of the increase in the female/male wage ratio found in the data. Since wage age proles are endogenous in this model, it is worthwhile to compare the female/male wage differential by age group predicted by the model for the 1990s (when only returns to experience are changed) with the values observed in the data. The results are summarized in Table 4. Entries in the table represent the actual and predicted female/male wage ratios. Average wages by gender are weighted by the fraction of hours worked by each group with respect to the total hours worked in the population. The model matches quite well the observed age-specic female/male wage differential. The model also makes predictions about childrens welfare. Under experiment I, childrens equilibrium lifetime well-being increases by 28% with respect to the baseline economy despite an eighty percent decline of the share of motherly time spent in childcare. The increase in family income allows a family to substitute motherly time with market-produced goods and services in the production of childcare. Under the second experiment, childrens welfare slightly increases (by 0.3%) even if motherly time spent in childcare drops by forty percent. Thus, childrens welfare does not decline as a consequence of the increase in hours worked on the market by mothers. Moreover, despite the extent to which a family substitutes parental time with market good and services, the childcare expenditure share increases only slightly in both experiments (up to 1%

Table 4 Gender wage differentials by age, actual and predicted proles Data 0.786 0.684 0.586 0.457 Model 0.764 0.611 0.501 0.447
under experiment I).35 The small increase is consistent with what is found in the PSID data when we compute the childcare expenditure share of family labor income for married household aged 3039 with young children. The model also generates values of the childcare expenditure share that are consistent with data from the 1970s and the 1990s.36 6.3. Single household I also study whether the observed change in returns to labor market experience can predict changes in the work hours age proles for single men and women that are consistent with the data. To this aim, given the 1970s estimates of the human capital production function, I parametrize the model to match the age proles for hours of work and hourly wages for single females and single males. As previously discussed, in 1970 single women displayed a work hours age prole similar to that of men. Moreover, lifetime average hours worked by single men and single women only slightly increased between 1970 and 1990. The experiments show that the model is capable of reproducing this feature of the data. In particular, this is true for a range of values of the intertemporal elasticity of substitution parameter (ranging from 0.84 to 5.6). The values of the intertemporal elasticity of substitution needed to match the 1970s data for single are slightly higher than those needed to match the behavior of married couples. The results for one of these experiments are presented in Fig. 6. Panel A presents the actual change of age proles of work hours for single men and women. Panel B presents the change predicted by the model. As in the previous experiments, the 1970s economy is parametrized in order to match the 1970s proles for single men and single women. The model can match the shape in the work hours age proles for both genders and the relative change in their 1990s proles, although the predicted increase in average hours worked per person is larger than what is observed in the data.
35 Attanasio et al. (2004) show that shifts in the cost of children relative to lifetime earnings are an important explanation for the change in labor force participation across cohorts. In the context of my model, even if child care costs do not change over time, we would observe a decline in the cost of children relative to lifetime earnings in the presence of increasing returns to labor market activities that increase hours worked by women. In addition to its direct impact on womens labor participation, a decline in the cost of childcare would reinforce the impact of the change in labor returns on womens labor supply when they have children. 36 Anderson and Levine (1999), using 19901993 SIPP data, nd that families with at least one child under 13 spend, on average, 7% of their family income in childcare, whereas families with at least one child under 6 spend on average 7.7% of their income. I nd similar values by computing the same shares using the PSID data for 1973 and 1993.

Panel A: Annual hours workeddata.
Panel B: Annual hours workedmodel. Fig. 6. Change in returns to experience: Single households.
6.4. Sensitivity analysis In this section I study the sensitivity of the results to changes in some of the parameters used in the calibration exercise.37 I perform the analysis by recalibrating the economy to the 1970s and then by changing the returns to labor market experience to their 1990s values. This allow to study whether the result (i.e. the strong contribution of the change in returns to labor market experience) is robust to the parameters chosen in the calibration exercise. In particular, I will study the robustness to changes in the real interest rate, the female/male wage ratio and the intertemporal elasticity of substitution for both men and women. In the literature38 , the female/male wage ratio for the 1970s ranges from 0.6 over all age groups to 0.764 when focusing on workers aged 20 to 24 and controlling for education, race and full-time labor market participation. Therefore, I recalibrate the economy setting f /m = 0.6 in the rst case, and f /m = 0.764 in the second case. Under the rst parametrization the model can account for about 66% of the observed increase in average hours worked per person observed in the data for the age group 3039. In the latter case the model accounts for 97.6% of the increase. Moreover, given the baseline economy discussed in Section 6.1, I show that in order for the increase in the female/male wage ratio to generate the change in the shape of the work hours age proles observed in the data one would need a 1.9 female/male wage ratio. That is, young women should be earning almost twice as much as men. In fact, when the rental rates to human capital for men and women are the same we still observe double-peaked proles for womens hours of market work. Hence, the increase in the female/male gender wage gap seems to play a secondary role in explaining the change in the working behavior of women over their life cycle. I also study the sensitivity of the results with respect to the real interest rate by setting the annual real interest rate r equal to 4 and 6% respectively. The model is not very sensitive to changes in the real interest rate (indeed the values of the preferences and childcare production parameters needed to match the 1970s economy change only slightly). In both cases, I obtain results roughly equivalent to the ones obtained when I set the yearly real interest rate, r, to 5%. For what concerns the intertemporal elasticity of substitution, in the micro literature the estimates of the elasticity for men range between 0.1 and 0.45 (see, for example, Altonji, 1986 and Pencavel, 1986). In the macro literature higher values of the elasticity are typically used to calibrate the model (for example, Prescott, 1986 uses 2 in his calibration exercise). More recently Imai and Keane (2004) obtain an elasticity estimate equal to 3.82 for prime-age men in the context of a dynamic labor supply model with learning-by-doing human capital accumulation. For women there is a wide array of estimates for the intertemporal elasticity of substitution ranging from negative values to large and positive values (see Mroz, 1987, and Killingsworth and Heckman, 1986). In general, there has been consensus that female labor supply wage elasticities are larger in absolute value than are mens. As a consequence, I perform two experiments. In the rst one I set both mens and womens elasticities to low values (0.3 and 0.35 respectively). In the second experiment I choose high elasticities (1.2 and 1.3 respectively). In both cases, the model generates a change in hours of market work over the life cycle of the same order of magnitude as the ones described previously. That is, the change in returns to labor market experience